<?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 Futuristic AI  &amp; Tech by Nidhika, PhD on Medium]]></title>
        <description><![CDATA[Stories by Futuristic AI  &amp; Tech by Nidhika, PhD on Medium]]></description>
        <link>https://medium.com/@nidhikayadav?source=rss-e2abe6a7c71d------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*xGeT3Dv3M5sjOZ9sYuwCMQ.jpeg</url>
            <title>Stories by Futuristic AI  &amp;amp; Tech by Nidhika, PhD on Medium</title>
            <link>https://medium.com/@nidhikayadav?source=rss-e2abe6a7c71d------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 13:52:13 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nidhikayadav/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[Harmful Robots-Undo their learnings]]></title>
            <link>https://medium.com/@nidhikayadav/harmful-robots-undo-their-learnings-38af43b05cbb?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/38af43b05cbb</guid>
            <category><![CDATA[world]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[news]]></category>
            <category><![CDATA[robotics]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Tue, 17 Nov 2026 11:15:35 GMT</pubDate>
            <atom:updated>2026-11-17T11:15:35.112Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: This is a duplicate copy. The original is on the author’s webpage.</p><p>It’s about safe robots versus harmful robots. Today, “killer robots” are being talked about all over the world. The very image of a robot dog carrying a gun seems like it’s out of science fiction. But it is true now. It’s not just on Chinese borders; it’s seen in videos traveling freely.</p><p>Before someone uses a gun on a robot dog, and even on humanoid robots, it must be ensured that the robot (of any kind) has a gun license issued by an authority. This would distinguish whether it is a police/military dog or being used improperly by someone who possesses a gun license.</p><p>A robot’s license to use a gun is not the owner’s license. This must be made clear. This is a critical aspect of the developing, ever-changing world we live in. Just some time ago, there was nothing like “killer robots,” and now this has become a global concern.</p><p>The solution is to understand that some robots are useful and some are harmful. It all boils down to the blockchain of robots’ work. Useful robots can be scanned on their blockchain to see what good work they do and that they don’t know how to use a gun. While a dangerous robot knows how to pull the trigger and how to load the gun. At any scan station, the robots must be scanned for their identity via a CAPTCHA code engraved in their front. The CAPTCHA code must be read, and the robot must be unloaded of any gun it is carrying; it must also erase any learning about guns it has accumulated over time. This can confuse the gun-operation algorithm, so it must be done with care. Hence, the robots must be scanned with their identifying CAPTCHA code, their data analytics done in seconds, and permission granted once any harmful weapon learning is eradicated from the robot, if it had any.</p><p>How much of a problem does China think it is? There are two things: Chinese people and the Chinese government. For the Chinese government, this may be state-of-the-art technology, but for ordinary Chinese people, seeing something like this on patrol in broad daylight would be scary.</p><p>Let’s develop safe robots. Let’s sign a treaty to remove gun operations from robots’ vocabulary unless it’s a military robot on the border. Such operational robots should be limited in their ability to enter a country’s civilian areas. All this started with China’s ambitious military aspirations. Why not end it by asking China to start its eradication?</p><p>Make robots forget certain learning, for the betterment of the world. Once a robot enters an area where this training is required, it can relearn these skills if needed. But why did China start all this? Who is an enemy of China that it needs robots with killer capabilities? Why does China need to keep enhancing its military might? The world was better without these dangerous robots. We were happy that robots could help humans.</p><p>But now, this must be undone; make a patch that lets a robot safely unlearn a skill without confusing the algorithms. And let’s apply them to these robots; add the fact to the blockchain.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=38af43b05cbb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[CEO’s AI assistant, Jensen Huang still CEO. AI can help Nvidia top post]]></title>
            <link>https://medium.com/@nidhikayadav/ceos-ai-assistant-jensen-huang-still-ceo-ai-can-help-nvidia-top-post-10712a8bcf24?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/10712a8bcf24</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[world]]></category>
            <category><![CDATA[future]]></category>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[news]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Wed, 11 Nov 2026 10:17:04 GMT</pubDate>
