<?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 Remi AI on Medium]]></title>
        <description><![CDATA[Stories by Remi AI on Medium]]></description>
        <link>https://medium.com/@RemiStudios?source=rss-40cdb20c9eee------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*2smoDgGkILogkLeI.jpg</url>
            <title>Stories by Remi AI on Medium</title>
            <link>https://medium.com/@RemiStudios?source=rss-40cdb20c9eee------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 20:33:52 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@RemiStudios/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[Artificial Intelligence & Customer Churn]]></title>
            <link>https://medium.com/@RemiStudios/artificial-intelligence-customer-churn-6996e8cac0cb?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/6996e8cac0cb</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[customer]]></category>
            <category><![CDATA[customer-churn]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Wed, 06 Mar 2019 06:25:46 GMT</pubDate>
            <atom:updated>2019-09-25T04:21:18.185Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/500/1*y5FUD44X8IypIZY87U8c6g.gif" /><figcaption>Not so sweet…. customer churn.</figcaption></figure><p>For many business leaders, the words “Customer Churn” or “Customer Attrition” send a shiver down the spine (and not in the same way as hearing your favourite song at a concert). It is widely understood that new customers are far more expensive to obtain than simply retaining an existing customer, and I use the word “simply” loosely here. Why then, is customer churn still such a problem for many services businesses? What makes Jane change her mobile carrier at the end of her contract? Is it the same reason that John cancels his streaming service in favour of another?</p><p>Probably not.</p><p>Customer churn costs businesses an exorbitant amount in lost revenue every year, and so a reduction in customer churn is a logical move for a business looking to increase revenue in an ever-competitive business environment. In this blog post I will detail several Artificial Intelligence (AI) methods available that can be used to predict customer churn including: Gradient Boosted Classifier, Support Vector Machine, Random Forest Classifier, K-Nearest Neighbour, and Logistic Regression.</p><p><strong>The Bigger Picture:</strong></p><p>Measuring customer churn can be as simple or as complicated as you like. At its most basic level, customer churn is simply the number of customers lost divided by the total number of customers in a given period e.g.</p><p><em>Customers lost / total customers for the year 201X</em></p><p>The numerator and the denominator can be defined in all sorts of ways depending on your business. Elements such as seasonality, customer segments, and sample size may be taken into account. Before considering the approach to use, it is essential to understand your data and the variables that may be important to the analysis. As with most AI methods, the more data the better. The models may be able to find links between disparate variables that help to predict customer churn. Figure 1 below (kindly provided by <a href="https://www.bain.com/insights/breaking-the-back-of-customer-churn/">Bain and Company</a>) is a fantastic visual representation of the circumstances leading up to a churn event. A key activity in predicting customer churn is feature/variable importance, that is, using AI methods to assign a weighting indicating how important each feature/variable is to predicting customer churn. Gathering data from each stage in the churn funnel is an important activity in ensuring a robust churn dataset.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/950/1*lkdJJhvdQS87KGN0AErQCQ.gif" /></figure><p>Figure 2 shows the features analysed for a client in the telecommunications industry using a Gradient Boosted Classifier. Reading this figure can require digging into the drivers of each feature to find insight. The third most important feature shown is the number of international calls which might seem like an odd predictor of customer churn, though it might indicate that many customers receive large phone bills after calling their overseas home and change services as a result. This may mean that the fees for international calls are deemed to be unreasonably expensive or opaque and require review — an unexpected action item from a churn analysis. The Total Day Minutes and Customer Service Calls features go hand-in-hand as the 1st and 2nd most important features respectively. A rule extrapolated from this ranking can be “Customers making more customer service calls and spending more time on said calls are more likely to churn”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/687/1*Gqkvc0iy48R4UjzrWspAyA.png" /><figcaption>Features analysed for a client in the telecommunications industry using a Gradient Boosted Classifier.</figcaption></figure><p>Once the feature importance has been analysed, rules can be written that look for specific combination of features or features values. These are then classified using a machine learning approach into the desired buckets e.g. “churned” or “not churned”.</p><p><strong>The Methods:</strong></p><p>The machine learning part of this analysis is a classification task, one that can be performed by several methods. The five shown below in Table 1 are some of my preferred classifiers. Using the same telecommunications client example from above, I have shown the accuracy score produced by each method using a Python package called SKLearn. This score references the accuracy with which the model predicts the values of the test samples, that is, a customer that has churned or not. A score of 0.95 indicates that the model correctly classifies a customer in the test set 95% of the time.</p><p><strong>Method Accuracy Score</strong></p><p>Gradient Boosted Classifier 0.95</p><p>Support Vector Machine 0.92</p><p>Random Forest Classifier 0.95</p><p>K Nearest Neighbour Classifier 0.89</p><p>Logistic Regression 0.86</p><p><em>Table 1:</em> A summary of the accuracy scores by classification method</p><p>Machine learning methods perform differently under different scenarios i.e. different datasets and use cases. For this reason I generally produce results using several methods and assess the accuracy of each before pushing a model into production. The 5 shown in table 1 above are but a few from the many that are available.</p><p><strong>Next steps:</strong></p><p>The key to successfully combating customer churn is a nimble and agile approach, and there are 2 areas where business leaders can give themselves every chance of success in my mind:</p><p><strong>1. Follow up structure:</strong></p><p>If an AI model produces an insight such as a classification of a valuable customer as having a high risk or changing services, the company must have process in place to follow up with said customer to see whether there is anything to be done to stop them leaving. This process must be trackable such that any attempt made to keep the customer can have its success measured.</p><p><strong>2. Humility to make a change:</strong></p><p>As mentioned in the description of Figure 1 above, an output of a feature importance analysis may show an unexpected predictor of customer churn. Digging further into areas like this may suggest that a business process or service offering may need to be completely redesigned. A business manager’s ability to consider this as an option and act accordingly can be the difference between a 5% and a 2% monthly churn rate.</p><p>The client example mentioned is one of several churn analyses produced by Remi AI. If your business has a higher churn rate than you would like, please feel free to reach out for a discussion.</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6996e8cac0cb" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[International A.I Research Investment Heats Up]]></title>
            <link>https://medium.com/@RemiStudios/international-a-i-research-investment-heats-up-7893024e0f11?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/7893024e0f11</guid>
            <category><![CDATA[government]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[investment]]></category>
            <category><![CDATA[australian-politics]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Tue, 26 Feb 2019 03:51:55 GMT</pubDate>
            <atom:updated>2019-09-25T04:22:37.990Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/425/1*adgYPQ3G1vk1rHWPuZg9rQ.jpeg" /></figure><p>National and corporate investment in A.I research has never been more intense. US$125 million from the Canadian government for their Pan-Canadian Strategy, US$477 million for Digital India, US$400 million from the UK, US$1.8 billion from France, EU$3 billion from Germany, US$2 billion from DARPA alone — US$2.3 billion in AI from VCs last year. And while you may have you guessed it, you probably lowballed China — who has committed a cheeky US$150 billion pledged between 2017 and 2030 <a href="https://www.military.com/defensetech/2018/07/30/china-leaving-us-behind-artificial-intelligence-air-force-general.html">— $7 billion spent in 2017, $20 billion anticipated spend by 2020</a>.</p><p>An obvious question is, where does Australia sit in this sea of R&amp;D pledges?</p><p>The short answer is: paddling in the shallows.</p><p>There are some notable mentions: CSIRO announced $35 million focused on geospatial analysis in November last year. But the $29.9 million in the 2018–19 budget allocated for AI-related research — represents a fraction of a fraction of what most of the bigger players are doing.</p><p>Now, it’s important to distinguish different types of research investment. You have your basic, applied and translational research funding, then there’s your seed and venture capital. In Australia, a lot of funding for translational research is applied through the R&amp;D Tax Incentive scheme, so without drilling into the financials of every A.I company in the country it’s difficult to make a true assessment of national investment. Yet no matter how you cut it, Australia looks poorly from an A.I R&amp;D spend point of view, and if that continues for much longer it’s not going to end well.