Next was an essential panel on the British public's attitudes on #AI at the #AlanTuringInstitute w/Florence Enock, Roshni Modhvadia, @helenzmargetts, @ShannonVallor, Ben Lyons, and Gina Neff. This report focuses on specific applications and moves beyond what various experts think and to what people actually want - finding attitudes much more accepting of AI than skeptics but also heavily prioritizing explainability over numerical accuracy. Highly recommend https://www.youtube.com/watch?v=sUG6y_E2UD4 (7/9)

Next was a nice talk by Santtu Tikka on identifying causal effects at the #AlanTuringInstitute. This is a bit technical, but a good look at a tough problem https://www.youtube.com/watch?v=kwjk8xwR-cI (3/7)

First was an excellent pair of talks by Anthony Cohn and Yarin Gal on commonsense reasoning in #LLMs and quantifying and expressing model uncertainty at the #AlanTuringInstitute. Both of these talks show the importance of understanding how models are created and trained as well as how designers should better communicate appropriate use cases. Highly recommend https://www.youtube.com/watch?v=p8mYlJt9Ba0 (2/11) #AI

Next was an engaging pair of talks on benchmarking #LLMs by @katie_m_collins and video summarization by Mirella Lapata at the #AlanTuringInstitute. These benchmarks, while not perfect, are better than most, and the challenges of video understanding shows how far #MachineLearning has to go. Highly recommend https://www.youtube.com/watch?v=1RgFz_Yzskk (5/7)

Next was an intriguing talk by Andrew Webster on causal attribution fractions in #health at the #AlanTuringInstitute. Starting with a simple premise - how much of a disease is caused by exposure - and then gets into the details of how to go about mathematical estimating that from data https://www.youtube.com/watch?v=5xaDz6KfMb8 (6/8)

Last was an intriguing talk by Zoe Kourtzi on using #AI for #dementia detection at the #AlanTuringInstitute. There's some encouraging results here, and I love the thoughtful incorporation of a potential diagnosis algorithm into a clinical pathway https://www.youtube.com/watch?v=nBU6Bj9svvg (8/8)

Next was a nice talk on evaluating algorithmic clinical recommendation systems by Baptiste Vasey and Lauren Morgan at the #AlanTuringInstitute. There's also a nice update at the beginning on the state of #AI regulation in the #UK https://www.youtube.com/watch?v=kYD2cofK8_8 (3/6)

Next was a nice primer on the history of #AI from the Clinical AI Group at the #AlanTuringInstitute. It's always nice to reflect on the origins of the field, which still profoundly shapes how computer scientists think about the space https://www.youtube.com/watch?v=969m0L0i8nA (3/10)

Last was a thorough examination of #fairness in commercial #FaceRecognition #algorithms by Santhosh Narayanan at the #AlanTuringInstitute. The talk uses a validated dataset to empirically test different algorithms, and as you might expect the results were not encouraging https://www.youtube.com/watch?v=SSKFFksC7ak (4/4) #AI

First was a nice panel and collection of talks on #art and #AI at the #AlanTuringInstitute. This included a lot of artists who use AI in their work, which I appreciate more than just hearing from technologists https://www.youtube.com/watch?v=5FIyaVRqfSI (2/8)

Last was a a group of shorts talks by Aldo Faisal, Damien Coyle, and Yulan He at the #AlanTuringInstitute on #neuroscience and #AI, AI-enabled #neurotechnology, and machine reasoning, respectively. The talk on AI-enabled neurotechnology, creating responsive #prosthetics for a variety of conditions, was particularly interesting https://www.youtube.com/watch?v=N9Xoe7xgehI (8/8)

Next was an expansive discussion of #data #visualization approaches by Senaka Fernando at the #AlanTuringInstitute. Designing effective visualizations requires considering a whole host of issues, and this talk touches on many of them https://www.youtube.com/watch?v=Kr4Cv65-70s (5/8)

First was a good panel on the role of #ComputationalSocialScience in policymaking with @omarguerrero, Leigh Tesfatsion, Josh Epstein, and Rob Axtell at the #AlanTuringInstitute. A bit too focused on traditional social science problems for my taste, but still a nice summary of some of the work in this space https://www.youtube.com/watch?v=SgMSRJvrOQ8 (2/8)

Next was a fantastic panel on the human #labor behind #AI with Madhumita Murgia, Wendy Gonzalez, Siddharth Suri, and Callum Cant at the #AlanTuringInstitute. Most #AI is driven by click workers training these systems, who are often underpaid and algorithmically managed. This essential conversation gets at how to change this approach and issues that #tech companies should consider when developing ever larger models that require this kind of work. Highly recommend https://www.youtube.com/watch?v=8pIrfngZCLA (3/4)

Last was a nice technical talk on interpreting deep #NeuralNetworks by Bin Yu at the #AlanTuringInstitute. Complex prediction models are always extremely challenging for humans to understand, with decisions that can seem nonsensical. The methods introduced here add more tools to ameliorating this issue https://lnkd.in/egqJ42Ne (7/7) #AI

Next was an important discussion on better images of #AI w/ @taniaduarte , Kanta Dihal, & @tristanf at the #AlanTuringInstitute. Changing misperceptions of the current capabilities of AI is a top priority, and reasoned argument alone won't get society there. #Art can have a huge effect, particularly the imagery organizations use when trying to represent these algorithms. This conversation introduces new images and an approach to develop them. Highly recommend https://www.youtube.com/watch?v=7el1K9Q4clc&t=6s (5/7)

Next was an engaging talk on decentralized #data #governance by Mark Lizar at the #AlanTuringInstitute. The current model of personal data siloed away in different databases with little meaningful user control is problematic to say the least, and the model of operational #transparency proposed here seems like a meaningful framework to start to address these problems. Highly recommend https://www.youtube.com/watch?v=ogjKwTMf4Zw (4/7)

Next was a fantastic talk by Cynthia Rudin on the challenges of using #MachineLearning for scoring systems at #AlanTuringInstitute. I always love a talk that takes a hatchet to the #Compass system, and this excellent talk does that and more, succinctly discussing when and how one should consider using machine learning, how problems in #data can significantly throw off scores, and the importance of #auditing systems even if you design it yourself. Highly recommend https://www.youtube.com/watch?v=sOCUP79m5lQ (3/7)

Next was a nice panel on the challenges of #data sharing at the #AlanTuringInstitute with Tim Watson, Fernando Lucini, Cynthia Dwork, Rachel Dugdale, and Zoe Kourtzi. Personal data sharing with any organization is fraught with ethical concerns, and even techniques like #DifferentialPrivacy can cause more problems than they solve. Lots of these topics are discussed here https://www.youtube.com/watch?v=17lYBh4dV3o (3/5)

First was an excellent talk and discussion on Sylvie Delacroix's new book on habitual #ethics, with @lawrennd, Emran Mian, and David Leslie at the #AlanTuringInstitute. Lots to unpack from this tremendous conversation, and I liked the idea brought up that mirrored @msbernst's recent work on jury learning systems. Highly recommend https://www.youtube.com/watch?v=OHweOGW5Ju4 (2/5) #AI #AIEthics