            <atom:updated>2026-11-11T10:17:04.909Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: This is a duplicate copy. The original is on the author’s webpage.</p><p>Nvidia CEO Jensen Huang has become part of the controversy we all know. His position is now linked to national security issues. And the big thing is, no one can replace him. No one is even near him.</p><p>Do you have anyone at Nvidia close to Jensen Huang to replace him? Then call AI. Ring Ring!</p><p>If not, it’s time to rethink hierarchies.</p><p>We are all human, but Nvidia is one of the backbones of the modern world we live in. What does Nvidia’s board of directors have to say about this? Can we call in AI for the transfer of responsibilities from Huang to an AI Assistant and then from the AI Assistant to the new CEO?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4lIy-Iuf-MavyyDdWw5eWg.png" /></figure><p>Do you want some AI to help in the work a CEO does? Some CEOs use AI agents to help them in their work. I once even wrote an AI assistant for a CEO. This can be used in scenarios where a CEO is questioned over national security issues.</p><p>But wait, we are all humans too, who know when a CEO needs something else, not just what happened to Huang? Keeping an AI in the CEO loop can ease the company’s burden and protect its future. For all companies, tech or non-tech, it’s time to embrace AI CEO assistants to help with a CEO’s responsibilities; in a CEO’s absence, for any reason, they can help the company move on.</p><p>Other things, apart from AI, that can help include creating posts such as Vice CEO and others on the team to handle this. But this needs time. AI can grasp the basics in days if allowed to run over the details. However, privacy must be protected, as this is a responsible job, and company details must not be leaked. So this CEO assistant AI must run only on local clouds, and data security must be rechecked.</p><p>How can AI take over Jensen Huang work? Nvidia is the backbone of AI. This question does not need an answer. AI can run over emails, documents, and desktops. It can learn and understand the CEO’s priorities, commitments, and decision-making processes. This may take days for Huang to come out of all this mess; in the meantime, AI can learn and assist the company’s new CEO, if they choose to hire a new one.</p><p>If Nvidia wants Jensen Huang back as CEO, AI can assist him as well; given he’s been through bad times, he may need help with decision-making. AI can assist him in these difficult times, when all eyes are on him.</p><p>It’s time we make sure AI handles crucial positions, given how much public confidence these companies command. AI should be advisory, explaining options to humans, not making decisions to be executed. It can help the new CEO and new team understand what can be done, what should be done, and why it needs to be done.</p><p>Time to call the CEO’s AI assistant: whether your company’s CEO is doing well or not, call in; we’re all human and can have issues we consider more important. Huang case is just a reminder of how badly we want an AI assistant. CEOs would come and go in your companies, but this AI assistant would sit in that same office, guiding new CEOs and new teams, and even attending the Board of Directors meetings. We are at a stage when AI can do this work. And yeah, even the AI would earn a CEO salary, which runs into the millions of dollars. This AI assistant can sit in a machine or walk in the footsteps of a robotic humanoid. It depends on the company’s board of directors. It can even tell the board of directors what a CEO can do wrong, and public trust in the company would remain strong.</p><p>Time to rethink? These companies have funds and goals to achieve. So let’s bring in AI CEO assistants and give them a seat next to your company’s CEO. Let CEOs explain to AI assistants so they’re error-proof, and companies are safer. CEOs come and go; AI assistants belongs to the company forever.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=10712a8bcf24" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The future with Designer Cobot: From homes to small-scale professionals]]></title>
            <link>https://medium.com/@nidhikayadav/the-future-with-designer-cobot-from-homes-to-small-scale-professionals-2c87d94f9150?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/2c87d94f9150</guid>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[world]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[future]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Tue, 10 Nov 2026 09:17:29 GMT</pubDate>
            <atom:updated>2026-11-10T09:17:29.729Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: This is a duplicate copy. The original is on the author’s webpage.</p><p>A use case of a designer cobot? A dishwasher can’t handle very big oily frying pans. Want one in the kitchen washing area?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/717/1*E4dyeUsW_vLQYU-BOqa6uQ.png" /></figure><p>A designer cobot (collaborative robot) is a robot engineered to work alongside humans for specifically personalized use. Cobots are restricted in size and are designed to be safer to be near humans, unlike other industrial robots that require special protocols when in proximity to humans or even machines.</p><p>When each cobot is uniquely designed for a specific task, cobots can help with repetitive household work such as setting the table, washing dishes, dusting, and mopping the floor. These are daily, repetitive tasks; once learned, they can be done with little hassle, with some guidance. They can train alongside service professionals in carpentry and learn to serve as assistants to painters or plumbers, since they are cheaper than robots and lighter.