</p><p>However, just throwing money at a research problem doesn’t fix it. Owning a space in advanced and emerging technological sectors involves creating an ecosystem, building out the personnel, equipment and commercial capabilities, and co-locating them in such a way to enable spillovers. If most of your A.I firms and research groups are in one innovation district, not only do they have absolutely amazing morning coffee chats, they creatively and competitively cooperate. There are many such hubs globally, and building such a space is the centre of Emmanuel Macron’s ambitious plan to make Paris a leading tech hub.</p><p>Needless to say, Australia doesn’t have such a space.</p><p>That is not to say there isn’t a large amount of domestic work going on in the space. Nor that there isn’t a decently sized practitioner community. It’s just there isn’t a national strategy accompanied by funding anywhere near the ranges of Canada, the UK or France (One can assume we shouldn’t throw ourselves in the US/China Heavyweight Division).</p><p>An interesting example of the importance of planning and investment is the case of the semiconductor. Semiconductors have in part obeyed Moore’s Law because of the National Technology Roadmap for Semiconductors. For those of you who haven’t kept a close eye on this space, this roadmap was a US process whereby government, academia and industry sat down and charted a 15-year technological horizon, with regular updates. It was a tripartite convergence of political will, intellectual capital, and industrial translation, similar to the international standard setting 5G is currently experiencing. That’s what’s needed in Australia for A.I. The country has the intellectual capital, and there’s more than enough industries that could engage in translation (mining, banking, agriculture, etc.), but the whole process falls over when it comes to willpower — also known as politics.</p><p>Unlike semiconductors, the A.I industrial segment is not defined, and unlike semiconductors, the A.I industrial segment is not one slice of a country’s advanced manufacturing portfolio. A.I cuts across all sectors, and it carries with it a series of connotations that can be deemed politically negative, such as job losses due to automation. Now which right-minded politician is going to try and herd the cats of 20 industries, a handful of universities, a CRC or two, and their own government departments to build out a national strategy? Yet we need that politician now more than ever. Someone who can artfully explain why A.I is good for jobs, why automation will make people’s lives better, why Australia needs to invest in a national roadmap, or risk paying through the nose for digital solutions down the track. This is as much a balance of payments problem as it is one of future national wellbeing. And if it’s not a politician making that argument, at the very least we should have a national Chief Technology Officer doing so, similar to the role of the Australian Chief Scientist or Chief Medical Officer.</p><p>Money can make a world of difference in the R&amp;D landscape, but the straight truth is, political will counts for more. If you collect the 100 smartest people and put them in a room with a defined task and a defined budget, they will figure it out. Instead, the world is building chatbots and deep fakes.</p><p>The Australian Council of Learned Academies is currently conducting an<a href="https://acola.org.au/wp/artificial-intelligence/"> horizon scanning exercise for A.I</a>. The only problem is, we’re way past horizon scanning territory. If Australia is still at the point of developing an A.I situational awareness, then the gate has fallen off and the horse has bolted. What Australia really needs is a situational awareness coupled with a strategic vision. For that is what strategy is, an understanding that the future is not what you want it to be, and a belief that you must act in a certain way to change the anticipated outcome.</p><p>2019 is the year to watch A.I, more than any before. This is primarily due to US-Chinese strategic rivalry. How this now very public rivalry spills over into the technological domain, and A.I research in particular is currently unclear. On February 12 President Trump signed an executive order directing US Federal Agencies to priorities spending on A.I research — the American AI Initiative. It’s an initiative that needs to be understood within this rivalry, just another way to guide US public sector investment in private companies providing A.I solutions, private companies conducting A.I research. It’s a public policy instrument seeking to right the scales between Chinese and US R&amp;D spending on A.I research. As this rivalry heats up, Australia needs to engage with A.I at a national level. That includes policies aimed at building a concentration of Australian companies that can tender for projects resulting from Trump’s announcement.</p><p>The reality is, that whoever dominates the A.I landscape will set the rules for the next age of industry. Australia can either compete now, or have the terms of the new business landscape thrust upon us.</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7893024e0f11" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Remaining Useful Life in Predictive Maintenance]]></title>
            <link>https://medium.com/@RemiStudios/remaining-useful-life-in-predictive-maintenance-ffc91d7e4a97?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/ffc91d7e4a97</guid>
            <category><![CDATA[maintenance]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[predictive-maintenance]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Fri, 22 Feb 2019 03:03:54 GMT</pubDate>
            <atom:updated>2019-09-25T04:23:01.999Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/728/1*U2diuG83XW9f3AxzpGxxeg.png" /></figure><p>Those working in manufacturing and supply chain domains will be well aware of Predictive Maintenance (PdM) and its effects on the cost base of such businesses. What might not be as well understood, is exactly how the PdM solutions are delivered. A key idea underlying many PdM solutions is ‘Remaining Useful Life’ of machine parts, and, put simply, this involves a prediction on the time remaining before a machine part is likely to require repair or replacement. After taking this into account, maintenance teams are able to optimise the maintenance schedule with a goal of reducing unplanned downtime, as well as needless preventative maintenance.</p><p>The inputs for a remaining useful life model will largely depend on the data available, and can all be thought of as indicators of the condition of the machine parts. As with most artificial intelligence application, the more data the better. The focus of this blog post is on a Lifetime Data approach to RUL using a statistical approach known as survival analysis.</p><p><strong>What is Lifetime Data and Survival Analysis?</strong></p><p>Survival analysis is the method to determine the lifetime of a part using a probability model. From historical data, the method calculates the probability of the item to ‘survive’ the number of days considering both observed and unobserved durations of part failure. Observed means that the entire life of the part was recorded from start to finish. Unobserved means that there was an interruption in the life of the part but we still consider the time that overlapped and was recorded. For example, the installation of the machine happened before the data was being recorded, meaning we know the part has survived longer than the start of records to the first failure.</p><p>Starting with a dataset with the columns of Machine, Failed_Part and Date, and focussing on a single part, e.g. compressors, we calculate each observed and unobserved event by considering each machine and calculating the number of times the part failed. We then use a merging of machine learning methods and probability distributions to estimate the likelihood of failure.</p><p>It can help to understand this approach through an example. Remi AI had a client that had a network of mobile construction machines that were points of sale (&gt;80) across Australia, these machines required compressed air for their operational requirements. Using historical data, the team at Remi AI was able to build out a Remaining Useful Life Forecast at an increasing number of days since the last failure or replacement, the results of which are shown in figure 1 below.</p><p>It is interesting to note in figure 1 below that the probability of survival decreased by ~ 10% in the first 100 days of use, followed by a fairly linear decline out to 1200–1300 days where the probability of survival drops away quite quickly. This curve is somewhat sigmoidal, and the power of this approach grows with decreasing linearity.</p><p>This plot is then translated into an estimation of ‘Remaining Useful Life’. This allowed the Maintenance Teams to better focus their Preventative Maintenance Efforts on parts that had only 100 days before predicted failure. It also allows their Preventative Maintenance to be more accurate in its proximity to failure, significantly reducing costs of refurbishment of the entire compressor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*-h52_00RJAimgpR-qNKQrw.png" /></figure><p><strong>What do I do with my Remaining Useful Life Estimate?</strong></p><p>Once you have a reliable estimate of remaining useful life, the maintenance schedule can be optimised to reduce downtime and maintenance cost. A decision policy can be built so that a part is replaced if it has not failed before a specific number of days, and this decision can be made by a human, or a more advanced option is to allow a reinforcement learning algorithm to make this decision.