<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Anup Khanvilkar on Medium]]></title>
        <description><![CDATA[Stories by Anup Khanvilkar on Medium]]></description>
        <link>https://medium.com/@anupkhanvilkar?source=rss-62bf8434b066------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*iRnEf9UVkQWeFAQUHqbNFw.jpeg</url>
            <title>Stories by Anup Khanvilkar on Medium</title>
            <link>https://medium.com/@anupkhanvilkar?source=rss-62bf8434b066------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 22:18:29 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@anupkhanvilkar/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Hospital That Taught an AI to Fight Ovarian Cancer — and Got 297% ROI]]></title>
            <link>https://anupkhanvilkar.medium.com/the-hospital-that-taught-an-ai-to-fight-ovarian-cancer-and-got-297-roi-fe1b4f01da4b?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/fe1b4f01da4b</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[healthcare-ai]]></category>
            <category><![CDATA[clinical-ai]]></category>
            <category><![CDATA[cancer]]></category>
            <category><![CDATA[ai-use-cases]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Wed, 13 May 2026 06:05:01 GMT</pubDate>
            <atom:updated>2026-05-13T06:05:01.131Z</atom:updated>
            <content:encoded><![CDATA[<p>By Anup Khanvilkar | Series : AI Case Files | Case # 1</p><p>Inside Sheba Medical Center’s clinical AI deployment: the technology, the numbers, and the lessons every enterprise AI leader needs to hear.</p><p>There is a number in oncology that haunts every physician who works in women’s health.</p><p><strong>Seventy percent.</strong></p><p>That is the proportion of ovarian cancer cases diagnosed at Stage III or Stage IV — the point at which the cancer has already spread beyond the ovary, metastasised into the abdomen, wrapped itself around organs, and made curative treatment a long shot at best.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/752/1*w_1ROHmyjQ7oPNpF7GXz6A.gif" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/783/1*nxVOk1tT4H6DDG2OY5bzsw.png" /></figure><p>The five-year survival rate at Stage I is 92%. By Stage III, it collapses to 29%. This is not a treatment failure — oncologists have powerful tools. The problem is <em>time</em>. The disease presents with vague, easily dismissed symptoms: bloating, fatigue, a sense of fullness. It is misdiagnosed as IBS, stress, perimenopause. By the time imaging arrives, the window has often closed.</p><p>This is the problem Sheba Medical Center — Israel’s largest hospital and one of the top 10 globally ranked medical institutions — decided to attack with artificial intelligence.</p><h3>Why This Problem, and Why Now</h3><p>Sheba treats over a million patients annually and is a rare institution: large enough to generate clinical data at scale, research-oriented enough to validate rigorously, and operationally mature enough to integrate new technology without chaos. Its oncology department had a specific and quantifiable problem — patients were arriving late, referral pathways were long, imaging interpretation was inconsistent, and treatment decisions required synthesising data volumes no single physician could fully hold in their head.</p><p>The clinical team had a hypothesis: if they could build a system that ingested all of this data simultaneously, flagged high-risk patients earlier, and surfaced evidence-based treatment recommendations in real time, they could meaningfully shift the stage-at-diagnosis distribution. They were right. But the path from hypothesis to 297% ROI was not linear, and the technical decisions they made along the way explain almost everything about why it worked.</p><h3>What Sheba Actually Built</h3><p>This is the section most enterprise case studies omit — or bury in impenetrable jargon. Let’s not do that.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/783/1*4QPMG32PFFPmq0eYCS52vg.png" /></figure><h3>The Data Foundation: Google Cloud BigQuery</h3><p>Everything starts with data — and Sheba’s challenge was representative of any large hospital: patient information sitting in silos. Imaging, EHR, pathology, genomics, and outpatient notes all generated data independently, in different formats, on different timescales. BigQuery became the central warehouse, unifying previously siloed datasets into a single analytical layer. This is not glamorous. It is foundational.</p><h3>The Modelling Layer: Google AutoML on Local Data</h3><p>Rather than purchasing a pre-trained clinical AI model and deploying it hospital-wide, Sheba used Google’s AutoML to train models directly on Sheba’s own patient population — real clinical outcomes, real treatment decisions, real survival data. This distinction is critical, and we will return to why when we discuss what went catastrophically wrong at a competing institution that took the opposite approach.</p><h3>The Safety Layer: Hippocratic AI</h3><p>Clinical AI faces a category of risk that consumer AI does not: a confident wrong answer can kill someone. Hippocratic AI is a safety-focused LLM built with explicit constraints around clinical recommendation guardrails and uncertainty disclosure. Integrating it meant that when Sheba’s system surfaced a treatment recommendation, a physician could see not just <em>what</em> the system recommended, but <em>how certain</em> it was — and the system flagged when human expertise should take over.</p><h3>The Multimodal Fusion Engine: Imagene AI</h3><p>Ovarian cancer diagnosis requires synthesising fundamentally different data types: imaging (CT, MRI, PET), histopathology slides, and multi-omic biomarkers. Imagene AI provided the multimodal fusion capability — aligning data modalities that exist at different spatial and temporal resolutions, with different noise profiles and uncertainty levels, into a unified risk assessment.</p><h3>The Genomic Intelligence Layer: NVIDIA Genomic Foundation Model</h3><p>Developed in partnership with Mount Sinai Health System, NVIDIA’s Genomic Foundation Model provided pre-trained genomic sequence representations that Sheba fine-tuned on their patient population — dramatically reducing the data volume required to train genomics-aware models. In oncology, labelled genomic datasets are expensive. Pre-trained foundations change the economics entirely.</p><h3>The Predictive Risk Engine: Medinsight.ai</h3><p>The final layer — patient risk stratification and clinical pathway prediction — integrated outputs from all upstream components to generate patient-level risk scores and flag individuals for expedited clinical review.</p><h3>The Business Case: Five Years of Real Numbers</h3><p>Enterprise AI projects are often evaluated on soft metrics. Sheba’s programme was evaluated on hard financial and clinical outcomes, which is why the ROI story is credible.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/902/1*xtnvVQ4fzVJQ6X_-p4a5Lw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/783/1*8g8HV2xpOzUl4hmg3EKPPA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/783/1*Ig82KWw6qH1PXLofQiRyog.png" /></figure><blockquote>“The NPV figure is the honest metric. At $3.1M over five years with a 10% hurdle rate, this is not a marginal investment. It is a demonstrably value-generating programme by any standard capital allocation framework.”</blockquote><h3>What Made It Work: Four Decisions That Separated Sheba From the Field</h3><ol><li><strong>They built on real patient data, not synthetic proxies.</strong> Sheba trained their models on Sheba’s patients, with Sheba’s outcomes, in Sheba’s clinical context. This cannot be emphasised enough — and we will return to exactly what goes wrong when organisations don’t do this.</li></ol><p>2. <strong>They embedded uncertainty disclosure from Day 1.</strong> Hippocratic AI’s integration was a design requirement, not an afterthought. Clinicians always knew how confident the system was, which built trust incrementally rather than destroying it with a single high-profile wrong recommendation.</p><p>3. <strong>They insisted on multimodal input.</strong> A cancer diagnosis is not a single-channel problem. Sheba’s insistence on fusing imaging, pathology, and genomics — rather than optimising for one modality — is directly reflected in the clinical accuracy metrics.</p><p>4. <strong>They planned for a negative Year 1 and committed anyway.</strong> The financial model shows Year 1 at -121% cumulative ROI. Leadership knew this going in. Most enterprise AI programmes fail because they are evaluated on a 12-month horizon. Sheba’s was evaluated on a 60-month one.</p><h3>The Bigger Picture</h3><p>Sheba’s case is important not because it is exceptional but because it is <em>replicable</em> — with the right conditions. The conditions are not exotic. They require an institution with sufficient patient volume to generate training data, leadership with sufficient patience to accept a negative Year 1, clinical stakeholders with sufficient buy-in to change their workflows, and a technology stack that is honest about uncertainty.</p><blockquote>The 297% ROI is not an AI story. It is a leadership story, expressed through AI.</blockquote><p>The best model in the world, deployed in an institution that lacks those four conditions, will underperform. A modest model, deployed in an institution that has all four, will overdeliver.</p><p>Sheba had all four.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/752/1*svfKtNGjht9wy6A93CEnCA.gif" /></figure><p><em>Next in The AI Case Files: Apollo Hospitals deployed AI radiology screening across a network serving a country with 1 radiologist per 100,000 patients. 3.2 million screenings later — here’s what the numbers actually show.</em></p><p><em>The AI Case Files — Real Deployments. Real Numbers. No Hype. Follow the series on Medium.</em></p><p>#Artificial Intelligence #Healthcare #Enterprise Technology #Clinical AI</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fe1b4f01da4b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Top Healthcare-Specific AI Models (LLMs): Analytical Comparison and Future Impact]]></title>
            <link>https://anupkhanvilkar.medium.com/top-healthcare-specific-ai-models-llms-analytical-comparison-and-future-impact-7f8de83a8f80?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/7f8de83a8f80</guid>
            <category><![CDATA[healthcare]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Mon, 25 Nov 2024 10:29:24 GMT</pubDate>