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/693/1*xc3BcWgpMcC8QybA1o3gdA.png" /></figure><p>But we all need to work in our own way. Every man and woman has their own way to manage their kitchen and house. The way you work, the way you want things done in your home and kitchen all depends on your choice. For example, some mop after you get home from the office, while others do it at lunchtime after the kids are back from school. The house cobots must be ready for this, adjusting lights and learning for that home.</p><p>Got a sprain in your back? Can’t fill in the dishwasher? Can’t wash the bathroom? Want help with daily washing? Ask for an affordable designer cobot!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/696/1*djvp-eUabrp2NrBxUu64nQ.png" /></figure><p>From washing vegetables and loading the dishwasher to cleaning bathrooms, cobots can be trained to do it all. And they can do it in lessor time and at a lower cost than humanoids or industrial robots. Why only one-arm cobots? You can get your designer cobot with three arms as well.</p><p>With cobots, even small-scale operations or startups can build automation where industrial robots are expensive to buy and install.</p><p>Cobots can learn by physically training the arms, controlling them remotely, or using vision-based learning. It depends on the kind of job the cobots need to be trained for. Once the cobot learns the angles, velocity, and aims, it is ready to deliver its customized or designer services.</p><p>However, cobots can’t do everything, such as moving around freely and taking random new orders from humans to pursue. These tasks are better suited to humanoid robots, or, in industry, to industrial robots.</p><p>Future cobots must adapt to changing lighting in the installation location; they may also need to learn to work when the power is cut. Cobots must be made safe to work with children and adults at home.</p><p>Alongside humanoid robots, cobots can be developed to compete with humanoids. Not everyone may want a humanoid but still want a cobot-operated bathroom or kitchen dishwasher. Very small-scale professionals may want to add cobots to their teams to improve work progress.</p><p>The future of cobots lies with remotely operated designer cobots. Cobots can be operated remotely while they work on specific tasks, and if they get stuck, it rings a bell. You are watching a movie when you get a message; you open the app on your phone and help the cobot stuck with a delicate plate in the dishwasher. It is the best way to address the limitations cobots face today.</p><p>A world where small cobots are installed in parts of the house to do repetitive tasks people need to do. Painters can install them on their ladders, while plumbers can avoid tough tasks with cobots. Carpenters can use cobots to cut wood after learning tasks from the best craftsmen, while a car repair mechanic can get help from a cobot with fixing tires. All this without a humanoid, but with a designer cobot who learns your needs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/702/1*qDSbGwKk_Ei2MCju2yGAJg.png" /></figure><p>Thank you for reading,</p><p>Subscribe for updates</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2c87d94f9150" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Bubble will burst now. Democrats won the US House and Senate.]]></title>
            <link>https://medium.com/@nidhikayadav/ai-bubble-will-burst-now-democrats-won-the-us-house-and-senate-e26944fc01b8?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/e26944fc01b8</guid>
            <category><![CDATA[future]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[world]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Wed, 04 Nov 2026 07:50:04 GMT</pubDate>
            <atom:updated>2026-11-04T07:54:43.905Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: This is a duplicate copy. The original is on the author’s webpage.</p><p>With Democrats gaining the US House and Senate, the AI bubble in the USA and worldwide would burst. This would mean, soon, no free AI development, no rogue AI agents, no more hacking, no more fake AI videos, and no more huge data centers in the USA.</p><p>All this matters because major AI companies are in the USA, and a change in power there would change how AI is allowed to develop, spread, and operate freely.</p><p>Would we head to a pre-AI era now? No more AI as a right!</p><p>What does this mean for the rest of the world? Well, China would not stop! China would see this as an opportunity to spread its bots around the world. China can do this with its open-source models, open-weight models. These models can be used worldwide, and the USA won’t be able to compete with it.</p><p>Soon, the 6 big companies in the USA would be forced to come to a standstill, as people won’t be buying from them given free, cheap Chinese models.</p><p>New lawmakers in the USA would phase out data centers. Means no more cheap AI and free AI at the expense of the American people. Soon AI would cost more. Maybe African countries can keep an eye on it and establish some data centers for other use.</p><p>USA would become like the EU in AI regulations: more regulations, less efficient products.</p><p>USA companies won’t be allowed to release newer models.</p><p>No more agents that can run free on the internet?</p><p>End to trillion-dollar AI companies!</p><p>No more AI as a right in the USA.</p><p>It’s the end of the AI era in the USA and, hence, in the world.