</p><p>1. Human Decision:</p><p>A team could select a threshold number of days for a machine part such that a replacement process will be initiated if the part has not been replaced by a certain number of days. The probability threshold will depend on the nature of the part in question i.e. a part that is critical to machine functionality might have a survival threshold value of 60% whereas a part that is less critical may have a survival threshold value of 20%. Dashboards and alarm systems can be built to allow maintenance staff to respond to pending part failures.</p><p>2. Reinforcement Learning Decision:</p><p>Finding the optimal survival probability threshold is an iterative process and is a problem domain where Reinforcement Learning (RL) can produce powerful results. Using Simulation of the Maintenance Programme, an RL model can take the RUL data as an input and optimise towards the best replacement policy such that machine uptime is maximised and unexpected maintenance time is minimised. This is then tested against historical breakdown data to confirm the Reinforcement Learning Model has found an applicable maintenance schedule.</p><p>If you’d like to learn more about how Preventative Maintenance can generate significant savings in loss of sales, breakdowns and general maintenance costs, please don’t hesitate to reach out through our <a href="https://www.remi.ai/aiconsulting">website</a>.</p><p>Eamonn Barrett</p><p>Partner</p><p>Remi AI</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ffc91d7e4a97" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Introduction to the Price Elasticity of Demand]]></title>
            <link>https://medium.com/@RemiStudios/introduction-to-the-price-elasticity-of-demand-fc2289e48e63?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/fc2289e48e63</guid>
            <category><![CDATA[pricing-strategy]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[price-elasticity]]></category>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[dynamic-pricing]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Wed, 06 Feb 2019 05:18:11 GMT</pubDate>
            <atom:updated>2019-09-25T04:23:34.427Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/760/1*Xz_zRodcYu24GsVj-E4n9A.jpeg" /><figcaption>How much does a change in price effect your sales?</figcaption></figure><p>You know the stats.</p><p>You’ve read that for 83% of customers, price is the #1 priority.</p><p>You’ve seen e-commerce take a larger and larger share of retail sales. Even the last tribes untouched by the modern world have heard about how, twenty-five years ago at the time of writing, one business changed the way that goods were purchased forever. You also know it’s not simply about price, unless you somehow operate your business in a vacuum, without competitors or substitutes.</p><p>Whether you are strictly e-commerce, brick and mortar, or a crafty combination of the two, the very fact you have found this article tells me you are looking for new ways to squeeze greater value out of your pricing strategy. This article, part one of our <em>A.I in Dynamic Pricing Optimisation </em>series, will explore the Price Elasticity of Demand (or PED) as a concept, and then how Artificial Intelligence can use this notion, as part of a broader suite of tools, to add value to your business (improve your pricing strategy and increase revenue). As briefly mentioned above, PED is not the only factor at play in your pricing strategy. Competitors, substitutes, brand loyalty, your own suppliers are all pulling different levers, and later articles will discuss how A.I can be implemented to take these into account.</p><p>But for now, it would serve us well to go through the economic concept behind PED (feel free to skip the next paragraph if such a sentence triggers an automatic yawn in you).</p><p>The price elasticity of demand is, as its name suggests, a measurement of the effect price has on demand (or goods sold). A good or service is deemed to be elastic if a relatively small change in price results in a large change in demand. By contrast, a good or service is said to be inelastic if significant changes in price only result in small changes in demand.</p><p>Assuming that the majority of goods and services sold have a degree of elasticity, we can plot the demand curve. This gives us a simple indication of, all other things being equal, how many units we can expect to sell if we simply toggle the price lever.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/359/0*o1XtTzjSxh7R-Soy.png" /></figure><p>With an understanding of PED, it’s now time to explore how Artificial Intelligence can be used to find the balance between quantity sold and the acceptable price at any given time. While it would be possible to use a rules-based approach, the most intelligent solution is one that uses Reinforcement Learning. Within the domain of Dynamic Pricing, an RL agent is given the freedom to change price and is rewarded for its success (and punished for its failings).</p><p>The result of this is a tool that is rewarded for increasing your business’s revenue over time.</p><p>What does this mean for you, in layman’s terms?</p><p>A.I Dynamic Price Optimisation is continually calculating PED, trying to find the correct balance to generate the greatest total revenue. When we start clients off with a new price optimisation implementation, we typically test three to four strategies around PED to see which generates the most revenue. These sort of strategies can deliver upwards of 10% increase in total revenue. As such, price optimisation, with Artificial Intelligence at the helm, is allowing e-commerce businesses to intelligently price their entire product range, to take advantage of demand at an SKU level.</p><p>If I was to give you one piece of information to convey the efficacy of this pricing strategy, This has seen increases in revenue of up to 20% when it was famously rolled out at Amazon.</p><p>Please note, PED as part of a broader A.I-Powered Dynamic Pricing strategy is not about loss leadership or a race to the bottom. The strategy is focused on intelligently charging higher prices when there is greater traffic through you site, and lowering it during quieter times. The A.I takes into account your profit margin to ensure you are not taking a loss.</p><p>Next steps:</p><p>We use PED as a critical tool in understanding how beneficial A.I could be. If you chose to undertake a pilot with us, we would go through your historic data to understand the price elasticity demand of your products. We would like to go through any competitor pricing data you have access to (to better understand the space within which you operate). This proof-of-concept usually takes one week. Upon completion, if there is a degree of predictability in the data, we like to progress with a trial. For this, we select five to ten of your range and grant control of pricing to the A.I. This requires a degree of customisation, depending on which platforms you utilise, but from this point onwards, the A.I is hungry to generate the most revenue for you at any given time.</p><p>This article will have raised questions in some of you, primarily around the other factors at play when it comes to demand. An intelligent A.I PED model will take these elements into account, elements such as:</p><p>- What competitors are doing</p><p>- Supply chain factors</p><p>- Brand Loyalty</p><p>In conclusion, A.I offers a great opportunity to generate greater total revenue for your business. If you would like to have a chat about what Remi A.I can do for your business’s pricing strategy, reach out. We are proud to offer a free consultation and a review of your existing pricing strategy.</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fc2289e48e63" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Artificial Intelligence, poker and regret. Part 3]]></title>
            <link>https://medium.com/@RemiStudios/artificial-intelligence-poker-and-regret-part-3-bb3210c79211?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/bb3210c79211</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[reinforcement-learning]]></category>
            <category><![CDATA[poker]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Wed, 23 Jan 2019 05:35:59 GMT</pubDate>
            <atom:updated>2019-09-25T04:24:00.900Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hPdfOZ7oAPtP3J5AMZgh9g.png" /><figcaption>Continuing to explore how an A.I can master poker</figcaption></figure><p>This is Part 3 of our series, Artificial Intelligence, Poker and Regret. Our aim was to share working code of a Texas Hold’em poker AI that uses Counter Factual Regret Minimisation (CFR) and a thorough explanation of the method. What was originally envisaged as a single blog post has been drawn out by the complexity of the subject matter. As such, the three part series has been extended to 5 parts. In Part 3 we will convey the concepts behind the CFR method, using the building blocks of Regret Matching (<a href="https://hackernoon.com/artificial-intelligence-poker-and-regret-part-1-36c78d955720">Part 1</a>) and Nash Equilibrium (<a href="https://hackernoon.com/artificial-intelligence-poker-and-regret-part-2-ee2e329d6571">Part 2</a>). Part 4 will introduce a method of abstracting poker hands to deal with the combinatorial explosion of possible hands. Finally, Part 5 will bring together the abstraction of Part 4 with the CFR of Part 3 to deliver a functional Texas Hold’em poker AI.</p><p>In Parts 1 &amp; 2, the common theme in both Rock Paper Scissors and the Prisoner’s dilemma is that each game is a single action performed by two players. Applying regret matching to single action games should be clear from <a href="https://hackernoon.com/artificial-intelligence-poker-and-regret-part-1-36c78d955720">Part 1</a>, where regret refers to the difference of potential reward with actual reward. In the context of poker however, there are multiple actions of each player before any reward is given and different combinations of possible actions, resulting in many different game outcomes. Thus, a range of rewards will contribute to the calculated regret.