            <atom:updated>2024-11-25T10:31:21.312Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Cv7hliFO27tkoa40" /></figure><h3>Introduction</h3><p><em>The integration of Large Language Models (LLMs) into healthcare has been transformative, with several advanced models now accessible on Hugging Face. These LLMs offer solutions across a wide range of medical applications — from clinical documentation to diagnostics and research. This article dives into the specifics of seven prominent healthcare-specific AI models, analyzing their features, parameters, and real-world applicability to evaluate why they stand out in the healthcare landscape.</em></p><p><em>As Gen AI is significantly striding different avenues addressing multiple use cases, some common use cases in the healthcare sector are mentioned below.</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/792/0*zZnj3kLdgQDtYC9r" /><figcaption>LLMs application in Healthcare</figcaption></figure><h3>1. Med42–70B: A Pioneer in High-Quality Medical Question Answering</h3><p>Parameters: 70 billion Suitability: Med42, built by M42 Health, has been trained extensively on a curated dataset of medical materials, including flashcards, exam questions, and open-domain dialogues. With 70 billion parameters, this model is finely tuned to handle complex medical questions and generate high-quality answers that match expert-level comprehension.</p><p>Analysis: Med42’s immense parameter size and comprehensive training make it a powerhouse in medical question answering. It outperforms smaller models in accuracy and relevance, making it ideal for use cases such as patient consultations, EHR summarization, and decision support. Med42 is especially beneficial for healthcare providers seeking reliable AI assistance in diagnosing or providing health advice, setting it apart as one of the most suitable models for direct clinical support.</p><h3>2. BioMistral-7B: Specialized in Biomedical Research Support</h3><p>Parameters: 7 billion Suitability: BioMistral, built on the Mistral architecture, is specifically designed for the biomedical domain and is pre-trained on PubMed Central data. With a smaller yet efficient parameter size of 7 billion, BioMistral is optimized for biomedical research, clinical decision support, and interpreting research data.</p><p>Analysis: While BioMistral may not match Med42’s sheer size, its focus on biomedical research data makes it an excellent choice for healthcare providers and researchers in need of quick insights into medical studies and complex medical literature. This model bridges the gap for those in academia or research-driven healthcare environments, offering a streamlined approach to analyzing large datasets. BioMistral’s specificity in training data positions it as an asset in research-driven healthcare sectors.</p><h3>3. Meditron-7B: Robust in Medical Reasoning and Differential Diagnosis</h3><p>Parameters: 7 billion Suitability: Meditron-7B, a model fine-tuned from Llama-2–7B, uses curated medical literature, including articles from PubMed and clinical guidelines, making it exceptional for medical reasoning and diagnostics. Its strength lies in differential diagnosis and medical exam questions, particularly in interpreting nuanced healthcare information.</p><p>Analysis: Meditron’s fine-tuning with trusted medical literature places it at the forefront for healthcare environments requiring accuracy and interpretative capabilities in diagnostics. Its lower parameter count compared to Med42 ensures efficiency without compromising performance, making Meditron highly practical for differential diagnosis tools, patient-facing applications, and assisting medical professionals in rapidly diagnosing conditions.</p><h3>4. ClinicalBERT: Precision in Extracting Data from Clinical Notes</h3><p>Parameters: 110 million Suitability: ClinicalBERT, based on the foundational BERT model, is refined with clinical data from the MIMIC-III database. Although smaller with 110 million parameters, it excels in extracting critical information from clinical notes, offering high precision in clinical data handling and language understanding.</p><p>Analysis: For healthcare providers focused on information extraction and EHR analysis, ClinicalBERT’s compact model structure makes it highly efficient. Its training on MIMIC-III clinical notes makes it invaluable for tasks that involve unstructured data within healthcare records. While it lacks the scale of Med42 or Meditron, ClinicalBERT’s targeted approach allows for streamlined clinical note handling, making it an ideal model for EHR systems and data extraction tasks.</p><h3>5. ClinicalCamel-70B: Expert-Level Support for Clinical Decision-Making</h3><p>Parameters: 70 billion Suitability: ClinicalCamel-70B is designed as an expert-level conversational model for clinical decision support. With 70 billion parameters and a training focus on medical dialogues, ClinicalCamel stands out as a model designed for aiding healthcare professionals with expert insights in real-time.</p><p>Analysis: ClinicalCamel is ideally suited for scenarios requiring high-level clinical decision-making support. Its extensive training on clinical dialogues equips it to handle complex patient interactions and provide expert-level insights, making it a valuable asset for both doctors and clinical staff. With the increasing demand for real-time clinical support, ClinicalCamel proves to be critical for advanced healthcare applications, though it may be more resource-intensive.</p><h3>6. Palmyra-Med-70B: Leading in Biomedical Benchmarks</h3><p>Parameters: 70 billion Suitability: Palmyra-Med-70B, developed by Writer, is specifically crafted for healthcare use and trained on a comprehensive biomedical dataset. It has achieved top performance on biomedical benchmarks, often outperforming even Med-PaLM-2 base models and demonstrating particular strength in research and clinical applications.</p><p>Analysis: Palmyra-Med’s focus on healthcare benchmarks and biomedical data gives it a significant advantage in specialized research environments and clinical decision-making. With its competitive edge over general-purpose models, Palmyra-Med stands as a strong candidate for academic and clinical settings that demand accuracy, relevance, and performance. It’s poised to shape future healthcare innovations, especially in environments focused on cutting-edge medical research.</p><h3>7. OpenBioLLM-70B: Advanced Biomedical Research and Support</h3><p>Parameters: 70 billion Suitability: OpenBioLLM-70B is a model from Saama AI Labs specifically tuned for the biomedical domain. It excels in providing insights into complex medical research topics, clinical diagnostics, and patient record summarization.</p><p>Analysis: OpenBioLLM’s capability to perform in-depth biomedical research tasks makes it a top choice for healthcare professionals engaged in high-level medical research and diagnostic support. Its large parameter count and precise biomedical focus ensure a deep understanding of medical information, offering a substantial advantage for future biomedical advancements and precision healthcare.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/540/0*cwp3IzYxA1HfI_Yr" /><figcaption>Analyzing the right AI model</figcaption></figure><h3>Conclusion: The Future Impact of LLMs on Healthcare</h3><p>The LLMs on Hugging Face are reshaping healthcare by offering tailored solutions across various medical use cases. For future developments, models like Med42, Palmyra-Med, and ClinicalCamel will likely dominate due to their parameter scale and in-depth medical training. On the other hand, models like ClinicalBERT and Meditron serve niche roles that fill critical needs within healthcare systems.</p><p>In conclusion, the adoption of LLMs in healthcare is poised to grow, with each model providing distinct advantages based on its parameter scale, training data, and design focus. As LLMs evolve, their impact on personalized medicine, diagnostics, and medical research will be profound, paving the way for a more precise, efficient, and patient-centered healthcare system.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7f8de83a8f80" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AlphaFold 3 Open-Source Release Revolutionizes Drug Discovery and Molecular Biology]]></title>
            <link>https://anupkhanvilkar.medium.com/alphafold-3-open-source-release-revolutionizes-drug-discovery-and-molecular-biology-58b3a0aff8de?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/58b3a0aff8de</guid>
            <category><![CDATA[alphafold-3]]></category>
            <category><![CDATA[drug-discovery]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[artificial-inteligence]]></category>
            <category><![CDATA[open-source]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Mon, 18 Nov 2024 07:26:25 GMT</pubDate>
            <atom:updated>2024-11-18T07:26:25.013Z</atom:updated>
            <content:encoded><![CDATA[<p><em>The latest advancement in artificial intelligence, AlphaFold 3, marks a revolutionary step forward in drug discovery and molecular biology, signaling a new era for scientific research. Google DeepMind’s decision to open-source this updated version of AlphaFold continues their pioneering efforts in the AI sector. By making AlphaFold 3 available to the public, DeepMind is fostering a spirit of collaboration and innovation within the scientific community, transforming how scientists approach the previously complex challenge of predicting protein structures.</em></p><p><em>In this blog, we’ll dive into the significance of AlphaFold 3, explore the impact it holds for drug discovery and molecular biology, and analyze how this open-source initiative could accelerate scientific advancements.</em></p><h3>The AlphaFold Evolution: From Concept to Cutting-Edge Tool</h3><p>When DeepMind introduced the first version of AlphaFold in 2018, it made waves by achieving remarkable success in predicting protein structures, a task traditionally considered extremely difficult. Proteins, the fundamental building blocks of life, fold in highly complex ways that are essential for their function. Understanding protein folding has been a central challenge in biology for decades. Scientists need accurate structural information to advance in areas like drug discovery, genetic engineering, and molecular biology.</p><p>AlphaFold 2, released in 2021, took the technology a step further by predicting protein structures with atomic-level accuracy for thousands of proteins. This breakthrough received widespread acclaim, with DeepMind releasing the predictions for nearly all proteins known to science, cataloged in an extensive database hosted by the European Molecular Biology Laboratory (EMBL).</p><p>Now, AlphaFold 3 promises even greater capabilities. By improving its accuracy, speed, and accessibility, this latest release can empower scientists to study protein structures at an unprecedented level. DeepMind’s decision to open-source AlphaFold 3 provides researchers worldwide with the tools they need to address critical challenges in medicine, agriculture, and environmental science.</p><h3>How AlphaFold 3 Works and What Sets It Apart</h3><p>AlphaFold 3 leverages powerful machine-learning algorithms trained on vast datasets of protein structures. Here’s a closer look at the core elements that distinguish AlphaFold 3:</p><ol><li>Enhanced Accuracy: AlphaFold 3 has been fine-tuned to predict protein structures with even greater precision. By training on an extensive dataset of known protein structures, AlphaFold 3 produces highly accurate predictions, enabling scientists to study intricate molecular interactions with confidence.