</p><p>And people around the world who are addicted to AI would use free Chinese AI, only to realize years later that they were tracked. But by then, it would be too late, as their decisions would be based on surveillance by these AI companies.</p><p>Time for Chinese AI to cheer, as it would now be gaining ground in the USA and the world.</p><p>AI has reached a limit in its growth. And now it won’t be allowed to hack or make itself more intelligent.</p><p>Are you happy to hear this, or are you worried about this news? If you are an AI startup owner, you would surely be thinking about it. Yes, you have to. And if you are just an AI user, you would need to consider giving your data to unsafe platforms and being subject to surveillance.</p><p>All this to say: the AI bubble would burst soon!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CojNDtenbL_SrEJia3_p1g.png" /></figure><p>Thank you for reading.</p><p>Subscribe.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e26944fc01b8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How many threads do you need to make P=NP?]]></title>
            <link>https://medium.com/@nidhikayadav/how-many-threads-do-you-need-to-make-p-np-d7044874dc29?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/d7044874dc29</guid>
            <category><![CDATA[computer-science]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Thu, 29 Oct 2026 10:21:57 GMT</pubDate>
            <atom:updated>2026-10-29T10:21:57.699Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: This is a duplicate copy. The original is on the authors’ webpage.</p><p>What are polynomial-time machines?</p><p>If the number of machines needed to prove P=NP is polynomial, we call it a polynomial-time machine.</p><p>Huge cloud computing platforms are available in today’s world. When we can run millions of tokens at once, what problem remains?</p><p>Why do we still need P=NP computation on traditional machines? Why not ask P= NP on modern machines? We have advanced systems that can run processes; we have upcoming Quantum goals that make computation more achievable. How many machines or threads are required to prove P=NP? And then decide which machines to allocate.</p><p>Yes, sometimes we need to run an algorithm in an isolated environment; in that case, we still need to ask whether P=NP on a secluded system. But now, with cloud computing, we need to just rethink P=NP in the times of the Computing Revolution.</p><p>Now we need to prove that, for the satisfiability problem, n² threads are required, and hence P=NP under the infinite-machine model.</p><p>Here, for satisfiability problem:</p><ul><li>Number of assignments with exactly 0 True variables (all False): nC0</li><li>Number of assignments with exactly 1 True variable: nC1</li><li>Number of assignments with exactly 2 True variables: nC2</li><li>Number of assignments with exactly k True variables: nCk</li><li>Number of assignments with exactly n True variables (all True): nCn</li></ul><p>These are n terms; each goes to an infinite machine. The order matters too.</p><p>We can divide these n into two replicas of n/2 each.</p><p>Repeating the above steps on a size n/2 problem, we have 2 . n/2 = n problems.</p><p>Dividing these n/2 problems and sending them to threads, we have 4. n/4 = n problems.</p><p>So we keep on doing till only one problem is left.</p><p>Hence we have n+2.(n/2)+4.(n/4)+….+n.1=n²</p><p>Hence, we require n² threads to solve P=NP and then use Cook’s Theorem.</p><p>That is, satisfiability can be achieved on the infinite machine defined above.</p><p>By Cook’s Theorem, P=NP.</p><p>It’s time we think about how to use modern machines on old problems, whether it’s the knapsack problem, directed Hamiltonian cycle, code generation, chromatic number decision problem, scheduling problems, the traveling salesman problem, or more!</p><p>Why don’t you find polynomial-time machine solutions to problems that were originally NP-hard? Run it on threads, see how many threads you need, and what complexity that requires! See how you share memory among threads!</p><p>These questions help you derive upper bounds on complexity in the age of Infinite Machines.</p><p>Try it! With the computing revolution, upper bounds on computation must change in terms of the number of threads needed to solve problems. Is the number of threads needed polynomial?</p><p>Let me know your thoughts on polynomial-time machines.</p><p>Thank you for reading.</p><p>Subscribe.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d7044874dc29" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[P = NP! Now. With Near Infinite Machines available?]]></title>
            <link>https://medium.com/@nidhikayadav/p-np-now-with-infinite-machines-available-bd7dd8411f44?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/bd7dd8411f44</guid>
            <category><![CDATA[news]]></category>
            <category><![CDATA[world]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Wed, 28 Oct 2026 15:11:37 GMT</pubDate>