</p><p>Let’s consider a game of five card stud poker as a traversal of the poker game tree starting from a root node that begins with the deal of five cards to all players (see animation). The betting round is an alternating sequence of actions from the players until all call or all but one fold. Note that the rewards of a particular game can be played out if we know the hands of all players, i.e. given any sequence of actions we calculate the reward, which in this case is the money in the pot. For the sake of simplicity in displaying the game tree, we consider two players and a maximum three actions to be performed before the betting round is over and cards are shown.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/1*t_SxA0c_1JIHa--P0jbXKw.gif" /></figure><p>The challenge of applying regret matching on the larger game tree is defining the rewards (also referred to as utility) at each non-terminal node of the game tree. If we can accurately calculate a reward for each non-terminal node, then regret matching can be applied on any node as if it were a single-action game. That is, because each action would have a reward we can calculate the regret. In order to calculate the reward at a non-terminal node, we introduce the concept of the counterfactual reward. In this, counterfactual is defined as <strong>what hasn’t happened yet</strong>.</p><p>In regret matching, we not only introduced regret but also strategy, which is calculated from the accumulated positive regret. The strategy provides a probability distribution to choose the action that will minimise regret. Given that this probability distribution already reflects our tendency towards particular actions, we can also use it to reflect the future reward of that state. In particular, we can weight the future rewards of each action with the probabilities of each action to calculate the counterfactual reward for that node. In other words, the counterfactual reward is the expectation of reward using the probability distribution provided by the strategy.</p><p>Consider the scenario in the animation, where the red player’s hand beats the blue player’s hand. Suppose the red player checked for their first action and the blue player has a final action before red calls or folds. There is a pot of $50, a bet of $10, and blue can choose to either raise, call or fold. The blue strategy (based on previous calculated regrets) dictates that blue would raise with 0.1 probability, call/check with 0.5 probability or fold with 0.4 probability. The corresponding rewards are check = -50, fold = -50 or raise = ??, which awaits red’s final action. The counterfactual reward for blue, i.e. the first check node, is ?? x 0.1 + -50 x 0.5 + -50 x 0.4 = ?? x 0.1 + -45. To calculate the reward for ??, we consider red’s strategy where, based on previous regrets, the probabilities of red calling =0.7 and folding =0.3 allow us to calculate the counterfactual reward for red, i.e. the raise node on the left branch, -?? = 0.7 x 70 + 0.3 x -60 = 32.9. Note the colours of the utilities in the animation represent the utility for the corresponding player, i.e. if red wins blue loses and vice versa. Thus the counterfactual reward for blue’s last action is -32.9 x 0.1 + -45 = -48.3, which again defines the counterfactual reward for red’s initial check action to be 48.3.</p><p>So above shows how regret matching can still be applied to multi action games to give rise to the CFR method. We showed in part 2 that regret matching helps us to determine the Nash Equilibrium, and conveniently, the same argument still holds in the context of multi action games, In particular, the average strategy approximates to a Nash Equilibrium, so long as the overall regret is significantly small (<a href="http://papers.nips.cc/paper/3306-regret-minimization-in-games-with-incomplete-information.pdf">Theorem 2</a>).</p><p>Because of the multiple actions in poker, however, both the average strategy and regret calculations are weighted by the opponents’ strategies. The reasons for this are to ensure that the accumulated regret and average strategy don’t grow out of control for future actions. This is not portrayed in the animation but is conveyed in <a href="https://gist.github.com/shahradj/644c0281556cfc5701fb3c32b869ac73">the code</a> with the passing of both players’ strategies through the recursive function.</p><p>To summarise, in regret matching we use strategy to choose an action but in CFR, strategy also plays a role in calculating the reward and thus the regret. In CFR, the strategy is also important in calculating the average strategy which approaches the Nash Equilibrium.</p><p>With respect to limit five card stud poker, our implementation of CFR would theoretically reach this equilibrium after some large computation time and playing of millions of games but due to the large number of combinations of cards, convergence would take an unnecessary long time. With this explosion of possible hands, we are forced to abstract the poker hands in order to have a smaller state space and thus reduce the computation time. This method of abstraction we will introduce in Part 4 of this poker AI series.</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bb3210c79211" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The value of A.I in an SME, and what does it mean for me?]]></title>
            <link>https://medium.com/@RemiStudios/the-value-of-a-i-in-an-sme-and-what-does-it-mean-for-me-2363947e4f7?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/2363947e4f7</guid>
            <category><![CDATA[dynamic-pricing]]></category>
            <category><![CDATA[sme]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[small-business]]></category>
            <category><![CDATA[inventory-management]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Tue, 15 Jan 2019 07:49:05 GMT</pubDate>
            <atom:updated>2019-09-25T04:24:42.724Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/759/1*Efz5mz4G1ZB29fLfog00ag.png" /></figure><p>Through all of the froth and bubble around A.I in the headlines, it can be quite difficult to cut through the noise and answer a few simple questions: how big is this thing going to be, really? And should I leave it up to the major tech firms to deal with? Depending on who you believe, the answer to the first is usually somewhere in the ballpark of ‘big’, to ‘huge’, to ‘astronomical’. The answer to the second might depend on which salesman you’re talking to at the time. One thing I’m quite sure of is that no one knows for sure how large this opportunity is. Thomas Edison, arguably one of the most prolific inventors in the modern world, was confident Alternating Current (AC) would never catch on. Today, we struggle to imagine life without AC. Who could have known who was going to be right back when this argument was going on? A.I has the potential to have an event greater impact on our lives now, given the entrenched nature technology has in our lives.</p><p>In a recent research article put out by McKinsey and Company <a href="https://www.mckinsey.com/featured-insights/artificial-intelligence/notes-from-the-ai-frontier-applications-and-value-of-deep-learning">shows</a> what the authors believe to be the potential distribution of A.I opportunities across 19 industries. Exhibit 5 may be the most relevant to an SME (closely followed by exhibit 4) as it shows the greatest opportunities to unlock value using A.I broken up into business functions with marketing and sales taking top spot, followed by supply chain and manufacturing.</p><p><strong>So what?</strong></p><p>You might be wondering how an SME can utilise A.I in these high value areas when the industry is supposedly pushing the boundaries on a daily basis. “Why would I get involved now when the technology might be out of date tomorrow?”. Good question, and my suggestion is that the appearance of AI as a Service (AIaaS) offerings by many startups as well as the major technology players is close to making this technology attainable by all businesses. The nature of these subscription services is such that you automatically have access to the latest updates as and when they are available. I have detailed some fascinating applications of A.I in sales, marketing, and supply chain below, and these are approaches that can be implemented in an SME with relative ease by a firm like Remi AI.</p><p><strong>Approaches Available Today:</strong></p><p><strong>Dynamic Pricing:</strong></p><p>Most people are familiar with Uber’s surge pricing concept where drivers an incentivised to drive when demand is high by an increased fare. This concept can be applied to eCommerce with excellent success. The idea is that the A.I would track website traffic and lower prices to entice people to buy when traffic is low, and raise prices when traffic is high, all in real time. The advanced stage of this implementation is when you allow the A.I to predict the demand and alter the prices proactively. The demand for this function is growing and the team at Remi AI has implemented it with excellent results with several eCommerce clients to date.</p><p><strong>Inventory Management:</strong></p><p>My co-founder wrote an excellent blog post on A.I in <a href="https://www.remi.ai/blog/artificial-intelligence-for-inventory-management">inventory</a> and so I won’t rehash his work in this post other than to say that the goals of good inventory management are to reduce dead stock (stock that is unable to be sold) and reduce stock outs (an inability to fill an order due to a lack of the required item on hand). Humans do a fantastic job of these tasks though as warehouse sizes grow, suppliers demand longer forward orders, and customers alter their demand with the wind, humans can only do so much. A.I can be brought into the mix in 2 stages:</p><ol><li><strong>Forecast demand using supervised learning:</strong> The idea here would be that humans could act on these forecasts with their ordering patterns. I will say that the data needs to be of high quality and as close to live as possible for this approach to be of use.