</li><li>Faster Prediction Times: AlphaFold 3 is designed to make protein structure prediction faster, enabling high-throughput analyses that were previously too time-consuming or computationally prohibitive. This improvement will help researchers screen thousands of potential drug candidates quickly, accelerating the drug discovery process.</li><li>Increased Accessibility and Collaboration: The open-source release of AlphaFold 3 aims to democratize access to this powerful tool. By sharing AlphaFold 3’s source code, DeepMind allows researchers worldwide to customize the technology, integrate it into their workflows, and experiment with new applications.</li><li>Cross-Platform Compatibility: AlphaFold 3 has been optimized for various computing environments, enabling labs with different levels of computational resources to use the tool effectively. This accessibility could have a transformative effect on smaller research facilities that previously lacked access to cutting-edge protein structure prediction tools.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/0*KBnUkjyKZAwCMWBN" /><figcaption>AlphaFold 3 Features analysis</figcaption></figure><h3>Impact on Drug Discovery: From Hypotheses to High-Speed Insights</h3><p>One of AlphaFold 3’s most anticipated applications is in drug discovery, where understanding protein structures is crucial. Drug discovery often involves identifying compounds that bind to specific proteins in the body to influence their function. For decades, pharmaceutical companies have relied on trial-and-error processes to design effective drugs, which can take years and cost billions.</p><p>AlphaFold 3’s improved accuracy allows researchers to predict the structures of disease-related proteins with high precision. This capability can help identify druggable sites on proteins, providing researchers with critical information to design targeted treatments. Here’s a breakdown of the expected benefits in drug discovery:</p><ul><li>Accelerated Lead Discovery: With AlphaFold 3, scientists can rapidly identify promising drug candidates by simulating how small molecules interact with target proteins. This capability shortens the lead discovery phase, where the search for potential compounds typically takes years.</li><li>Precision Medicine: AlphaFold 3’s predictive capabilities could facilitate precision medicine by helping scientists design drugs tailored to patients’ unique genetic profiles. Understanding protein structures can reveal mutations associated with diseases, enabling the development of personalized treatments.</li><li>De-Risking Drug Development: The ability to predict protein structures accurately allows researchers to assess a drug’s efficacy earlier in the development process. By ruling out ineffective candidates sooner, pharmaceutical companies can invest their resources more efficiently, reducing the risk of late-stage failures.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/984/0*clqmVuIwNuppUBmi" /><figcaption>Alpha Fold 3Impact on Drug Discovery</figcaption></figure><p>The open-source nature of AlphaFold 3 also means that smaller research organizations and startups can access this advanced technology, potentially spurring innovation in drug discovery outside the traditional pharmaceutical giants.</p><h3>Transforming Molecular Biology Research</h3><p>Beyond drug discovery, AlphaFold 3 holds vast potential for molecular biology research. Protein structures are critical to understanding many cellular processes, from DNA replication to signal transduction. AlphaFold 3 can aid researchers in deciphering these mechanisms, offering insights into fundamental biological processes. Key applications include:</p><ul><li>Protein-Protein Interactions: Many biological processes rely on proteins interacting with each other. AlphaFold 3’s accuracy helps scientists visualize these interactions in detail, supporting research into complex cellular pathways.</li><li>Genetic Engineering and Synthetic Biology: AlphaFold 3 can help scientists design proteins with specific functions, advancing genetic engineering. In synthetic biology, this capability can accelerate the development of custom organisms for purposes such as sustainable biofuel production or environmental cleanup.</li><li>Understanding Evolutionary Relationships: By analyzing protein structures across species, AlphaFold 3 offers insights into evolutionary biology, revealing how proteins have evolved over millions of years. This information is crucial for understanding how organisms adapt to different environments.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/781/0*GSQxbRGjrglB4FBJ" /><figcaption>AlphaFold 3 contribution to Molecular biology</figcaption></figure><h3>The Open-Source Advantage: Fostering Collaboration and Innovation</h3><p>DeepMind’s decision to open-source AlphaFold 3 is as noteworthy as the technological breakthrough itself. Open-source software fosters a spirit of collaboration, allowing scientists, developers, and researchers to modify, improve, and build upon the original code. This approach benefits the entire scientific community, as researchers worldwide can:</p><ul><li>Share Findings and Collaborate: Open access encourages the exchange of ideas and data, helping researchers build upon each other’s work. Scientists can share modifications, improvements, and applications with the community, accelerating progress in protein research.</li><li>Expand the Scope of Research: With access to AlphaFold 3, researchers can apply protein prediction models to new areas, such as studying climate resilience in plants or understanding infectious disease mechanisms. This flexibility could result in discoveries beyond the initial scope envisioned by DeepMind.</li><li>Adapt AlphaFold 3 for Specialized Applications: The open-source model allows developers to customize AlphaFold 3 for specific applications. For instance, a research group studying neurological diseases could adapt the software to focus on brain-specific proteins, creating specialized models that further enhance predictive accuracy.</li></ul><h3>The Road Ahead: Challenges and Future Prospects</h3><p>While AlphaFold 3 presents remarkable opportunities, it also brings certain challenges. Protein folding is still a highly complex process, and AlphaFold 3’s predictions may not always account for the dynamic nature of proteins within the human body. Factors such as cellular environments, post-translational modifications, and protein flexibility can influence protein behavior, areas that AlphaFold 3 may need to address in future iterations.</p><p>Looking forward, there is immense potential for integrating AlphaFold 3 with other AI-driven tools. For instance, combining it with AI platforms that predict drug interactions, toxicity, and efficacy could create a holistic approach to drug discovery, accelerating the journey from hypothesis to treatment.</p><h3>Conclusion</h3><p>AlphaFold 3’s open-source release is more than just an update; it’s a catalyst for innovation. By making this advanced tool available to the global scientific community, DeepMind is reshaping the landscape of drug discovery and molecular biology. AlphaFold 3’s ability to predict protein structures with high accuracy, its faster processing speeds, and its accessibility across computing platforms will empower scientists and researchers to tackle some of humanity’s most pressing challenges.</p><p>As researchers worldwide embrace AlphaFold 3, we are likely to see breakthroughs that will have profound impacts on medicine, environmental science, and our understanding of life at the molecular level. AlphaFold 3 stands as a testament to the power of open-source collaboration in driving forward the frontiers of science and technology.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=58b3a0aff8de" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Leveraging AI tools for Sustainability in Consumer Healthcare: Top Tools and Companies]]></title>
            <link>https://anupkhanvilkar.medium.com/leveraging-ai-tools-for-sustainability-in-consumer-healthcare-top-tools-and-companies-1b1e65c43406?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/1b1e65c43406</guid>
            <category><![CDATA[ai-tools]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[sustainability]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Fri, 15 Nov 2024 07:16:24 GMT</pubDate>
            <atom:updated>2024-11-15T07:16:24.841Z</atom:updated>
            <content:encoded><![CDATA[<p><em>As global industries increasingly prioritize sustainability, the consumer healthcare sector is no exception. By integrating Artificial Intelligence (AI) into their operations, companies are making significant strides toward reducing waste, minimizing environmental impact, and optimizing resource use. Here’s an in-depth look at the AI tools driving these changes, the companies using them, and how each tool contributes to a sustainable future.</em></p><h3>How is AI helping achieve Sustainability in Consumer Healthcare companies</h3><p>AI plays a critical role in achieving sustainability by enabling:</p><ol><li>Predictive Analytics — Forecasting demand and reducing excess inventory.</li><li>Supply Chain Optimization — Reducing transportation emissions and energy consumption.</li><li>Efficient Resource Management — Minimizing water, energy, and raw materials use.</li><li>Waste Reduction — Identifying and minimizing waste in product design and packaging.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/0*hDIV9zIaix6Mo6wR" /><figcaption>How AI is helping achieve sustaianabiity in CHC</figcaption></figure><p>How AI helps achieve Sustainability</p><p>So, the question which pops up is — Which AI tools are top consumer healthcare companies using to help sustainability.</p><h3>Top AI Tools in Consumer Healthcare for Sustainability Goals with their Use cases</h3><h3>1. SAP Leonardo for Supply Chain Optimization</h3><ul><li>Company Usage: Johnson &amp; Johnson</li><li>Application: SAP Leonardo helps companies streamline supply chains by optimizing inventory and reducing transportation emissions.</li><li>Sustainability Benefit: Decreases waste and emissions across the supply chain, essential for reducing the environmental impact of transporting consumer healthcare products.</li></ul><h3>2. Microsoft AI for Environmental Risk Assessment</h3><ul><li>Company Usage: GlaxoSmithKline (GSK)</li><li>Application: Microsoft’s AI tools, such as Azure Machine Learning, help assess and mitigate environmental risks in manufacturing processes.</li><li>Sustainability Benefit: Enables early identification of potential hazards, allowing for preventive measures that safeguard the environment and reduce waste.</li></ul><h3>3. IBM Watson for Carbon Footprint Analysis</h3><ul><li>Company Usage: Pfizer</li><li>Application: IBM Watson is used for analyzing and tracking the carbon footprint of manufacturing processes.</li><li>Sustainability Benefit: Provides data-driven insights to reduce emissions, supporting more sustainable manufacturing practices.