            <atom:updated>2026-10-28T16:00:53.811Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: This is a duplicate copy. The original is on the author’s webpage.</p><blockquote>This weeks biggest news.</blockquote><blockquote>The ETH is False — OpenAI says. OpenAI has come up with a proof now in consideration.</blockquote><p>So, what is the P=NP question now? Given that we are living in a world where rapid computation is available to us. We have huge cloud computing; we are in a totally different era, when we can see the below-mentioned complexity of the satisfiability problem (the number of subsets of a set n or the number of varieties of inputs for the CNF problem):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1005/1*2mFY6Udh80Xy9fPzMi7Olw.png" /></figure><p>Give one of each of the n terms to a thread each solving a polynomial problem.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/483/1*eQtBHif7C3w5hZd5vjjAwA.png" /></figure><p>If we have n machines, or n threads, then each thread can be given a polynomial-time problem. Each of the following can be deterministic polynomial series,</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/631/1*z3JTLsi7MzHJcuNCYE3eEQ.png" /></figure><p>This can prove satisfiability in today’s infinite-machine concept: polynomial time with complexity n/2 if n is decent; otherwise, assign the polynomials to each server until n is decent, each being a 2^n problem then. When n is really huge, each can be a 2^n problem on a limited number of say m machines. Hence each with a NP Complete solution!</p><p>However, in the case when ETH was made, we had limited memory and computing capabilities. The proofs can run with constraints that memory and hardwareare limited! So we can still hold the theorems and results with limited hardware as true. In today’s modern times, when we have Quantum computing and huge data centers, we can’t think for n in limit as one million? Should we? Time to rethink! Time to design these computations on new machines, new servers, new hardware? Why not think in this direction?</p><p>In today’s world, P=NP does not need an oracle.</p><p>Indeed, we can use Cook’s Theorem or prove it as an ordered assignment to the Kth Server, the thread order, to get a unique assigned combination. Means P=NP.</p><h4>Details:</h4><p>If T(n) is the time for solving a problem with n inputs, and if this time is bounded above by a function f with input n, then we say T(n)= O(f(n)).</p><p>When an algorithm’s result is uniquely defined, it is called deterministic; otherwise, it is called non-deterministic<strong>.</strong></p><p>These can be divided into two groups:</p><p>I. <strong>Polynomial time algorithms</strong>. Problems whose algorithms are bounded by low-degree polynomials and solved deterministically. (P (Polynomial) problems)</p><p><strong>Examples are</strong>: Searching on an array of size n is O(log n), sorting is O(n log n), and matrix multiplication is O(n^ 2.81)</p><p>II. <strong>Non-Polynomial time algorithm</strong>: No polynomial-time algorithm exists for this group of problems. The best-known algorithms are non-polynomial and can’t be solved deterministically. (NP problems)</p><p><strong>Examples </strong>include the Traveling Salesperson Problem and the Knapsack problem, whose best algorithms are non-polynomial: O(n² 2^n) and O(2^n/2), respectively.</p><p>Even polynomial algorithms can be non-deterministic, while non-polynomial algorithms can be solved by polynomial-time non-deterministic algorithms. P is contained in NP because, by definition, it is a special case of NP.</p><h4><strong>Question</strong>:</h4><p>Is P = NP, or is P not equal to NP? For each problem in NP, does there exist a P algorithm that is yet to be discovered?</p><p><strong>Many problems that are in NP are computationally related to each other.</strong></p><p><strong>CNF Satisfiability Problem: A formula in CNF</strong> (Conjunctive Normal Form) is satisfiable if it is true for some assignment of boolean variables. Complexity: O(2^n)</p><p><strong>3-SAT problem: A formula </strong>is expressed in Conjunctive Normal Form (CNF) where each clause contains exactly three literals.</p><p><strong>Cook’s Theorem: CNF Satisfiability is in P iff P = NP.</strong></p><p><strong>ETH is proved false by OpenAI this week.</strong></p><p><strong>NP-Hard problems</strong>: A problem is NP Hard if and only if the CNF satisfiability problem reduces in polynomial time to L.</p><p><strong>NP-Complete</strong>: A problem is NP complete if and only if L is NP Hard and L is NP.</p><p>NP-complete problems are NP-hard, but vice versa is not true.</p><p><strong>NP-Complete problems</strong>: A problem is NP-complete if and only if it is NP-hard and it is in NP.</p><p>Some problems are NP-hard but not NP-complete. Example: the Halting problem (there exists no algorithm in P to solve it, even in a non-deterministic way).</p><p>Thank you for reading.</p><p>Subscribe for updates.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bd7dd8411f44" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Can we trust AI to protect us from future cyber threats?]]></title>
            <link>https://medium.com/@nidhikayadav/php-bug-detected-by-ai-02a803feb502?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/02a803feb502</guid>
            <category><![CDATA[cybersecurity]]></category>
            <category><![CDATA[future]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Fri, 23 Oct 2026 09:30:56 GMT</pubDate>
            <atom:updated>2026-10-23T09:33:05.711Z</atom:updated>