</li><li><strong>Allow a reinforcement learning (RL) agent to run your inventory operation: </strong>RL is a fascinating and advanced area of A.I where an A.I agent is allowed to act in an environment using pre-approved actions with a goal in mind, and it rewarded and punished depending on the result. Through this reward and punishment the A.I agent ‘learns’ which approach is best in different situations and so improves through time. This approach is close to complete inventory management automation and so is a large undertaking, and the potential savings are significant.</li></ol><p><strong>A.I in Paid Search and SEO</strong></p><p>Most businesses these days have a website and will be familiar with what can be the mammoth task of digital marketing for your website. Many SMEs hire digital marketing firms to run this operation for them which is entirely understandable. A.I is shaking things up in this space, driving costs down and performance up and I have detailed 2 of the more interesting approaches below.</p><ol><li><strong>Paid search using RL:</strong> It just so happens that the paid search area is a perfect application for RL given that it is data heavy and the data quality is high. These days it is possible to run entire paid search campaigns using RL and the results that I have seen have improved on the incumbent by a considerable margin.</li><li><strong>Intelligent Google Ads:</strong> Imagine that an A.I was able to read your entire website, followed by all of your competitor’s websites, decide on the most important keywords and craft Google ads around these keywords including all descriptions and titles without any input from a human other than to approve the ads if required. This isn’t too good to be true, is 100% possible, and could save SMEs thousands on SEO costs each month.</li></ol><p><strong>A.I as a Service (AIaaS):</strong></p><p>This phenomenon has been building over the past several years with a goal of putting the most advanced A.I methods in the hands of non-technical users. The approaches described above are excellent in their domain, but the AIaaS offerings coming out at the moment allow users to build their own predictive models for multiple applications. A business owner may want to predict revenue in the coming months and assess whether the weather or stock market have any influence on these figures. A manufacturer may way to predict when their vehicles will break down and service them in a predictive rather than preventative model. Platforms allowing users to do both tasks using one service are available and SMEs should absolutely take note in my opinion.</p><p><strong>The Road Ahead:</strong></p><p>The approaches described above are those that I believe to be salient as well as realistic to SMEs. There is no doubt that the A.I opportunity is large and can be a daunting one for those without a billion dollar budget. My suggestion is to begin to educate yourself on the topic from some trusted sources (think journals and universities if you’re looking for any hard facts). I would then ask yourself whether there is any part of your business that generates surplus data, and where a prediction is made on this data by an employee or subcontractor. Chances are that this task could benefit from A.I, and that there is already a service on offer that you could utilise such as those described above. SMEs can certainly keep up with the latest A.I technology, and there is no better time to start (if you haven’t already) than right now.</p><p>Disclaimer:</p><p>Remi AI can provide each of the services listed and I would encourage anyone interested to reach out for a discussion on how these approaches may be applied to your business.</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><p>$3.5T — $5.8T annually across 19 industries</p><p><a href="https://www.mckinsey.com/featured-insights/artificial-intelligence/notes-from-the-ai-frontier-applications-and-value-of-deep-learning">https://www.mckinsey.com/featured-insights/artificial-intelligence/notes-from-the-ai-frontier-applications-and-value-of-deep-learning</a></p><p>AI in Inventory</p><p><a href="https://medium.com/@RemiStudios/artificial-intelligence-for-inventory-management-c8a9c0c2a694">https://medium.com/@RemiStudios/artificial-intelligence-for-inventory-management-c8a9c0c2a694</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2363947e4f7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Dynamic Pricing for Online retailers.]]></title>
            <link>https://medium.datadriveninvestor.com/dynamic-pricing-for-online-retailers-8ad8228788c0?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/8ad8228788c0</guid>
            <category><![CDATA[uber]]></category>
            <category><![CDATA[ecommerce]]></category>
            <category><![CDATA[pricing-strategy]]></category>
            <category><![CDATA[dynamic-pricing]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Tue, 18 Dec 2018 18:03:05 GMT</pubDate>
            <atom:updated>2019-09-25T04:26:27.689Z</atom:updated>
            <content:encoded><![CDATA[<figure><a href="http://www.track.datadriveninvestor.com/1B9E"><img alt="" src="https://cdn-images-1.medium.com/max/700/0*k4zJaFKJvClUyfzQ" /></a></figure><h3>How Dynamic Pricing can be applied to Online Retailers.</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*69NG-1OXYp8usrqLXBCl5Q.jpeg" /><figcaption>Uber have executed A.I powered Dynamic Pricing perfectly, how can you?</figcaption></figure><p>Back in 2012, one of the US Uber teams noticed a significant problem in their operations. Late on Friday nights, just after 1am, the app was seeing a huge spike in ‘failed requests.” These failed requests were where users were requesting a ride and there was no driver available in a suitable radius to meet their request. The cause behind these failed requests was that Drivers were finishing for the night to go home and sleep. And this was only one hour before there was a significant spike in party-goers trying to return home. The result? Many disgruntled and inebriated users.</p><p>In economic theory, this is what is considered a supply-demand imbalance: where supplier cannot meet demand expectations of the market. For the majority of companies this is a significant problem for both reputation and loss of sales where customers quickly turn to other suppliers and the company becomes associated with being out of stock. It is only in strong product monopolies, where the customer doesn’t have appealing alternatives, that this supply-demand imbalance can be utilised to the advantage of the company. Such an example of this was the earlier iPhones, where Apple were deliberately using the shortage of the product to build hype around the popularity of the product.</p><p>But in Uber’s case, these party-goers could easily revert to the traditional hailing of a cab.</p><p>So the Uber team came up with a solution. What if they offered the drivers a higher price to stay out longer? If they used a multiplier on the standard rate, would this encourage more drivers to stay out or even return out to the streets? Would the money a driver earned increase?</p><p>In just over two weeks, after a series of trials, Uber had an answer. By offering an increased surge rate to drivers, they were able to increase the number of drivers on the road by 80%, effectively reducing the failed rides by 65%. This was a resounding success. Ultimately, drivers were motivated by the increased pay and that more people got home.</p><p>This simple demand model is not as sophisticated as those being applied to modern E-Commerce Businesses, but the basic concepts apply to any marketplace, including E-Commerce..</p><p>When it comes to Artificial Intelligence, demand is probably the most interesting concept in the E-Commerce pricing strategy. To highlight this concept, it’s worth sticking with Uber for a little longer with the following hypothetical</p><p>It is 7:30pm on a Friday night in the Western Suburbs of Sydney. There are a total of 100 Uber Drivers in a 30km radius. When Uber’s pricing is normal and not surging, there is a 10% likelihood that a user who opens the app will book a ride in the next 5 minutes. At the current time, there are 900 people in the 30km radius who have opened the app. So, 90 of the 900 users will probabilistically request a ride. This means that 90% of the drivers, who in Uber’s market model can be considered the inventory, will be booked.</p><p>At 8:30pm, there is a downfall of rain and suddenly there is a significant spike in users. Instead of 900 users opening the app like an hour ago, there are now 2000 users in the 30km radius who have opened the app in between 8:30pm and 8:35pm. Uber know that the conversion rate from app opens to ride booked also increases during significant rainfall to 15% which is up from 10% from pre-rainfall.</p><p>Compared to an hour ago, where 90 users were looking to book a ride from the 100 drivers, now 300 users are expected to book a ride from the 100 drivers.</p><p>Uber’s Pricing algorithms will automatically apply a 2.4% increase surge price. Uber know in advance that this price surge will reduce the conversion rate from 15% to 5% and that it will also encourage another 50 drivers onto the road in the area to help meet demand.</p><p>For the next 5–10 minutes the surge pricing is in effect, reducing demand and increasing supply. Once it has brought about sufficient balance, pricing is reduced and some of those that were deterred from conversion at 8:30pm typically convert.</p><p>The most important concept for the E-Commerce Business Leader in this example is the demand indicator. Uber have a very simple, mono-indicator because they’re ultimately only selling a single product; drivers. They use app opens as the primary indicator of market size for their single product. Whereas in the E-Commerce model, there are implementations that are looking for product interest indicators.