</li></ul><h3>4. C3 AI for Energy Optimization</h3><ul><li>Company Usage: AstraZeneca</li><li>Application: C3 AI’s predictive analytics optimize energy consumption in manufacturing and storage facilities.</li><li>Sustainability Benefit: Reduces the energy footprint of consumer healthcare production, contributing to lower emissions and more efficient resource use.</li></ul><h3>5. Symphony RetailAI for Waste Reduction in Inventory Management</h3><ul><li>Company Usage: Walgreens Boots Alliance</li><li>Application: Symphony RetailAI manages inventory more accurately, reducing overproduction and waste.</li><li>Sustainability Benefit: Reduces excess inventory and waste in healthcare retail, contributing to a circular economy.</li></ul><h3>6. H2O.ai for Product Lifecycle Analysis</h3><ul><li>Company Usage: Unilever</li><li>Application: H2O.ai analyzes product lifecycles to find ways to make packaging and materials more sustainable.</li><li>Sustainability Benefit: Helps reduce single-use plastics and encourages recyclable or biodegradable materials, minimizing landfill waste.</li></ul><h3>7. IBM Environmental Intelligence Suite for Climate-Resilient Infrastructure</h3><ul><li>Company Usage: Bayer</li><li>Application: Bayer uses IBM’s Environmental Intelligence Suite to prepare for extreme weather events that could disrupt supply chains.</li><li>Sustainability Benefit: Supports proactive risk management, reducing environmental disruption and promoting climate-resilient operations.</li></ul><h3>8. Salesforce Sustainability Cloud for Carbon Accounting</h3><ul><li>Company Usage: Merck</li><li>Application: Merck employs Salesforce’s Sustainability Cloud to track, analyze, and report carbon emissions across its operations.</li><li>Sustainability Benefit: Improves transparency and accountability in carbon reduction efforts, helping Merck meet regulatory and sustainability goals.</li></ul><h3>9. Google Cloud Healthcare API for Data-Driven Environmental Solutions</h3><ul><li>Company Usage: Mayo Clinic</li><li>Application: Google Cloud Healthcare API enables the analysis of patient data to develop eco-friendly healthcare solutions.</li><li>Sustainability Benefit: Promotes data-sharing for sustainable healthcare innovation and reduces infrastructure-related emissions.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/744/0*WTOpOdkubPYOokcy" /><figcaption>AI tools plotted on their impact on sustainability</figcaption></figure><p>Plot of a few AI tools based on the impact and where they stand in organizational strategy</p><h3>Analysis</h3><p>Consumer healthcare companies are adopting AI tools that are cost-effective, tailored to healthcare-specific needs, and developed by trusted vendors with whom they have existing relationships. The current trend leans toward modular and domain-specific tools that can be easily integrated into existing infrastructure. Companies value versatility, with a strong focus on tools that enable both compliance with sustainability standards and long-term cost savings. This selective, strategic adoption is positioning the consumer healthcare sector as a leader in sustainable innovation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/732/0*MreAsjqUPsFkVR8g" /></figure><p>AI adoption for Sustainability in CHC</p><h3>The Future of AI in Sustainable Consumer Healthcare</h3><p>AI’s integration into consumer healthcare has only just begun, and as more companies adopt these tools, the sector will continue advancing its sustainability goals. Here are some key future directions:</p><ul><li>Enhanced Data Sharing: AI tools that encourage collaborative data sharing across companies can help create industry-wide sustainability standards.</li><li>Smart Packaging Solutions: AI is already pushing the boundaries of packaging with recyclable and biodegradable materials.</li><li>AI-Driven Patient Care: Optimized patient-care models reduce unnecessary tests and hospital visits, reducing the overall environmental footprint of healthcare.</li></ul><h3>Conclusion</h3><p>By leveraging AI, consumer healthcare companies are setting new standards in sustainability. From supply chain optimization to carbon footprint reduction, these tools demonstrate the immense potential AI has to help businesses achieve their environmental goals. Through strategic partnerships with AI providers and a focus on sustainable innovation, the consumer healthcare sector is on a promising path toward a greener future.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1b1e65c43406" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Sora - Everything you DONT need to know about this AI model.]]></title>
            <link>https://anupkhanvilkar.medium.com/sora-everything-you-dont-need-to-know-about-this-ai-model-275105ff77ec?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/275105ff77ec</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[open-ai-sora]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[sora]]></category>
            <category><![CDATA[text-to-video-ai]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Thu, 22 Feb 2024 15:46:55 GMT</pubDate>
            <atom:updated>2024-02-22T15:46:55.872Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qi-saYKfKeGUQtGvlpBV5w.jpeg" /><figcaption>Sora is Out of the hat</figcaption></figure><p>Magicians hold audiences spellbound with the awe and intrigue they spark. No matter one’s age, the anticipation of what will emerge from the magician’s hat is universal. Initially, a vibrant ball appears, next a pigeon gracefully takes flight, and then a rabbit hops into view. Everyone is left in suspense, eager to see what marvel will come next!</p><p>AI companies have been pulling off these stunts since the past few years and all those enthusiastic about this technology are left perplexed and spell bounded with what has been coming out of this magic hat. While ChatGPT shook the world with the capabilities this technology could bring, Sora has swept off a lot of industries like a tsunami.</p><h4><strong>What is Sora?</strong></h4><p>Sora is an AI diffusion model from OpenAI which has the ability to generate realistic, high-definition imaginative videos from text. Currently its capabilities are limited to creating 1 min long videos. Nonetheless, the gates are opened!</p><p>While ChatGPT was able to generate meaningful responses from text prompts, Sora is able to generate videos. The distinctive feature is that those videos are realistic, comparable to those created by Hollywood cinematographers.</p><p>The below video was generated by a simple text prompt: “Historical footage of California during the gold rush.”</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FJkyJnmsi0DM%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DJkyJnmsi0DM&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FJkyJnmsi0DM%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/059530110d812f3da6ac471d5c91b366/href">https://medium.com/media/059530110d812f3da6ac471d5c91b366/href</a></iframe><p>Imagine the number of resources spent on capturing this shot in a traditional way. The set, the infrastructure, the lighting, videography, cinematography, drone footage, editing etc. Most importantly the time.</p><p>With Sora — it&#39;s just a text prompt and a few seconds. While it might not suffice the exact requirements in some cases, I am sure customized LLMs, and smart prompt engineering will solve these problems in near future.</p><p>Sora is a diffusion model (models which are used for generating images and videos) which also use the transformer architecture helping the video scale its performance. Its architecture involves using the Dall-E model to generate text descriptions for the input videos, then dividing the videos into small chunks of space time before passing them through the diffusion model. The novelty of the architecture shows how OpenAI innovates cleverly. I will cover the details of their architecture in a separate post to avoid making this one longer.</p><h4><strong>Is Text to Video new?</strong></h4><p>The concept of Text to Video models is not new. There are many diffusion models which generate short length videos including Ali-vilab’s — modelscope, Meta’s -Make a video, Runway’s - Gen 2, Googles- Lumiere. When these models were already generating videos from text, Sora’s popularity seems rather biased and unjustified. But when we compare the video quality, length, its synchronicity over its length, Sora is a leap forward in this area.</p><p><a href="https://www.youtube.com/watch?v=f9ThAzZs32M&amp;list=PPSV">(159) Introducing Lumiere: A space-time diffusion model for video generation — YouTube</a></p><h4>What is so special about Sora?</h4><p>Videos generated by SORA are high quality outputs.</p><p>We can see how the model created a video from the below text prompt incorporating every detail from the prompt.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FUk7d_kWl8y8%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DUk7d_kWl8y8&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FUk7d_kWl8y8%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/2542f5a0ccf79bebe40dcef6ddd968ed/href">https://medium.com/media/2542f5a0ccf79bebe40dcef6ddd968ed/href</a></iframe><p>Text prompt: The camera follows behind a white vintage SUV with a black roof rack as it speeds up a steep dirt road surrounded by pine trees on a steep mountain slope, dust kicks up from it’s tires, the sunlight shines on the SUV as it speeds along the dirt road, casting a warm glow over the scene. The dirt road curves gently into the distance, with no other cars or vehicles in sight. The trees on either side of the road are redwoods, with patches of greenery scattered throughout. The car is seen from the rear following the curve with ease, making it seem as if it is on a rugged drive through the rugged terrain. The dirt road itself is surrounded by steep hills and mountains, with a clear blue sky above with wispy clouds.</p><p>The immense level of precision captured in videos which Sora creates can be seen in the clip below.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FjxFLfJyz4pk%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DjxFLfJyz4pk&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FjxFLfJyz4pk%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/73461fbc36af6979440732222cbc62d6/href">https://medium.com/media/73461fbc36af6979440732222cbc62d6/href</a></iframe><p>Prompt: A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse. She wears sunglasses and red lipstick. She walks confidently and casually. The street is damp and reflective, creating a mirror effect of the colorful lights. Many pedestrians walk about.</p><h4>Sora vs Others</h4><p>OpenAI bumps their competitors with amazing quality output from their models. They provide something par excellence than the best in the market and it forms their USP.</p><p>Although there are multiple Text to Video models in the industry, if we compare the videos they produce, Sora emerges far better.</p><p>Below we can see a fair comparison between Runways Gen 2 and Sora depicting why Sora is way better.