In This Week’s SuperDataScience Newsletter: The Limitations of Deep Learning. Ukraine Invasion Raises Questions about AI in Warfare. AI Ultrasounds Could Reduce Pregnancy Risks. The Failures of AI in the Age of Covid-19. AI and Psychedelics. 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. --------------------------------------------------------------- [The Limitations of Deep Learning]( brief: In this in-depth article published in Science magazine Nautilus, AI scientist, entrepreneur, and professor in the Department of Psychology at New York University, Gary Marcus, argues that despite deep learning currently being the fashionable side of AI, which may potentially only be a tiny fragment of what trustworthy AI will eventually be able to incorporate, it is experiencing diminishing returns. This is despite the vast amounts of resources poured into it and the buildup surrounding it, claiming “Few fields have been more filled with hype than artificial intelligence.” In this essay entitled ‘Deep Learning Is Hitting a Wall’ Marcus stresses the propensity of language models to be drawn to toxic language. He also questions the real-life applications of the technology, citing the fact that AI has yet to replace a single radiologist as cause for concern, and offers his viewpoint on what may be next for the application of deep learning. Why this is important: Marcus is a New York Times best-selling author and renowned expert in his field. This article is a committed read at over 6,000 words but offers a world of insight into a contentious issue that all data scientists should be aware of. [Click here to sign up!]( [Ukraine Invasion Raises Questions about AI in Warfare]( brief: In the SuperDataScience newsletters, we’ve often discussed the possible uses of AI in wartime situations. Until now this has largely been in abstract terms but the Russia-Ukraine conflict has turned this possibility into reality for many of us in the West. News has come from Ukraine that KUB-BLA kamikaze drones have been intercepted in Ukraine, with leaked images appearing to show damaged drones that have either crashed or been shot down. The KUB-BLA’s are known as “suicide drones” as they are fired from a portable launch and can travel up to 130 kilometers per hour for 30 minutes before deliberately crashing into a target, detonating a 3-kilo explosive. The drones are built by ZALA Aero, who when revealing them for the first time at a Russian air show in 2019, claimed that they feature “intelligent detection and recognition of objects by class and type in real time.” Why this is important: As this Wired article states, these drones are likely to do little to affect the outcome of the war. However, the use of them raises many questions about the role that AI may take in the conflict and whether autonomous weapons may soon determine targets. [Click here to read on!]( [AI Ultrasounds Could Reduce Pregnancy Risks]( In brief: This exclusive story from The Guardian reveals that the UK’s NHS has begun trialing scanning technique that uses AI to analyse ultrasound images. The scheme could help assess the risk of adverse pregnancy outcomes and ultimately reduce them. The trial will use AI to analyse ultrasound images taken during a woman’s 12-week scan and assign them a risk score. Patients who are considered to be at high risk of negative conclusions to their pregnancies, such as stillbirth or pre-eclampsia, could then be offered additional scans or drugs to reduce their risk of adverse outcomes. This is similar to the first-trimester risk assessment for Down’s syndrome which is routinely offered to NHS patients at this point in their pregnancy. Developer of the technology, Prof Sally Collins, said: “What we’ve come up with is a fully automated, artificial intelligence method for seeing and measuring the placenta.” Why this is important: Currently in the UK, approximately eight families are affected by stillbirth per day and approximately 6% of pregnant women develop pre-eclampsia. This trial has the potential to be rolled out more widely and prevent heartbreak for many families. [Click here to discover more!]( [The Failures of AI in the Age of Covid-19]( In brief: In this article from the Harvard Business Review, Bhaskar Chakravorti, the Dean of Global Business at The Fletcher School at Tufts University and founding Executive Director of Fletcher’s Institute for Business in the Global Context, argues that the coronavirus pandemic was an opportunity for AI to shine and show its true potential, which ultimately it failed to live up to. He argues that the pandemic necessitated the use of technology to enable fast, evidence-based decisions and large-scale problem-solving with datasets. However, what was instead revealed was the abundance of bad datasets, embedded bias and discrimination, susceptibility to human error, and a complex, uneven global context – eventually leading to critical failures. Chakravorti outlines AI’s shortcomings and argues that we need to learn from these errors, offering four key steps to improve AI and ensure that these failures won’t be replicated in any future crisis. Why this is important: After two long years of living with Covid-19, the world is beginning to settle again. However, we cannot simply return to the way things once were. Experts like Chakravorti can offer us compelling insights into the failures of technology, which can help us to ensure they never happen again. [Click here to see the full picture!]( [AI and Psychedelics]( In brief: Psychedelic drugs have long been touted as a potential treatment for a range of mental health issues; including (PTSD), addiction, and depression. A sticking point is to be found in the fact that, until now, scientists have had very little understanding of how these drugs - such as LSD, ketamine, and psilocybin - actually work and affect the brain. A new study headed by Danilo Bzdok at McGill University is attempting to rectify this by creating a pattern recognition algorithm that has scoured 6850 accounts of people’s experiences with 27 drugs in order to learn more about how they alter consciousness. The accounts were found on Erowid’s website, a non-profit educational organization that provides information about psychoactive plants and chemicals. The researchers used natural language processing to study written “trip reports” of drug users’ experiences and cross-examine them with brain receptors that the drug is known to interact with. Why this is important: The aim of the experiment is to develop a greater understanding of how psychedelic drugs trigger specific mental states. This applies to both good and bad experiences, such as euphoria, anxiety, or a sense of being at one with the world. Once these links are understood it is hoped that hallucinogens will be able to be used to develop new treatments for mental health issues. [Click here to find out more!]( [Super Data Science podcast]( this week's [Super Data Science Podcast](, Pandas expert Matt Harrison joins us for a review of his best Pandas tips, how to squeeze more data into Pandas, and his recommended software libraries. Tune in and take note! --------------------------------------------------------------- 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