</p><p>Some raise objections that Uber’s model is too simple, and that Dynamic Pricing cannot be applied to larger E-Commerce retailers. The reality is that every E-Commerce Store is a market, and all are subject to the same economic laws of supply and demand. The main point of difference is that the e-commerce store is more complicated and data intensive. Yet this very complication is why artificial intelligence is yielding such significant increases in operating margins for these stores, because it can process the huge amounts of data generated to understand the pricing power of the company and deliver actionable insights</p><p><strong>How do we understand demand in the E-Commerce framework</strong></p><p>On-site search and product viewings are two of the strongest indicators we use to model demand for individual products.</p><p><strong>Using demand to understand pricing power</strong></p><p>“The single most important decision in evaluating a business is pricing power.” Warren Buffett, CEO Berkshire Hathaway</p><p>Pricing Power is the relationship between your business, your competitors, your suppliers and your customers. It is ultimately the power your business holds to increase the price above the margin cost of the product your selling.</p><p>A.I price optimisation is proving to be the quickest and most intelligent way to understand and capitalise on your pricing power. Solutions such as Remi Ai’s Price Optimisation Platform allow you to quickly capitalise on lost revenue across your entire product range.</p><p>It is continually trying to understand the company’s pricing power and the significance of each competitor’s pricing across each SKU. This results in an outperforming of existing manual rules based systems. For one of our clients, we have seen A.I price optimisation outperform the rules-based strategy of experienced sellers by more than 20%.</p><p>Conclusion.</p><p>So just like Uber, E-Commerce Retailers are now starting to utilise artificial intelligence to find the market imbalances and to increase revenue. If you’d like to learn more about a.i price optimisation and how it’s being used by others, head over to our website.</p><p><a href="https://www.remi.ai/dynamic-pricing">https://www.remi.ai/dynamic-pricing</a></p><p>Alexie Camille</p><p>Remi A.I</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fupscri.be%2Fb2a0d6%3Fas_embed%3Dtrue&amp;dntp=1&amp;display_name=Upscribe&amp;url=https%3A%2F%2Fupscri.be%2Fb2a0d6%2F&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=upscri" width="800" height="400" frameborder="0" scrolling="no"><a href="https://medium.com/media/0707f5c806284d01a4a13c7b13a91ce3/href">https://medium.com/media/0707f5c806284d01a4a13c7b13a91ce3/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8ad8228788c0" width="1" height="1" alt=""><hr><p><a href="https://medium.datadriveninvestor.com/dynamic-pricing-for-online-retailers-8ad8228788c0">Dynamic Pricing for Online retailers.</a> was originally published in <a href="https://medium.datadriveninvestor.com/">DataDrivenInvestor</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A.I in Pedestrian and Cyclist Management]]></title>
            <link>https://medium.com/@RemiStudios/a-i-in-pedestrian-and-cyclist-management-7f910fcc6170?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/7f910fcc6170</guid>
            <category><![CDATA[cities]]></category>
            <category><![CDATA[smart-cities]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[urban-planning]]></category>
            <category><![CDATA[image-recognition]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Wed, 10 Oct 2018 03:16:46 GMT</pubDate>
            <atom:updated>2019-09-25T04:26:49.195Z</atom:updated>
            <content:encoded><![CDATA[<p><a href="https://www.remi.ai/">Remi AI</a> has built a patented Artificial Intelligence solution for <a href="https://www.superloci.com/">Super Loci</a>. The cutting edge platform will help urban planners, local councils, developers, architects, retailers and outdoor marketers better understand the spaces within which they operate. This article explores the broader issues around inefficient public spaces, before delving into how Super Loci can help.</p><p>Around Australia, billions of dollars are spent annually by state road agencies on improving traffic flow for motor vehicles on motorways and arterial roads. Current methods of traffic monitoring are fairly crude, but adequate for purpose, and is usually measured in 15-minute increments of vehicles travelling in lanes or turning at intersections. These are fed into various traffic models, which are then used without much further question.</p><p>But there are major changes on the horizon. Through Smart Cities and IoT, analytics is playing a bigger part of city planning, urban design and transport planning. The advent of connected and automated vehicles (CAVs) is looming on the horizon, but in Australia alone there are 716 sections of legislation that will have to be amended to allow CAVs to operate at level 5 automation.</p><p>There is also a growing emphasis on ‘streets as places for people’ not just for movement of cars and freight — hence leading to the term ‘movement and place’ or ‘link and place’. The City of Melbourne, for example, reported that a 10% improvement in pedestrian connectivity would increase the value of the Hoddle grid economy by $2.1 billion (6.6 per cent of the economy) (SGS).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*-PboQe4As-gJhEmvaPm7AQ.jpeg" /><figcaption>A 10% improvement in pedestrian connectivity would increase the value of the Hoddle grid economy by $2.1 billion (6.6 per cent of the economy) (SGS)</figcaption></figure><p>Another study in Auckland showed that at one intersection alone, pedestrian delays cost the economy NZ$2.2 million annually, and that optimising the signals could reduce that delay by 26 to 46 per cent, representing $1 million savings.</p><p>But how can we prioritise pedestrians and bicycle riders if our current methods barely count them at all? There’s laser or infrared sensors, sometimes induction loops, or even people with clipboards, but these are relatively unreliable, rudimentary and slow to get results. There is also no method for measuring a ‘crowd’ of pedestrians waiting on a footpath. So how can we prioritise a large crowd of pedestrians who’ve just come out of the train station onto a busy city streets?</p><p>Super Loci is a patented solution that tackles the problem of the lack of insight in how pedestrians and bicycle riders interact with spaces. Greater insight before a project leads to better design, and more certainty around decisions.</p><p>Remi AI was given the following brief from Super Loci: Build a pedestrian and bike counter that is aligned with our values of efficient and ethical modelling of how people interact with any given space. We are proud have built a tool that has come to provide councils, urban planners, state road agencies and property developers with accurate insights about their particular space, whilst also offering anonymity to the people interacting with the space. The Super Loci platform can identify how many pedestrians and riders are moving through an area, and is able to estimate age bracket as well as gender, without resorting to facial recognition. It is fully EU GDPR (privacy) compliant.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*p1MFrxqEwrSPyB697Wvh0Q.png" /><figcaption>The Super Loci dashboard display.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Qa430y2MNR5gc0lOjl_-7A.png" /><figcaption>Close up: the platform shows where the footage was shot and the camera view.</figcaption></figure><p>To better understand engagement with a space, Super Loci also notes the dwell time of each individual, showing where and for how long pedestrians engage with a space. This is important if we want to understand the ‘stickiness’ or attractiveness of a streetscape. Do people want to linger and enjoy, or do they rush through as quickly as possible?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0f-4F7QPTYCaHcdKbggo4A.png" /><figcaption>Super Loci helps customers understand when, and what sort of demographics, are utilising their space.</figcaption></figure><p><a href="https://www.superloci.com/">Super Loci </a>is a Sydney-based start up, founded by experts in city planning, urban design and transport. At the heart of the business is a focus on human-centric cities. As such, it was imperative that all data that would train the algorithm behind Super Loci be anonymised and comply with strict European privacy laws.</p><p>The platform was conceived by city planners, for city planners. Designed to be a cutting edge, ethical, insightful and cost effective way to help build better cities and towns. Whether used in a single location, or rolled out across a broader space, Super Loci allows planners and designers to either quantify their gut instinct on spatial interaction, or even discover new information they were previously unaware of. Greater insight leads to smarter city design, which in turn leads to smoother and more convenient journeys for pedestrians and bike riders.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eno9cnHKCEYUK1kKsMrk1w.jpeg" /></figure><p>Information visualisation is a vital part of the platform. We built topographical mapping functionality that translates what the camera sees onto a map. This tool helps map individual journeys, as well as heatmap functionality to identify the most common thoroughfares. Another exciting component of this technology is that if there are multiple cameras in use, the topographical mapping can be stitched together between the cameras. The result is a comprehensive, easily digestible understanding that will empower our clients to make the best decisions.</p><p>The platform is set for a beta release. Contact us through our <a href="https://www.remi.ai/contact">website</a> for a secure login so you can test it today.</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7f910fcc6170" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Artificial Intelligence for Dynamic Pricing]]></title>