</p><p><a href="https://youtube.com/shorts/PS8zM1UakXI">Sora vs Gen 2</a></p><h4>Limitations of SORA</h4><p>SORA is a diffusion model which works by adding static noise to a video and then working backwards removing this noise to generate the required output. So, there are problems with the model understanding the causal relationship or with the general physics of a complex scene.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FGrM-tJTEsJc%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DGrM-tJTEsJc&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FGrM-tJTEsJc%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/630ec1ea50e09a934d4633f911a40f62/href">https://medium.com/media/630ec1ea50e09a934d4633f911a40f62/href</a></iframe><p>Or can be seen the video below.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FnfKq99y0Xps%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DnfKq99y0Xps&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FnfKq99y0Xps%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/e0077a5b591f66da07685e8dec261b6d/href">https://medium.com/media/e0077a5b591f66da07685e8dec261b6d/href</a></iframe><h4>Where are we going with Sora?</h4><p>It is quite evident that the race to AGI is on! Nearly all the tech giants, mid-sized or small companies in this space are competing ferociously to reach AGI, provided they have the resources.</p><p>What is AGI? — Artificial General Intelligence is a term often deciphered contextually by different people in different ways. Here, by AGI, I mean a machine having intelligence to understand, learn and apply knowledge at a wide level of tasks possibly better than humans.</p><p>Is ChatGPT (GPT4) AGI? No. Although the model is extremely clever to understand and answer, it is only trained on vast amount of <strong>textual</strong> data. Although its capabilities to answer factual questions is arguably better than most humans, it is lacking the visual, auditory, sensory, emotional experiences that we human beings collect over our span of life. AGI is imparting all those capabilities to a machine. So, when we ask this AGI model a question or give him any possible humane task, it will consider all the perspectives when giving an output or produce the output in that format.</p><p>This model could probably do anything and everything that a normal human could do, probably even better. Although this might seem something plausible, we need to consider that human beings have extremely complex minds and supernatural capabilities. We have produced some great works over the period of time.</p><h4>Road ahead</h4><p>AI companies like OpenAI have a clear roadmap of achieving this goal of AGI and they have a unique way of sharing their milestones with us. With Google publishing their updates silently or Altman tweeting about Sora, they keep the people excited about the journey, yet skeptical about the destination. Nevertheless, we, humans have never felt an excitement of welcoming an artificial counterpart in our World. Our own ruled World!</p><p><em>P.S : Content 100% human generated.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=275105ff77ec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Predicting AI Trends in 2024: A Glimpse into the Future]]></title>
            <link>https://anupkhanvilkar.medium.com/predicting-ai-trends-in-2024-a-glimpse-into-the-future-8be2b1403faa?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/8be2b1403faa</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[2024-trends]]></category>
            <category><![CDATA[2024]]></category>
            <category><![CDATA[trends]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Thu, 04 Jan 2024 12:18:49 GMT</pubDate>
            <atom:updated>2024-01-04T12:23:26.668Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*r5VGB-g-MBam9oq3qC-wxw.png" /></figure><h3><strong>Introduction</strong></h3><p>As we step into 2024, Artificial Intelligence continues to reshape our world. This post delves into the most promising AI trends for the year, offering insights into how these technologies are set to evolve and impact various sectors.</p><p>1. <strong>Advanced Natural Language Processing (NLP)</strong><br>Justification: NLP technologies have seen rapid advancements, largely driven by more sophisticated models and increased computational power.<br>Real-life Example: GPT-4, an AI model by OpenAI, is revolutionizing content creation, enabling more nuanced and context-aware interactions.</p><p>2. <strong>AI in Healthcare: Predictive Analytics</strong><br>Justification: With the growing need for personalized medicine, AI’s role in predictive analytics for patient care and treatment is increasing.<br>Real-life Example: IBM Watson Health is using AI to analyze medical data and provide tailored treatment recommendations.</p><p>3. <strong>Autonomous Vehicles</strong><br>Justification: Advancements in machine learning and sensor technology are accelerating the development of fully autonomous vehicles.<br>Real-life Example: Waymo, a subsidiary of Alphabet, is leading the charge in self-driving car technology, already conducting public trials.</p><p>4. <strong>AI Ethics and Governance</strong><br>Justification: As AI becomes more integral to our lives, the need for ethical guidelines and governance frameworks becomes paramount.<br>Real-life Example: The European Union’s AI Act is a pioneering legislative effort to set standards for AI ethics and safety.</p><p>5. <strong>AI-powered Cybersecurity</strong><br>Justification: With cyber threats becoming more sophisticated, AI is increasingly being deployed for proactive threat detection and response.<br>Real-life Example: Companies like Darktrace are using AI to detect and respond to cyber threats in real-time.</p><p>6. <strong>Quantum Computing in AI</strong><br>Justification: Quantum computing promises to exponentially increase processing power, potentially revolutionizing AI capabilities.<br>Real-life Example: Google’s Quantum AI lab is exploring the integration of quantum computing with AI for solving complex problems.</p><h3><strong>Conclusion</strong></h3><p>As these trends indicate, 2024 is set to be a pivotal year in the AI landscape, bringing both opportunities and challenges. The integration of AI in various sectors will continue to drive innovation, reshape industries, and influence our daily lives.</p><p>Follow us on <a href="https://www.linkedin.com/company/acuriousbot/?viewAsMember=true">Linkedin</a> or <a href="https://twitter.com/CuriousBot14">Twitter</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8be2b1403faa" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[ChatGPT — 9 key facts]]></title>
            <link>https://anupkhanvilkar.medium.com/chatgpt-9-key-facts-5297d3d2f4e7?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/5297d3d2f4e7</guid>
            <category><![CDATA[chatgpt3]]></category>
            <category><![CDATA[facts]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[openai-chatgpt]]></category>
            <category><![CDATA[openai]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Sun, 25 Dec 2022 15:34:06 GMT</pubDate>
            <atom:updated>2022-12-25T15:34:06.922Z</atom:updated>
            <content:encoded><![CDATA[<h3>ChatGPT — 9 key facts</h3><p>ChatGPT is OpenAI’s endeavor in optimizing a language model for dialogues.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/242/1*vwAKh0-NbDtdz_Z3oLNpzQ.png" /></figure><p>It is interesting to understand some of the key facts of this immensely viral language model which is released in the form of a chatbot.</p><h4>Fact #1 <strong>It costs approximately $100,000 each day to run ChatGPT.</strong></h4><p>OpenAI which is funded by Microsoft is spending a fortune to keep this model available for public use.</p><p>A single chatGPT query uses 8 NVIDIA A100 GPUs</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/265/1*t2nCVCb5kBya6cLOUccNSg.png" /></figure><h4>Fact #2 <strong>ChatGPT is trained on data only till December 2020.</strong></h4><p>Yes. Feels quite odd that this knowledge machine does not know anything that happened after December 2020. When asked about its counterpart Google’s LaMDA (released May 2021) — ChatGPT responds that it is aware of only BERT language model released by Google.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/413/1*B0yxM0_fUeSR1ywshprzlQ.png" /></figure><h4>Fact #3 <strong>ChatGPT is the 5th model released by OpenAI in terms of finetuning their GPT model series</strong></h4><blockquote>GPT was released in Jun 2018</blockquote><blockquote>GTP-2 was released in Feb 2019</blockquote><blockquote>GPT-3 was released in May 2020</blockquote><blockquote>GPT-3.5 was released in Jan 2022</blockquote><blockquote>ChatGPT was released in Nov 2022</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/622/1*9K6gwR5XYZy8SuCkVXZ2Lg.png" /></figure><h4>Fact #4 <strong>GPT-3 model is finetuned with RLHF (Reinforcement learning using human feedback) and then trained on dialogues and rules to avoid inappropriate answers. That gave us ChatGPT</strong></h4><p>GPT3.5 — released in Jan 2022 is also called instructGPT</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/645/1*tXfvCxtDpfHxFktgsXQfPA.png" /></figure><h4>Fact #5 <strong>Is ChatGPT the largest language model till date?</strong></h4><h4><strong>No. There are numerous language models released which are trained on larger datasets. However ChatGPT is a model finetuned for dialogues.</strong></h4><p>Below — A representation of various models in terms of the number of parameters (in Billion) they are trained on (Till Oct 2022). ChatGPT is based on GPT-3 which is trained on 175 B parameters</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/653/1*dKy1wo_mowLP5O_vRhM42w.png" /></figure><h4>Fact #6: <strong>Language models are build and released on a large scale in recent years. In 2022 alone 52 language models were released</strong></h4><p>The significant increase in language models can be attributed to the encouraging results shown by OpenAI’s GPT model released in 2020.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/611/1*xpFtmadavuT9vvYBo5zJPg.png" /></figure><h4>Fact #7 <strong>The training data for GPT-3 on which ChatGPT is finetuned from consisted of -</strong></h4><h4><strong>410B tokens from Web data</strong></h4><h4><strong>55B tokens from online books</strong></h4><h4><strong>12B tokens from books</strong></h4><h4><strong>3B tokens from Wikipedia.</strong></h4><p>Around 480 tokens were fed to formulate 175B parameters to train this model.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/592/1*yRuoxiY9By_e5sKkgcyZZA.png" /></figure><h4>Fact #8 <strong>ChatGPT can lie (unintentionally??)</strong></h4><p>On account of huge amount of unreliable data fed to train this model, ChatGPT often provides incorrect answers to our questions.</p><p>The dialogue model although make it sound very convincing.</p><p>If you correct ChatGPT, it quite humbly accepts its mistake of providing misinformation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/350/1*jUWwB6OVBv3ximU7mR1MxA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/380/1*ham9J2byhQJK3HnbTYE51g.png" /></figure><h4><strong>Fact # 9 GPT-4 is coming soon</strong></h4><p>OpenAI plan to launch GPT-4 soon. The possibility is that this model will be trained on a gigantic dataset formed from the web and also from the interactions people had with chatGPT.</p><p>It is quite fitting that the huge interactions chatGPT has with people will generate humongous dialogue training data which OpenAI can use to train a new language model/ With this data, the model will hopefully be more finetuned and enhanced on account of some finetuning techniques (which openAI has not disclosed so far)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/333/1*3NNGQBu0zvCgVXrRjNJPRA.png" /></figure><p>Thanks</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5297d3d2f4e7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Do we really need a Metaverse?]]></title>