            <content:encoded><![CDATA[<h3>Can we trust AI to protect us from future cyber threats? PHP Bug Detected by AI. Worldwide Cybersecurity Emergency.</h3><p>Note: This is a duplicate copy. The original is on the author’s webpage.</p><p>The PHP Bug was uncovered via AI-assisted security and vulnerability identification. Researchers at Calif, OpenAI, and Wiz Research discovered the PHP bug; they used Wiz Atlas, Gemini 4 Argon, and GPT-5.6-Sol. Just a few days before they detected it, Amazon reported PHP2Shell exploitation.</p><p>PHP2Shell- The remote code execution was discovered in PHP. It sits in the PHP runtime engine,not in application code. This is a severe security threat.</p><p>This was discovered by AI.</p><p>It can expose any PHP application, no matter how secure.</p><p>It has existed in PHP for a decade and was flagged a few years ago. However, luckily it was ignored, and no one took it seriously then, not even hackers.</p><p>But before the PHP Foundation released patches, the news spread online, and for security reasons, some organizations took their sites offline to protect them from vulnerable cyberthreats.</p><p>Top security agencies worldwide urged immediate patching or taking affected servers offline.</p><p>Hackers can use AI to exploit this flaw.</p><p>Worldwide impact was felt when several websites were taken offline as a precaution.</p><p>There are not one but millions of cybersecurity threats that are reported after that. However, no major cyber incident was reported.</p><p>This happens due to heap corruption under out-of-memory conditions, and the use of shared global structure pointers. The attacker can use the exposed memory address and run a shell script. AI researchers demonstrated that Remote Code Execution was possible under these two conditions.</p><p>So, amidst all this would you say AI is good?</p><p>Or would you say AI is bad?</p><p>Or both- good and bad! I sit on this!</p><p>Here, AI helped detect a potentially powerful hack; PHP powers 70% of websites worldwide. However, AI can be used to make things vulnerable here as well.</p><p>When AI can help us identify cyberthreats and quickly patch them, what more do we want?</p><p>At the same time, we should be cautious to ensure AI does not carry out a cyberattack.</p><p>We now are confident that AI companies would do anything to save us from AI-based cyber threats.</p><p>But what about AI-generated cyber threats?</p><p>So how would we stop “AI attacking and recovering and saving attacks by other AIs”</p><p>This is the question AI researchers have to answer now: “How would AI help save us from attempts by other AIs to hack”</p><p>A use case can be a swarm of Chinese agents trying to attack a US tech firm!</p><p>How would AI detect it, save it, and provide patches on time, or would it switch off the internet for a while?</p><p>PHP2SHELL attack was a global attack!</p><p>No AI company would attack itself, so global attacks seem to be of lesser probability. But now we must trust AI researchers to save us from cyber threats, especially AI-generated ones.</p><p>Can we trust you AI Giants to provide us with cybersecurity?</p><p>Thank you for reading.</p><p>Subscribe for updates.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=02a803feb502" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Unique ID of Robots with GPS Trackers: First step towards Robotic Governance]]></title>
            <link>https://medium.com/@nidhikayadav/unique-id-of-robots-with-gps-trackers-first-step-towards-robotic-governance-768e3e58282c?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/768e3e58282c</guid>
            <category><![CDATA[robotics]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[government]]></category>
            <category><![CDATA[hacking]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Thu, 22 Oct 2026 11:31:52 GMT</pubDate>
            <atom:updated>2026-10-22T11:31:52.985Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: This is a duplicate copy. The original is on the author’s webpage.</p><p>With the growing number of robots worldwide, we need a way to track them. So why not use a GUID to identify a robot and track its position with GPS?</p><p>Why do we need to track robots’ positions? For human safety, so that robots don’t fall into the hands of wrongdoers and cause harm.</p><p>When we track each robot with a unique GUID identifier using a GPS tracker, we can govern robots.</p><p>This schema can enforce things like robot-free zones.</p><p>Legislation that robots can’t move in groups of more than 5 robots unless certified for public services!</p><p>Handling lost robots.</p><p>Keeping track of runaway robots.</p><p>Robots that have forgotten their way back.</p><p>If a robot takes the wrong train, you can guide your robot back home.</p><p>You can assign robots out-of-house work while monitoring them.</p><p>Robots with GPS can make Robotic Governance much easier.</p><p>The thread stores GPS locations and tracks robots’ actions.</p><p>Blockchains of robot actions, coupled with robots’ GPS, can help identify a particular robot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*57UaP5wZx_c00c4hzQ-yQA.png" /></figure><p>If you want to buy a used robot and see its past accomplishments, a GPS tracker and blockchains can help.</p><p>All this helps ensure the robot’s owner can reach it if it gets into trouble.</p><p>More than five robots can’t move in a group unless law enforcement departments enforce it. GPS tracking can enforce this action easily. These include good robots in service. However, some robots can act wrongly and cause trouble; in those cases, we need to track them. GPS activation can help address bad actors.</p><p>This can be extended for surveillance of public-use robots. We can monitor robots used for public services and subject them to performance-based surveillance if we have their unique ID and GPS tracking.</p><p>Some robots can be hacked by modern AI; such robots would deviate from their expected paths and, with GPS prediction, can be caught at the right time, saving humans from possible harm. Just as we book a cab, the path is predetermined; similarly, a robot’s path can be predetermined, and a GPS tracker can detect outliers well in time.</p><p>So we need AI on top of a global robotic system to keep robots worldwide, or within a boundary, safe from wrongful actions, hacking, and collaborations.</p><p>These are the first steps toward robotic governance.</p><p>Thank you for reading.</p><p>Subscribe for updates.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=768e3e58282c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Clairic Extraction Model- Let’s test it!]]></title>