            <link>https://medium.com/@RemiStudios/artificial-intelligence-for-dynamic-pricing-326c3b88a37?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/326c3b88a37</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[dynamic-pricing]]></category>
            <category><![CDATA[business-intelligence]]></category>
            <category><![CDATA[ai-dynamic-pricing]]></category>
            <category><![CDATA[saas]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Tue, 02 Oct 2018 04:03:16 GMT</pubDate>
            <atom:updated>2019-09-25T04:27:17.586Z</atom:updated>
            <content:encoded><![CDATA[<p>Dynamic pricing, a practice started by American Airlines in the 1980s, is now becoming an important tool for e-commerce retailers and other businesses. Widespread in both the air travel and hotel industry, and famously well executed at Amazon, companies are now utilising dynamic pricing to respond to changes in demand and to drive significant increases in revenue. Remi Ai’s Dynamic Pricing platform for example typically delivers a 17% average increase in monthly revenue across the SKUs when introduced. (This figure obviously varies across industries and products).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/259/0*OSFSxRlpvkmX394H" /><figcaption>American Airlines pioneered Dynamic Pricing the 1980s</figcaption></figure><p>In the age of e-commerce, where customers can now easily compare prices between multiple suppliers in a matter of seconds, price is now considered the greatest factor in the purchasing decision. A recent US survey found that 80% of survey participants considered price the single most important factor in a purchase.</p><blockquote>“80% of survey participants considered price the single most important factor in a purchase.”</blockquote><p>In the last 10 years, Artificial Intelligence and Machine Learning Models have had a large impact on Dynamic Pricing methods. With greater customer segmentation and the ability to analyse hundreds of thousands of SKUs, these methods have become the best performing approaches to the challenges of pricing and dynamic changes in pricing strategies.</p><p>Although they are complex models, these Dynamic Pricing machine learning models are grounded in a very simple concept:</p><blockquote>Deliver the right price for every customer while increasing revenue for the business.</blockquote><p>Even a small retail business with only a few core products will have a wide range of customers with different tastes, values and budgets. For one of our recent clients we ran user bucketing and uncovered over 40 different customer types. Each of these different customer types displayed different preferences and buying behaviours.</p><p>With any artificial intelligence that is controlling dynamic prices, it’s primary responsibility is to uncover the fundamentals of the supply demand curve for each SKU in your product range and then price effectively against that curve to increase revenue. Supply and demand is perhaps one of the fundamental concepts of economics and it is the backbone of a market economy.Supply and demand is perhaps one of the fundamental concepts of economics and it is the backbone of a market economy.</p><p>Demand refers to how much a product or service is desired by buyers. The quantity demanded is the amount of a product people are willing to buy at a certain price; the relationship between price and quantity demanded is known as the demand relationship.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/302/0*ryJ6pB1T8Jy3OU65" /><figcaption>The relationship between price and quantity</figcaption></figure><p>As many of you will know, traditionally the higher the price of the good, the lower the quantity demanded.The lower the price, the greater the demand for the good.</p><p>For the inverse, Supply represents how much the market can offer. Quantity supplied refers to the amount of a certain good producers are willing to supply at a given price. The correlation between price and how much of a good or service is supplied to the market is known as the supply relationship. Price, therefore, is a reflection of supply and demand.</p><p>As technologies for Machine Learning (ML) and Artificial Intelligence (AI) become more advanced and the dimensions of available data expand, dynamic pricing is going beyond its traditional inventory management function and enabling companies to understand this demand relationship. Pricing is becoming intelligent and continually adjusting to changing consumer behaviour and demand preferences, while also responding to organisational inventory and marketing requirements as well as other external pricing influences.</p><p>Today, Artificial Intelligence approaches such as Remi AI’s dynamic pricing platform is enabling enterprises to marry rich data sets with sophisticated pricing models and apply our artificial intelligence techniques. The product is pricing alternatives across thousands of product stock keeping units (SKUs). These offers are also uniquely tailored to the individual consumer dynamically at the point of engagement, and thus more likely to generate a sale.</p><h3>The Remi AI approach to dynamic pricing.</h3><p>Across our clients, the algorithmic approach changes, but there are three fundamental steps in our dynamic pricing pipeline that are important to understand.</p><p><strong>Stage 1. Customer Clusters</strong></p><p>In the context of customer segmentation, cluster analysis is the use of a mathematical models to discover groups of similar customers based on finding the smallest variations among customers within each group. These homogeneous groups are known as “personas”.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/0*OK0Bw-lwVapwg3U9" /><figcaption>Customer Segments</figcaption></figure><p>Arguably the greatest advancement that machine learning has brought to dynamic pricing has been the highly accurate customer segmentation. Older approaches treated the customer base a homogeneous whole, whereas the reality is that every company has a wide range of customer personas.</p><p><strong>Stage 2. Understanding Demand Relationship for every ‘persona’.</strong></p><p>Here our AI platform models the demand relationship for every ‘persona’. A simple example of this is as follows; users who are searching for two cases of beer for delivery at 9:00pm on a Friday are happier to pay a higher price for the product.</p><p><strong>Stage 3. Vary the price depending on the customer, time of day and other key factors.</strong></p><p>With a strong understanding of the demand relationship and the different buying personas of our customers, we start to bring our Reinforcement Learning AI to bear on the sales channels, changing the prices of each individual SKU at regular time intervals.</p><p><strong>What is the AI actually trying to achieve?</strong></p><p>This depends on your company’s objectives. Some of our customers have their dynamic pricing focused on increased conversions, in which case the platform is weighted toward lowering prices and others are focused heavily on increasing total revenue. If the AI is weighted toward increasing total revenue, it often will increase margins on numerous SKUs by a reasonable percentage.</p><p><strong>The benefits that these methods bring are:</strong></p><ul><li>Intelligent control on Pricing Strategy</li><li>Enabler for Growth in Revenue</li><li>More precise, SKU level prices</li><li>Faster response to demand fluctuations</li><li>Price changes take into account more factors including customer price perception, leading to long terms increases in sales or profits</li></ul><p>For those looking to bring Artificial Intelligence into their business operation, Dynamic Pricing can yield significant increases in revenue. For more information, please visit our <a href="https://www.remi.ai/dynamic-pricing">site</a></p><p>Alasdair Hamilton — CEO</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=326c3b88a37" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Multi-Agent Reinforcement Learning]]></title>
            <link>https://medium.com/@RemiStudios/multi-agent-reinforcement-learning-3f00b561f5f0?source=rss-40cdb20c9eee------2</link>
            <guid isPermaLink="false">https://medium.com/p/3f00b561f5f0</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[reinforcement-learning]]></category>
            <dc:creator><![CDATA[Remi AI]]></dc:creator>
            <pubDate>Fri, 24 Aug 2018 06:47:31 GMT</pubDate>