            <link>https://anupkhanvilkar.medium.com/do-we-really-need-a-metaverse-b26b6745e8ae?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/b26b6745e8ae</guid>
            <category><![CDATA[metaversesociety]]></category>
            <category><![CDATA[facebook]]></category>
            <category><![CDATA[metaverse-nft]]></category>
            <category><![CDATA[metaverse]]></category>
            <category><![CDATA[meta]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Fri, 10 Dec 2021 14:49:12 GMT</pubDate>
            <atom:updated>2021-12-10T15:07:28.832Z</atom:updated>
            <content:encoded><![CDATA[<p>As the buzz around Metaverse grips the World, we all are excited to enter a new World where experiences will be defined differently.</p><p>But, how much do we really need the Metaverse?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*mnc621SfQQ3JkxPZPlEcQw.png" /></figure><p><em>In his dark room with his Quest headset on, Rhean had a distinguishable contented smile on his face. He was logged in Decentraland where he wore an expensive suite as he stepped down from his Tesla Model Z. While feeling the rich soft fabric of his coat, he was admiring the shiny glaze of his brand new Tesla Model Z car which he had bought from Horizon marketplace yesterday. He had just dropped off his girlfriend after giving her a ride, showing off his new Tesla to her, when suddenly, a screen popped in front of him. It was a notification from his friend Alberto, who lived in Spain. Alberto was inviting him in SoccerlWorld where his friends from different parts of the World were waiting for him to complete their playing 22 on the field in a game of soccer. Rhean gestured his hands to switch his screens in the VR world. The VR headset switched him from Decentraland to SoccerWorld changing his gear to his team’ s football kit in a matter of few seconds. He was on the ground. He looked at his muscular Avatar in SoccerWorld. The top-right corner of his VR glasses showed his skill level which was higher than anyone on the field that day.As his game was about to begin, his screen splashed again with a call notification. This time it was his mother.</em></p><p><em>“Rhean, I had been telling you since a week to fix your bathroom. The taps need repair and for heavens sake also allow us to fix this broken door.” She was outside the locked door of his tiny room. She peeped through the small crack in the rusty old door.</em></p><p><em>“Mom, I am in a game right now. I will call you later.”</em></p><p><em>Rhean never understood why his mother would bother him about things in the physical world. When the game was over, he switched his headset to the physical world and could see the mess around him. His room was distorted and the water running from his bathroom tap had encroached his bed. After a real tight soccer game, he was looking for his favourite chair to relax, until he realized he had sold it out last week to buy some Soccer coins. He had used them to upgrade his Avatar in Soccerworld. He sat down and looked around. His VR screen flashed him a notification showing the approximate second hand price for his old table.</em></p><p><em>“100$ is not bad deal. I could get around 1200 ICP coins which I could use to buy a glazing new color for my Tesla in Decentraland or a virtual sofa set for my living room in Horizon Home.” — He thought. He wondered why his mother always insisted to buy some decent furniture for his room in the physical World. “If only my parents could spend more time in metaverse, they would come to know the riches I own there” — He wished.</em></p><p>A decade back, the story above might have sounded like a futuristic Sci-Fi script probably featuring the year 2100, but right now this does not seems a so distant future.</p><h4>What is a Metaverse?</h4><p>Metaverse, in simple terms refers to all the virtual spaces or virtual worlds. For those who already know the gaming space, the concept of virtual world is not new. Games like Minecraft, SecondLife etc were created in the last decade where users could enter with their own Avatar (digital version of self) explore a world, build structures, interact socially, complete tasks or play games . In some of those games, users were not only able to develop their land or assets, but also advertise and monetize them by selling them in a marketplace.</p><p>With the advent of blockchain technology, some games like Decentraland, Sandbox allowed users to make those purchases over a blockchain. With immense advancement in AR/VR technology, a lot of projects have started complementing these games with immense real world-like experiences.</p><p>The worlds in those games could be considered as virtual spaces. All these worlds combined make up the Metaverse.</p><p>The term is coined from a sci-fi book named “Snow Crash” published in 1992 and now, is used to create a hype around the projects currently being developed around in this space.</p><h4>What can you do in a Metaverse?</h4><p>As many companies have jumped into this space, the possibilities explored are endless. Just putting an AR/VR set, sitting at home, a person can teleport anywhere virtually. He can not only visit, but also experience beautiful places, concerts around the World, or have some exotic experiences like skiing in the Alps or Skydiving in Dubai or simply have a delicious breakfast on a beachfront resort somewhere in Miami. All this while sitting in their home in any part of the World.</p><p>By entering their office virtual space right from the comforts of their homes, one can be at office and home at the same time. One can experience climbing Everest or being chased by Dinosaurs while staying safely in their rooms. For the real estate lovers, it provides the possibility of buying and experience living in a virtual house on a beautiful Spanish beach with ample sunlight throughout. You can also develop or build your own structures, houses over open land and sell and monetize it.</p><p>All those capabilities makes a very strong proposition for this technology and ensures that users are bound to spend time and money in these virtual worlds or Metaverse. This would also mean that people will start buying a lot of digital goods here. In order to ensure a secure decentralized experience, mostly the transactions in this space are planned to be based on the emerging blockchain technology. This also explains why NFT’s ( Non Fungible Tokens ) or virtual digital assets transacted on blockchains are gaining value recently.</p><h4>What is happening in this area</h4><p>While the World is moving towards VR (Virtual Reality), AR (Augmented Reality) adoption as part of the metaverse boom, businesses around the World are scripting their strategies around this space.</p><p>It was not much surprising for a company like Facebook, now ‘Meta’(whose business strategy always aimed at having maximum users spending maximum amount of time on their platform )to turn its head towards this undenyingly addictive market. Rebranding themselves to “Meta” or Meta Platforms in 2021 was corollary after they acquired Oculus (Virtual reality headset manufacturer) back in 2014.</p><p>Meta currently has numerous ongoing projects which include Horizon Home — where users can buy and decorate homes. Horizon World — which allows users to build their own virtual worlds. Horizon Marketplace — a shopping place. While some companies, in order to improve productivity at workplaces had banned social media sites like Facebook, Meta is planning to integrate 3D VR/AR work environment through their Horizon Works project.Here is a short video where Zuckerberg explains about his future project plans in this space -</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FgElfIo6uw4g%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DgElfIo6uw4g&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FgElfIo6uw4g%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/9dd5722b9b6bf0df9951c36529e075fb/href">https://medium.com/media/9dd5722b9b6bf0df9951c36529e075fb/href</a></iframe><p>Its no doubt that Zuckerberg is planning to enter in all those areas which were restricted for him due to the limitations Facebook as a social media platform had.</p><p>But its not only Meta, Microsoft introduced Mesh, its VR bet designed to be used in Teams application. Its alluring marketing video promises a lot of stuff. They also plan for virtual work experience through their “holoportation” technology. They introduced Hololens, their VR/AR set back in 2019.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FJd2GK0qDtRg%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DJd2GK0qDtRg&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FJd2GK0qDtRg%2Fhqdefault.jpg&amp;key=d04bfffea46d4aeda930ec88cc64b87c&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/fc157eea36eef2b72aaba21f0e021823/href">https://medium.com/media/fc157eea36eef2b72aaba21f0e021823/href</a></iframe><p>Chinese giant Alibaba has trademarked their metaverse and many giant companies have started working on projects around this. Adidas buying land in Sandbox to create adiVerse and Nike buying land to create Nikeland in Roblox, shows how retail giants dont want to miss out in this race. Virtual influencers are created by major brands like Prada — Candy or Yoox’s- Daisy to try out new marketing techniques.</p><h4>Advantages</h4><p>Undoubtedly there are numerous advantages — The convenience of being anywhere, anytime within a matter of seconds is special. Having a disabled person experience some exotic experiences would be a remarkable achievement. Since you could almost teleport to any place, it would reduce actual travel and hence could as well bring down carbon footprint over the years. It might open doors to new technological advents, new types of jobs and new ways to earn money and a lot of scope for creative ideas.</p><h4>But…</h4><p>But the question is what are we giving away with it?