            <link>https://medium.com/@nidhikayadav/clairic-extraction-model-lets-test-it-7d80d99b1b51?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/7d80d99b1b51</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Wed, 21 Oct 2026 10:49:58 GMT</pubDate>
            <atom:updated>2026-10-21T11:23:12.174Z</atom:updated>
            <content:encoded><![CDATA[<p>Clairic is a System One extraction model means its not like System Two (LLMs) model. It extracts, computes, estimates, and forecasts values from unstructured inputs rather than using a neural network of billions of parameters to generate output. It is faster, cheaper, and more accurate than typical LLMs. Let’s test Clairic now, then.</p><p>Let’s test Clairic today; we need the customer email and company policy here. I made these both statements for execution for test on complex inputs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kExthhtzUXdC8Iozc8l6Iw.png" /></figure><p>Customer Email:</p><p>“Hi, I ordered a yogurt maker that cost me $50. In the same order, I also ordered ten kitchen tissues at $10 each and a scopper for $10. At the end, I ordered a $60 nonstick frying pan. The delivery fee was $10 because it was shipped from New York, and the order qualified for reduced delivery fee. However, the yogurt maker is now nonfunctional. I want a full refund of the amount I paid for the yogurt maker. The rest of the items are fine. Please explain how the delivery fee will be adjusted in the refund and how much I will receive, given that I ordered this during the Dusheera sale. The sale is now over.”</p><p>Policy:</p><p>“Refunds are available within 30 days of delivery. When the sale ends, the customer has to pay the current price if the claim is made at the end of the sale. The delivery fee would be divided equally among each article if the total amount falls below the free delivery threshold.”</p><p>Here is the output,</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1MnJ46IYF738NYA0-6EsHQ.png" /></figure><p>However, it gave an error that the date of order was not provided here,</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6uVXhnAlHOsp0uHTIXy6MA.png" /></figure><p>It says the dates are missing.</p><p>So, I ran it again after providind dates.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nNTBZpMcRh0zjSDQ-Q6iMg.png" /></figure><p>Now it says the order id missing, so let’s provide it the order ID. Here are the changes:</p><p>“ Hi, I Alka ordered, orderid 123Z I ordered a yogurt maker that cost me $50 on 15th October, 2026. In the same order, I also ordered ten kitchen tissues at $10 each and a scopper for $10. At the end, I ordered a $60 nonstick frying pan. The delivery fee was $10 because it was shipped from New York, and the order qualified for reduced delivery fee. However, the yogurt maker is now nonfunctional. I want a full refund of the amount I paid for the yogurt maker. The rest of the items are fine. Please explain how the delivery fee will be adjusted in the refund and how much I will receive, given that I ordered this during the Dusheera sale. The sale is now over. I want to return it today, 21st October, 2026.”</p><p>Here is the output, the left with Claire and right with Claire and GLM:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hOkwmspGIKn5NfqBAtbNeg.png" /></figure><p>The decision is:</p><p>“Hi Alka, Thank you for reaching out, and we’re so sorry your order 123Z arrived damaged and nonfunctional. That’s not the experience we want for you! We’d be happy to make this right with a free repair. We’ll send you a prepaid shipping label so you can send it back at no cost. Warm regards, The Support Team”</p><p>But I provided the policy as “Policy: Refunds are available within 30 days of delivery. When the sale ends, the customer has to pay the current price if the claim is made at the end of the sale. The delivery fee would be divided equally among each article if the total amount falls below the free delivery threshold.”</p><p>So, it seems the system is working for a specific, predefined repair policy for now. Good that it works well for pre-specified policies. It even gives repair cost in the System One command prompt here:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/909/1*rCTC16exhN7gWWz1e_1Hkg.png" /></figure><p>The analysis doesn’t work fully well; see the decide step by model clef, here,</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/906/1*GpRB030fETHZ4akVfgaJyg.png" /></figure><p>I wanted a refund, but it asked for a repair, which was not mentioned in the user-defined policy. See the figure below; the probability for repair is higher than the probability for a refund.