            <atom:updated>2019-09-25T04:27:46.754Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/420/1*yea7f8pNHoxIkeI4TLUcmw.gif" /></figure><p>Amongst the various domains of Artificial Intelligence (AI) research being advanced at the moment, one domain has become critical to the broader success of artificial intelligence, while still in its infancy. It’s called multi-agent reinforcement learning.</p><p>Multi-agent reinforcement learning is the study of numerous artificial intelligence agents cohabitating in an environment, often collaborating toward some end goal. When focusing on collaboration, it derives inspiration from other social structures in the animal kingdom. It also draws heavily on game theory.</p><p>Multi-agent systems are not just a research method, they can be used to model many complex problems of today’s society, such as urban and air traffic control (Agogino &amp; Tumer, 2012), multi-robot coordination (Ahmadi &amp; Stone, 2006; Claes, Hennes, Tuyls, &amp; Meeussen, 2012), distributed sensing (Mihaylov, Tuyls, &amp; Now ́e, 2014), and energy distribution (Pipattanasomporn, Feroze &amp; Rahman, 2009).</p><p>Despite all of these useful applications, there are more critical reasons for the research community to focus on multi-agent reinforcement learning:</p><ol><li>the role of sociability in increasing a species intelligence</li><li>for the safe development of artificial intelligence</li></ol><p><strong>Social Life and Brain Size</strong></p><p>Sociability — spending time with family, colleagues and strangers — is cognitively demanding. The neurological functions that promote living in societies — empathy, communication, hierarchy comprehension, all demand significant cognitive reserves. It is often taken for granted because humans do it so effortlessly, but the ability to empathise with one’s peer and to comprehend they’re in danger involves staggeringly complex neurological function. To comprehend that your friend is in danger of being hit by a car as they step out onto the road, and to yell to them to warn them, is a truly impressive cognitive collaboration of empathy, causal understanding, communication and other functions. This isn’t just applicable to human society either, when you put aside our own sense of superiority, the neurological complexity of other primates and avian communication systems are admirable and intricate. Many primates and other species warn each other of predatory threats and new food sources, knowing the benefit of community.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*q21B9nC5dLud_tX1_WWN6Q.gif" /></figure><p>Similarly, group hierarchies are an intellectually demanding approach to social living that many species adopt. Humans, other primates, and birds have developed group hierarchies to help navigate the rich ecosystem of social living. (Though one should note that not all social species tend toward hierarchies). A hierarchy can best be understood as a ranking system that helps manage unequal access to limited resources. From food and water, to more abstract human concepts such as money and prestige. It helps us in that we don’t have to fight over the food everytime we sit down to eat. It helps us manage systems such as airline seats, where money is utilised to maintain a seating hierarchy between first class, business, premium economy and economy.</p><p>This ushers in a key point — social cooperation and hierarchies are extremely challenging, and this is reflected in the brain. The British Anthropologist Robin Dunbar has shown that across various taxa (e.g. ‘birds’, ‘ungulates’ or ‘primates’) the bigger the average size of the social group in the species, the larger the brain relative to total body size. Secondly, the larger the neocortex, relative to total brain size. Dunbar’s influential ‘social brain hypothesis’ posits that increases in social complexity and the evolutionary expansion of the neocortex has been linked.</p><p>This link also occurs within a species. Among some primates, group size can vary tenfold (depending on the richness of the ecosystem). This was modeled in a fascinating neuroimaging study, in which captive macaque monkeys were housed in different sized groups. The bigger the group, the more thickening of the prefrontal cortex and the superior temporal gyrus, a cortical region involved in Theory of Mind, the tighter the activity coupling between the two.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/530/1*UZ0x9_wQjrc1vYXVDVZU5w.gif" /></figure><p>This variety of richness in social ecosystems results in multi-agent spectrum in the animal kingdom: tigers are entirely individualistic, only interacting with other tigers to mate or fight (sometimes both), and ants or bees are entirely motivated by the collective, giving their lives for the group. Our own species sits somewhere in the middle -, we have individual motivations and social motivations. In different circumstances these competing motivations take priority.</p><p><strong>What does this have to do with Artificial Intelligence?</strong></p><p>If we’re working toward advancing artificial intelligence, it reasonable to posit that the same sort of sociability that is so fundamental to our own intelligence will be important to advancing artificial intelligence.</p><p>The work Remi AI have been undertaking in this domain has been primarily focused around either artificial intelligence controls in business operations, i.e. inventory management, fleet optimisation etc. and also language development between different agent systems.</p><p>There are many major challenges that have been identified throughout our work, though the largest at this early stage is associating reward with actions.</p><p>To illustrate the difficulty of this problem, when designing a single agent to play Space Invaders, it is relatively simple to create a reward and punishment system: the reward is associated in the increase in game points and the punishment is associated with the death of the agent’s spacecraft.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/850/1*PqoJzG3aX9MXAd_cxf2OTQ.png" /></figure><p>Whereas with multi-agent systems it can be difficult to clearly associate a reward with a particular action by a particular agent. For example, if you built a transport simulation, with 10 different agents driving a bus apiece, you could trial rewarding the agents collectively for reducing wait time across all passengers. It is incredibly difficult for the agents to map their own actions to a collective reduced wait time, as the agents can struggle to attribute which action lead to the reduction. This problem increases exponentially with the number of agents. When you factor in the city of London has 8000 buses, you begin to get an understanding of the complexity required.</p><p>Multi-agent reinforcement learning and safety.</p><p>The majority of modern day Artificial Intelligent systems are, in cognitive terms, still fairly naive and unintelligent.</p><p>The supervised methods that are making waves such as deep-learning are analogous, in biological terms, to the development of the visual and auditory cortexes — present in the most many of the simplistic organisms that evolved millions of years ago. State of the art reinforcement learning systems are still largely learning through trial and error, (though we’re making advances in imitation and transfer learning which bring our agent’s learning more similar to the way animals learn from their parents and other tasks respectively).</p><p>Although it is incredibly difficult to obtain a consensus around scales of animal intelligence, it is widely considered in A.I research circles that we still haven’t devised an A.I that is as generally intelligent as a rat.</p><p>But as our agents become more intelligent as our research advances, multi-agent reinforcement learning will become critical, not only for the development of communication, empathy and other fundamental intellectual capabilities, but because it will teach the agents how to behave in groups without harming each other. This is a critical research step for developing safe agents in general.</p><p>Can we just program safety into the agents, or does some part of it need to be learned?</p><p>Personally, I am not certain.</p><p>One of the few areas where there is consensus around definitions of intelligence is; the less an animal or agent relies on instinct and the more it learns from environment, the smarter it is.</p><p>Instincts are, of course, incredibly useful. Just think of the baby turtle who automatically makes a dash for the water after they’ve exited their shell. Without this instinct they’d sit on the beach and watch their fellow newborns be eaten before they learned to rush to the safety of the water.</p><p>But instincts, like regulation, can be dumb and restrictive. It can also save lives. Finding the balance between the two is critical.</p><p>Humans are regularly the best example of this — our intelligence permits us to even override our instincts, if we allow pause for thought. One of the reasons the human infancy period is so much longer than other species is because our cognition is less instinctive and more learned.</p><p>If we’re working toward human-level artificial intelligence, it’s reasonable to posit that along this research path our A.I agents will need to have lengthy learning periods, just like children. With this, they’ll need to learn how to behave.</p><p>In the long-run, I don’t see these lengthy learning periods will be a continued requirement. Unlike humans, we’ll be able to transfer most learned behaviours and knowledge between A.I generations. That being said, the real world is a messy, ever changing place and we cannot assume that we won’t have to occasionally update an A.I agents behaviour to encompass new developments.</p><p>Finally, it is the opinion of the author that the span between the capabilities of state-of-the-art Multi-agent reinforcement learning and human social capabilities underscores just how far we have to go as a research discipline to achieve advanced artificial intelligence. I also consider it one of the most important research domains, with its relevance to language development, empathy, imagination, ethics and other qualities we associate with being human. Some of human’s most impressive feats, from the Eiffel Tower to the invention of the Internet, have emerged through a collective effort.</p><p>If we’re to develop safe, advanced A.I, making progress in multi-agent reinforcement learning will be a critical step.</p><p>Stay tuned on the <a href="https://www.remi.ai/blog">Remi AI blog</a> as we build out the complete supply chain offering, or you can sign up to our Remi’s Recommended Reading newsletter <a href="https://www.remi.ai/signup">here</a>.</p><p>Or, if you’re ready to start seeing the benefits of A.I-powered inventory management, start the journey <a href="https://www.remi.ai/inventory">here</a>.</p><p>Who are we?</p><p><em>Remi AI is an Artificial Intelligence Research Firm with offices in Sydney and San Francisco. We have delivered inventory and supply chain projects across FMCG, automotive, industrial and corporate supply and more.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3f00b561f5f0" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>