</p><p>Today, as mobile devices have consumed us, we cannot let go ourselves from our mobile phones. <strong>How healthy would it be for our next generations to be addicted to 3D VR/AR sets</strong>. They would eventually become an integral part of us or a necessity. For someone in the 1970’s, it would had been a funny thought to have imagined humans occupied in a small handheld device for a considerable percent of their time. Yet, here we are. Similarly, today no matter how far fetched the idea of VR sets integrating with our body seems, we might not even know when we embrace this. No matter how sleeky or stylish they would look, integrating a digital product with our body sounds scary.</p><p>While its always exicting to embrace new technology, it sends shiver through our spines when we know that our data in the form of minute details will be available to those technology companies like Meta, which already have a shady past on their ways of using this data. Each movement of ours, likes-dislikes will be tracked with AI softwares on the context of providing a better user experience. Now if that did not threw privacy out of the window - <strong>Meta is also </strong><a href="https://tech.fb.com/inside-facebook-reality-labs-wrist-based-interaction-for-the-next-computing-platform/"><strong>planning to capture electric motor nerve signals</strong></a><strong> using electromyography</strong>, which means we would be relying on these tech giants with our thinking pattern data.</p><p>Virtual Worlds sound cool. But with huge number of people with different interests and opinions, we ought to have radical groups, crimes, thefts , hate speeches here which makes it extremely necessary for moderation. <strong>The question is how much can we trust these technological companies on a fair and unbiased moderation of their humoungous userbase</strong>. Someone could suggest of providing an opensource codebase for these metaverses as an option, but companies in this space would seldom agree with this, especially in this compeptitive world. Governments intervening to bring in regulations could might end up as good as those jokes like those recent trials some technological companies went through.</p><p>Finally these worlds would still be controlled by some central entity and having vast number of users managing their businesses and finances in these worlds would rest tremendous power in the hands of the creators of these worlds.</p><p>In my opinion the biggest fear would be of loosing the hold of reality in our physical world. As we would start focussing more on these virtual worlds, there is a huge risk of neglecting our health, life, relationships in the actual world.</p><p><strong>Do we really need a Metaverse?</strong></p><p>May be yes. May be no.</p><p>Metaverse opens up numerous doors for mankind to move further.Also provides us numerous options.</p><p>Like the story narrated at the start of this article, we might be able to virtually play whatever we want, anytime, anyplace and with anyone. Although, being a millenial, I always cherished those evening hours I spent under the sun playing with actual people on real grounds. When you dint had to log in from a dark room, but had to just get out in sun on a ground and start playing while your friends join in.</p><p>With the rate the disruption is taking place in this area, we might see an all different world in 2 decades. I just hope we dont have to explain what basking in the real sun or sweating profusely on a ground mean to our next generations.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b26b6745e8ae" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Rise and Rise of Cryptocurrencies]]></title>
            <link>https://anupkhanvilkar.medium.com/the-rise-and-rise-of-cryptocurrencies-5ba330280aac?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/5ba330280aac</guid>
            <category><![CDATA[cryptocurrency-investment]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[crypto]]></category>
            <category><![CDATA[bitcoin]]></category>
            <category><![CDATA[cryptocurrency-news]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Mon, 15 Nov 2021 11:21:09 GMT</pubDate>
            <atom:updated>2021-11-15T11:21:09.040Z</atom:updated>
            <content:encoded><![CDATA[<p>Regardless of the daily upheavels in the crypto market, the entire World has realized the utility of a distributed ledger system and the benefits it has over a traditional centralized system. As for we know, we associate cryptocurrencies with blockchain, the technology which lies underneath majority of the cryptocurrencies in the market.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*elGEA62aCngEURpjtcvz3Q.jpeg" /></figure><h4>Preface</h4><p>Before we delve into Cryptocurrencies, first we try to understand the reason why they came into existence.</p><p>The 2008–09 recession not only shook the real estate world, but pointed out the problems with the current monetary system. For those who well understood the financial markets and knew how the monetary cycle worked, could had it predicted long back. After the great recession of 1920’s, in order to boost the economy, the UK government removed the Gold standard in 1931 and the US, under the leadership of Roosevelt followed suite in 1933. In 1971, Nixon completely removed the remnants of this standard thereby converting the US dollar to a completely fiat (not backed by any tangible asset) currency. In a way, this allowed the Government to control the economic activity in the market by controlling the influx of money into the economy.</p><p><strong>What was the Gold standard and why was it removed?</strong></p><p>The Gold standard limited any country to print and put into circulation only the amount of money as corresponding to its actual Gold reserves. This helped placing a control on inflation or deflation, since the amount of money which could be printed was always tied to a tangible asset like gold. So, why was it removed?</p><p>The economic recession of late 1920’s and the huge spending of US government during the 60’s in order to combat the Cold war, showed that the necessity of money for keeping the standard far exceeded than the amount of Gold mined every year. Then the rise in population which did not grew proportionally to the gold reserves of a country, made the amount of money, per individual less. With the price of goods already high along with its demand, it made people to spend less in turn leading to economic instability. All this lead the governments around the World to remove the Gold standard in 1930’s. During the 60’s the US owed huge debts to its European counterparts and with its Gold supplies limited, they were left with the only option to print money for trading in the form of Government bonds without really worrying about the Gold reserves in the country. So, the standard associated with trade policies was removed as well.</p><p><strong>Why are we talking about the removal of Gold standard?</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/724/0*cytX2uwOK5rKZn_4" /></figure><p>Why are we talking about the removal of Gold standard?- Well, this decision, in future went far ahead to cause economic turmoils, fake price bubbles and pushed the US Government in a debt of more than 10 trillion dollars leading to a recession in 2008–09 (the debt in 2021 stand at 28 trillion dollar). The graph shows how the US Federal Reserves debt have grown over the years and everyone turns a blind eye towards it. It practically means that we are living in a World where some people hold the power to create money, distribute it, while nearly 99% of the rest dont have a idea where it is going or coming from.</p><p>The risks of having the power of controlling the influx of money in the hands of a few was identified by some people, who bought in an alternate medium to trade goods and services. This medium was called the Bitcoin, which was based on an algorithm or logic and used the SHA-256 cryptography function at its core. Since it used cryptography at its heart, the curreny was called cryptocurrency. To add to it, Bitcoin allowed anonymous transactions as the person holding bitcoins in his wallet did not needed to reveal his identity.</p><p>This medium of exchange, initially taken as a joke, caught the eye of Government when Bitcoin started finding some popularity and in turn gaining some value in the eyes of common people. Anticipating the disapproval from the Government for challenging their authority, the creator of Bitcoin has never revealed his true identity (till until now), and kept himself masked under the name of ‘Satoshi Nakamoto’. The idea of a currency created by an anonymous is enough to intrigue people into knowing more about it. Bitcoin gained enormous value over the years as it could be seen from the table below-</p><h4>Rise of Bitcoin — History</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/401/0*FZaqicmbHS6sBqtc" /></figure><p>As it can be seen in the table, Bitcoin rose from a value of 0.09 in 2010 to 770 USD in 2014, an unbelievable rise for any investment. In May 2021 it reached a peak of around 61000 USD per bitcoin, standardizing itself back at 29000 USD in around July -August and soaring back to 64000 in November 21 . Which means anyone who would had bought bitcoins worth 10 USD in 2010 would have BTC worth around 6 million USD in May 2021.</p><p>The graph also shows the extreme volatility of this market specifying the enormous money some people make out of it.</p><p>The popularity of Bitcoin cannot be however accredited to only its founder. The idea of a parallel monetary system gained huge popularity among some who well understood the power which came by holding something like Bitcoin which had a high probablity of gaining popularity.</p><p>Jed Mccaleb launched Mt Gox website which was later acquired by Mark Karpeles. The site became the first crypto exchange launched in 2010 and handled over 70 percent of all Bitcoin transactions at that time. Mark Karpeles, the CEO of Mt Gox crypto exchange was just in his 20’s when he founded Bitcoin foundation in order to promote the use of Bitcoin. However, the idea of anonymity and no regulations also attracted a lot of people who found it as a loophole in the monetary system to earn some quick money. Ross Ulbricht, who tried selling drugs on darkweb using Bitcoin as a way of exchange and many more like him who were apprehended set an example in front of us all of how this non regulated currency could be misused. Mt Gox, the first crypto exchange later ran into cyber security problems claiming that bitcoins on their sites were stolen. They filed for bankruptcy in 2014, when they declared of loosing around 850,000 bitcoins due to a breach in their website security. This was not a problem with the cryptocurrency algorithm, but rather a problem on how Mt Gox website stored bitcoins in wallets. Eventually this led to the arrest of Mark Karpeles as well. Bitcoin price fell sharply during this time of 2014–15. It was during this same time that a 19 year old skinny guy called Vitalik Buterin, who had been a founding member of Bitcoin magazine since 2011, founded Ethereum and the concept of Smart contracts.</p><p>Inspite of its shady past, cryptocurrencies gained momentum as some intelligenct investors started understanding its potential. The tech savvy people quickly understood the implications of smart contracts and by 2021 nearly 8000 different cryptocurrencies or tokens came into existence.