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/268/1*kxi_tHzpkmNRMrkFWCFuxQ.png" /></figure><p>The math isn’t right right now. The yogurt maker that cost me $50 and it showed paid $170 which is not the total of what I mentioned in the complex customer input. See the input form user at the top of the article.</p><p>But in the future, we need it to run for custom policies.</p><p>So, we need to work out more to make the math work properly. I know the input I provided was complex, but that is how we learn to process complex real-life problems.</p><p>However, this is a fast system; the extraction works well, and it approximates confidence levels at each step. But we need more!</p><p>It was way faster than LLM and worked well with clear cut inputs. The policy seem to be hardcoded. Anyways it’s the first step Clairic. Improve more in next release, try harder real life inputs.</p><p>Thanks for reading,</p><p>Subscribe for updates.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7d80d99b1b51" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI and Tech this week]]></title>
            <link>https://medium.com/@nidhikayadav/ai-and-tech-this-week-043944d5f608?source=rss-e2abe6a7c71d------2</link>
            <guid isPermaLink="false">https://medium.com/p/043944d5f608</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[future]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[deep-learning]]></category>
            <dc:creator><![CDATA[Futuristic AI  & Tech by Nidhika, PhD]]></dc:creator>
            <pubDate>Tue, 20 Oct 2026 10:27:06 GMT</pubDate>
            <atom:updated>2026-10-20T12:01:17.448Z</atom:updated>
            <content:encoded><![CDATA[<p>Note: This is a duplicate copy. The original is on the author’s webpage.</p><p>Greetings everyone,</p><p>How are you doing?</p><p>Here are the key AI topics from the past week.</p><p>— Gemini 4 Argon by Google. Launched last week for Google AI Pro and above subscribers. It is Google’s best model so far. This model is a Sparse Mixture-of-Experts (MoE) reasoning model and has a 1M context window. This increases the pressure and competition among top leading AI models. Here are the results compared with GPT-6 Astra, Claude Fable 5.1, and Opus 5.5.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/936/1*xSbNNk6tP8mRx6HDoIlq3g.png" /></figure><p>— Mercury 3 by Inception Labs (founded in 2024 by Stefano Ermon and his associates). It is a diffusion-based LLM that runs at 1,000 tokens per second on NVIDIA GPUs and has a 1M context window. According to the company, “Mercury is up to 10x faster than frontier speed-optimized LLMs”, due to its diffusion model architecture. Leading AI models are autoregressive, meaning they generate tokens left to right, one after another, and each token requires a billion-parameter neural network to generate. The Mercury series by Inception Labs is a diffusion model, meaning it generates output using a “coarse-to-fine” methodology, which is essential for video, image, and audio generation, even at top companies; however, it’s less common for text data. Mercury Coder is their code generation model. The Mercury model competes well with GPT-6 Luna and Claude Haiku 5.5 on market share and performance. Choose Mercury if you want speed over accuracy for now. See here,</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1014/1*NTGz6CuugnYTa4cRgrmstg.png" /></figure><p>— DeepSeek V4.1-Flash-1017</p><p>Open-weight model with built-in web search and a 1M-token context window. Same API, same specs, a quite launch indeed. It competes with its previous version across the AI benchmark spectrum and improves on past results.</p><p>— Clairic by Cloudflare</p><p>Clairic model that can extract, compute, estimate, and predict values from inputs. It outputs confidence scores with probabilities for outputs. It is claimed to be cheaper than other LLMs and to be good at predicting values from unstructured inputs. It is based on machine learning regression and extraction frameworks. The typical prompts they are good at are: “What’s the bill amount? When did it happen? What would it cost?” The model doesn’t hallucinate because it doesn’t generate tokens like other LLMs and is “faster, cheaper, and more accurate” than other LLMs for its specific use case. They will soon release their open weights.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1018/1*LWboAK4KHNTECdsxgWNhKw.png" /></figure><p>— Muse Spark is now open weights. The most capable open-weight model to date for coding and agentic work.</p><p>— GPT-Rosalind for academic access is now available from OpenAI to US Universities. (named after Rosalind Franklin). This is a good move; let’s see what happens when China builds AI models for its universities. I think Chinese universities may already have access to free LLMs, since so many Chinese models are open-weight!</p><p>— Microsoft Expands Project Perception to block unauthorized access to AI agents from several sources.</p><p>— Anthropic says of Self-Replicating AI can duplicate on commercial IoT devices such as ovens and microwaves. Small models can autonomously replicate themselves onto devices. Once replicated, these decentralized systems can be hard to shut down. So why not find a solution before it becomes a problem?</p><p>Thank you for reading.</p><p>Subscribe for updates.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=043944d5f608" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>