Data Science Insider: September 9th, 2022

In This Week’s SuperDataScience Newsletter: Experts Warn EU’s AI Act could Negatively Impact Open-Source. Amazon’s AI Sends Delivery Drivers on ‘Impossible’ Routes. Collaborative ML That Preserves Privacy. ML Helps Clear Land Mines. What Top Gun Can Teach Us About Data Science. Cheers, - The SuperDataScience Team P.S. Have friends and colleagues who could benefit from these weekly updates? Send them to [this link]( to subscribe to the Data Science Insider. --------------------------------------------------------------- [EU’s AI Act could Negatively Impact Open-Source]( brief: In this blog from The Brookings Institution experts have warned that proposed EU rules could limit the type of open-source research that produces cutting-edge AI tools such as GPT-3. The Brookings Institution is a nonprofit public policy organisation with a mission to conduct in-depth research. The blog warns against the proposed EU regulation of open-source AI and argues that it would create legal liability for general-purpose AI systems as well as undermine their development. The Institute’s research notes that under the EU’s draft AI Act, open-source developers would be forced to follow guidelines for risk management, data governance, technical documentation, and transparency, as well as standards of accuracy and cybersecurity. It argues that it would be possible that a company could sue the open-source developers on which they built their product if an open-source AI system that they were using led to a calamitous outcome. Why this is important: The EU’s draft AI Act would be the first law on AI by a major regulator anywhere and has the potential to set the tone for global legislation to come. The negative impact that it may have on open-source developments should be seriously appraised before any laws are passed. [Click here to learn more!]( [Amazon’s AI Sends Drivers on ‘Impossible’ Routes]( brief: Amazon’s AI software has been accused of sending them on ‘impossible’ journeys. The AI that manages driver routes is supposed to make deliveries more efficient but has been criticised by drivers for not considering real-life obstacles and it has been claimed that it “doesn’t account for real-world conditions like rivers or train tracks or [narrow] roads.” The discord comes from Japan and has led to 15 drivers in Nagasaki forming a labour union, the second after Yokosuka which was organised in June and started with ten members. Workers claim that they have to work long hours to deliver more and more parcels for no extra pay, due to software failures. The unions are campaigning for Amazon to raise daily pay, cover fuel expenses, and reduce overtime work. Amazon has claimed that the workers are subcontractors who technically work for a third-party logistics company. Why this is important: Amazon may have a legitimate reason to claim that they technically have no legal responsibility for the working conditions of the subcontractors. However, as a colossus of the technological world, Amazon has a moral obligation to ensure that its software doesn’t impede workers’ ability to do their jobs efficiently. [Click here to read on!]( [Collaborative ML That Preserves Privacy]( In brief: Researchers from MIT have collaborated with startup DynamoFL to increase the accuracy and efficiency of an ML method that safeguards user data. The researchers have achieved this by using federated learning, an ML technique that trains an algorithm across multiple decentralized edge devices or servers holding local data samples, without exchanging them. The technique keeps sensitive user data private, making it ideal for use when gathering enormous datasets is needed but privacy is a concern, such as with medical images. However federated learning has three drawbacks which this new research has addressed: Transferring a large ML model to and from a central server has high communication costs; each user gathers their own data, meaning that the data doesn’t necessarily follow the same statistical patterns, hindering the performance of the combined model and; The combined model is not personalised for each user but is instead made by taking an average. Why this is important: Vaikkunth Mugunthan the lead author of a paper that introduces this technique sums up its significance when he says: “This work shows the importance of thinking about these problems from a holistic aspect and not just individual metrics that have to be improved. Sometimes, improving one metric can actually cause a downgrade in the other metrics. Instead, we should be focusing on how we can improve a bunch of things together, which is really important if it is to be deployed in the real world.” [Click here to discover more!]( [ML Helps Clear Land Mines]( In brief: This Scientific America article discusses and contains a link to, a documentary short which explores how ML and drones are being used to identify and clear land mines so that humans don’t have to. The documentary shows an explosive-ordnance-disposal field laboratory maintained by Oklahoma State University, and led by researchers Jasper Baur and Gabriel Steinberg, co-founders of the Demining Research Community, a nonprofit organisation bridging academic research and humanitarian demining efforts. The article states that The Landmine and Cluster Munition Monitor reports that there were at least 7,073 people killed or injured by mines in 2020, across 54 countries. Often non-profits are tasked with identifying and eradicating mines that have been left behind after wars but they lack the resources needed to tackle all of them. Instead, by training a drone-based, ML-powered detection system it is possible to find and identify dangerous explosives with fewer resources. Why this is important: Land mines are unfortunately not a thing of the past and intelligence indicates that they are currently being deployed in Ukraine. By perfecting this technology we can hopefully save many people from death and injury. [Click here to see the full picture!]( [What Top Gun Can Teach Us About Data Science]( In brief: Top Gun: Maverick has been the blockbuster of the summer full of thrills, action, and adventure but not, you may think, of lessons on data science. This article by Data Science Central explains that we may have assumed wrongly and asserts that there are lessons to be learned around how fighter pilots gather, infer, analyse, and act on the multiple streams of real-time data that impacts their success as fighter pilots. The article discusses OODA Loop (observe–orient–decide–act), a real-life guide that fighter pilots use to make real-time decisions and increase their odds of successful engagement. OODA Loop is then combined with The Data Science Engagement Methodology, the article outlines five lessons that can be learned from this blend: Speed of decision is as important as accuracy; When “good enough” is “good enough”; The economic power of compounding; It’s all about improving odds and; Achieving your objective. Why this is important: The premise of this article may seem a little bizarre but it offers fascinating insights into how fighter pilots operate in a way that complements and incorporates data science methodology. [Click here to find out more!]( [Super Data Science podcast]( this week's [Super Data Science Podcast](, Dr Jennifer Hill, Professor of Applied Statistics at New York University, joins us to discuss the role of causality in data science applications and her favourite Bayesian and ML tools. ." --------------------------------------------------------------- What is the Data Science Insider? This email is a briefing of the week's most disruptive, interesting, and useful resources curated by the SuperDataScience team for Data Scientists who want to take their careers to the next level. Want to take your data science skills to the next level? Check out the [SuperDataScience platform]( and sign up for membership today! Know someone who would benefit from getting The Data Science Insider? Send them [this link to sign up.]( # # If you wish to stop receiving our emails or change your subscription options, please [Manage Your Subscription]( SuperDataScience Pty Ltd (ABN 91 617 928 131), 15 Macleay Crescent, Pacific Paradise, QLD 4564, Australia