</p><p><strong>What is the difference between a cryptocurrency and a token?</strong></p><p>Simply put, the main difference between the two is that a cryptocurrency has its own blockchain with its own set of rules, while a token uses an already existing blockchain and conform to its rules.</p><h4><strong>Finding huge Support</strong></h4><p>In 2019/20 many people like Mark Cuban, Robert Kiyosaki and Elon Musk came forward and supported the idea of decentralized financial systems. Bitcoin gained huge value in 2020/21 where it rose from 10k per bitcoin in July 2020 to around 29K in Jan 2021. It went at its peak to 64k in May 2021 before stabilizing at 40k at the time of writing this.</p><p><strong>Altcoins</strong></p><p>Any cryptocurrency apart from Bitcoin came to be referred as Altcoin (Alternate Coin). Now, as we can see there are around 8k altcoins. While Ethereum(ETH), Binance(BNB),Cardano (ADA), XRP have found some popularity on account of the unique technology their blockchain offer, some meme coins like Dogecoin,Shiba Inu saw a huge growth making many people millionaires in a few days.</p><h3>Problems</h3><p>Many cryptocurrencies operate in a way such that their transactions need to be validated by a lot of servers running on high power. This usage of huge electric energy if often considered a big hindrance to the functioning of cryptocurrencies and is often challenged by environmentalists. However, as per data, most of the servers which mine (validate transactions) those cryptocurrencies are found to be pumped by renewable sources of energy.</p><p>As some coins have gained popularity, there are some huge institutins or individuals who try to manipulate the price of these coins. They pump huge amount of money into the crypto market to fluctuate the price of some coins and make profits. These people are refered to as<strong> “Whales”</strong> in the crypto trading world owing to the huge size of investment they hold. They have the potential to fluctuate the prices of some cryptocurrencies massively to cause heavy profits or losses to people and they hold the advantage to use this power to get more rich. On account of this, many Governments around the World have started to plan on implementing regulations on cryptocurrency trading.</p><h3>Conclusion</h3><p>It is an established fact that value of anything depends on how majority of people percieve it. For instance, wood or iron has far more utility as compared to Gold, yet Gold has always been far more precious than either of them. A whole carton box of printed money could be more valuable than an entire house.</p><p>A cryptocurrency is a non tangible asset just like the paper money we deal with everyday. Its many abilities like providing secure non reversible transactions and digitally accessible records provides it the value it has. Moreover it is also about what people think about it.</p><p>A decentralized system which is governed by predefined rules which are not easily modifiable serves as a strong alternative to a centralized system governed by a few people often influenced by policies made by some individuals who hold power. But how comfortable are we to trust the unknown “Whales” or the market influencers in Crypto world against the known conservative stealers like banks, central banks or federal reserves is still for us to decide.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5ba330280aac" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Google announces LaMDA — What is it?]]></title>
            <link>https://anupkhanvilkar.medium.com/google-announces-lamda-what-is-it-452e82771f5?source=rss-62bf8434b066------2</link>
            <guid isPermaLink="false">https://medium.com/p/452e82771f5</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[nlp]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[lamda]]></category>
            <category><![CDATA[google]]></category>
            <dc:creator><![CDATA[Anup Khanvilkar]]></dc:creator>
            <pubDate>Sat, 12 Jun 2021 12:22:29 GMT</pubDate>
            <atom:updated>2021-06-12T12:22:29.423Z</atom:updated>
            <content:encoded><![CDATA[<h3>Google announces LaMDA — What is it?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/1*t_VPC6imoL1p3EXy5UxP_g.gif" /></figure><h3>Background</h3><p>The need and importance of conversational AI had long been identified by companies like Google, whose key products are based around natural language processing.</p><p>But have you ever tried interacting with Google or your Google home device? Did you found it exactly like interacting with another human?</p><p>Here is a sample chat a human has with Google -</p><blockquote>Anil — “Hey Google, Can you tell me where is the nearest I can watch Avengers Endgame?”</blockquote><blockquote>Google — “Its in the Cineplex theater near Pool plaza, just 1.5 km from your house. Do you want me to book a ticket there?”</blockquote><blockquote>Anil — “Yeah. What is the price?”</blockquote><blockquote>Google — “The ticket for the matinee show is 100 bucks.”</blockquote><p>Probably a 7 year old kid or our grandmothers would find it amusing that a machine responds to our queries, but honestly, the conversations with Google or Google home never are as close as an interesting conversation we have with other humans.</p><p>Have a look at a conversation between 2 office colleagues below.</p><blockquote>Anil — “Hey Sameer, Can you tell me where is the nearest I can watch Avengers Endgame?”</blockquote><blockquote>Sameer — “I think its in Cineplex. You should watch it. I loved Thanos.”</blockquote><blockquote>Anil — “What? Isn’t he the shady dark character trying to end half of the World with a snap of his finger?”</blockquote><blockquote>Sameer — “Yeah. But I wish I had the same power. I could make our manager disappear with a snap of my finger.”</blockquote><blockquote>Anil — “Haha. Sounds like a better world. By the way do you know how much it costs there?”</blockquote><blockquote>Sameer — “I guess 100 bucks. But the effects are better with a ticket of 300.”</blockquote><p>Which conversation did you like? With Google, You ask a question and it responds with a suitable answer. That is certainly not how humans interact. Unless you are having a conversation with your girl friends father, human beings tend to have conversations based on emotions, not tied to a single subject, often wandering off to a rather humorous or different topics and mostly without the responses being repeated. They might not look like the most optimal way to get answers, but they are emotionally satisfying and refreshing. That is probably why many studies ruled out the possibility of an AI robot replacing a human companion.</p><h3>Introduction</h3><p>Google developed 2 transformer based Natural Language Processing models based on BERT, in 2018 and has been using it since 2019 to better understand user searches. They also improved it, altered it to use it for their moderately popular Google Home or Google Nest as it is called today. But the focus has always been on understanding what the user wants to know or search and providing him with the appropriate search results rather than engaging him in a meaningful conversation. Yes, Google searches or Google Home do answer effectively and the conversations are sometimes pretty helpful for someone looking for answers, but we cant deny the fact that its conversational abilities are limited. Try telling Google Home some weird story or ask it some weird question and it will be left perplexed possibly answering its usual response -</p><p>“Sorry I did not understand”.</p><p>If you try asking the same question twice, you will most likely end up receiving the same response. This is because the language models which work behind have some predefined paths leading to a particular response as an output.</p><p>Last year <a href="https://anupkhanvilkar.medium.com/gpt3-a-peek-in-the-future-b40ee85ed982">GPT-3</a> opened the doors to an all exciting world of conversational AI, possibly giving a glimpse of countless possibilities we have in front of us by training a model on an extensive or rather humongous set of data.</p><h3>LaMDA</h3><p>On May 18th 2021, at Google I / O, Google announced LaMDA (Language Model for dialogue applications), a language model trained on dialogues or conversations. What is interesting about this model is its ability to collate information and synthesize meaningful responses. Google calls it instilled sensibleness. Similar to BERT and <a href="https://anupkhanvilkar.medium.com/gpt3-a-peek-in-the-future-b40ee85ed982">GPT-3</a> , it is based on transformers, which is a neural network architecture, where the model is trained to find the best next word or phrase suitable for the given context. I believe what it really stands out is the flow of dialogue or conversations it offers which gives it a more humane ability.</p><p>Below a small video on how LaMDA was trained to act as planet Pluto and get in a meaningful conversation with the team of Google.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FaUSSfo5nCdM%3Fstart%3D86%26feature%3Doembed%26start%3D86&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DaUSSfo5nCdM&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FaUSSfo5nCdM%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/0c3f0ee94a335badd6f7516c809a30f9/href">https://medium.com/media/0c3f0ee94a335badd6f7516c809a30f9/href</a></iframe><p>What is notable in the above conversation is that the model not only answered the questions asked about the planet Pluto, but it synthesized some responses to form a meaningful conversation.</p><h3>How is it different than GPT-3</h3><p><a href="https://anupkhanvilkar.medium.com/gpt3-a-peek-in-the-future-b40ee85ed982">GPT-3</a> and LaMDA, both are trained on huge datasets. LaMDA is more focused on conversations or dialogues as against <a href="https://anupkhanvilkar.medium.com/gpt3-a-peek-in-the-future-b40ee85ed982">GPT-3</a> , which has a lot more information to process. This definitely gives an advantage to LaMDA in being more refined and efficient in having interactive conversations, but it could possibly also mean that its knowledge base could be limited. Being a search provider, Google promises to keep the factual precision parameter of responses for LaMDA extremely accurate. It means that the focus of the LaMDA model would be to respond more accurately on any topic along with having an insightful and entertaining conversation.</p><h3>Conclusion</h3><p>LaMDA is definitely another step in taking conversational AI to the next level. With Google implementing it in their products, we can be definitely sure that these technologies are not far from reaching us. It should not be too long when we will get rid of those machine-like conversations we have with search engines or home assistants and it would be difficult to distinguish if the last conversation we had was with an AI or human.</p><p>Anup</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=452e82771f5" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>