AI researcher, Cooperative AI, AI for Social Good, Multi Agent Systems, Game Theory, Evolutionary Biology. Opinions are my own.https://bsky.app/profile/eaduenez.bsky.socialhttps://bsky.app/profile/eaduenez.bsky.social/post/3lrqzjq4rjs2mI once hoped discrimination was a dying relic. But what if it's a cognitive "bug"? At #DeepMind, we asked a critical question: Could bias emerge on its own, even in AI without any human social baggage? Our new research in @pnas.org has some startling answers. #AI #Bias #Psychology #CognitiveScience16 Jun 2025 22:25 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lrqzjq4rjs2mhttps://bsky.app/profile/eaduenez.bsky.social/post/3lkl5b4nrs22oHonoured to give a Keynote at Doing AI Differently organised by The Alan Turing Institute, @edinburgh-uni.bsky.social, and @ukri.org. "Defending human autonomy in the age of AI enshittification", what I consider to be the single most pressing problem facing humanity. https://vimeo.com/event/4981744/73d3306014#t=1h1m58s17 Mar 2025 12:19 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lkl5b4nrs22ohttps://bsky.app/profile/eaduenez.bsky.social/post/3lifgliwp2k2iAnother example of bad AI practices, with a great take from Pluralistic. TL;DR, no, you cannot train an AI Vision model to predict MBA performance from headshots (other than to measure preexisting human bias). https://pluralistic.net/2025/02/17/caliper-ai/17 Feb 2025 19:00 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lifgliwp2k2ihttps://bsky.app/profile/eaduenez.bsky.social/post/3lgsrkzlh222nHere's how to save trillions in government spending... Hint, Elon wouldn't like it: https://prospect.org/economy/2025-01-27-we-found-the-2-trillion-elon-musk-doge/28 Jan 2025 15:31 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lgsrkzlh222nhttps://bsky.app/profile/eaduenez.bsky.social/post/3lfzcrgdyas2rFantastic take on... well... rationality, mostly, but specifically about how revolutionaries need to think about consequences of their tearing down the status quo https://pluralistic.net/2025/01/13/wanting-it-badly/18 Jan 2025 12:29 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lfzcrgdyas2rhttps://bsky.app/profile/eaduenez.bsky.social/post/3lekujpoyjs2hIn collective intelligence smarter units are not always better: https://www.pnas.org/doi/10.1073/pnas.2414274121 For AI the lesson is that just making AIs more intelligent isn't necessarily going to make them better tools (or companions, or assistants)31 Dec 2024 01:12 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lekujpoyjs2hhttps://bsky.app/profile/eaduenez.bsky.social/post/3lbrti5tlf22nEnforced return to office leads workers to seek new jobs https://www.theguardian.com/business/2024/nov/24/enforced-return-office-rise-people-seeking-new-jobs Thank you, Captain Obvious! 😁25 Nov 2024 15:26 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lbrti5tlf22nhttps://bsky.app/profile/eaduenez.bsky.social/post/3lbrsuqwnps27Advice for budding AI researchers: read a paper every week. It almost doesn't matter which paper. The goal is not to know the whole field (that's impossible), but to get very comfortable quickly understanding what an article is about and evaluate it's contribution in the context of the field.25 Nov 2024 15:15 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lbrsuqwnps27https://bsky.app/profile/eaduenez.bsky.social/post/3lbrqpbqjuk27Anyone knows how to quickly follow back people who follow me? Bonus points if there's a way to do it by having some profile information alongside the decision point. I find it hard to just click on every person one by one, look at their profile, decide, and go back.25 Nov 2024 14:37 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lbrqpbqjuk27https://bsky.app/profile/eaduenez.bsky.social/post/3lbrpg4cgcc27What are the best examples of AI products that are trying to explicitly empower humans? Maybe search, maybe codegen? I think most use-cases are (actually) empowering corporations, and that worries me.25 Nov 2024 14:14 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lbrpg4cgcc27https://bsky.app/profile/eaduenez.bsky.social/post/3lbqdlqtrnr2iHey 'ma, that's me! [contains quote post or other embedded content]25 Nov 2024 01:09 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lbqdlqtrnr2ihttps://bsky.app/profile/eaduenez.bsky.social/post/3lbqcw4b6hc2iI cannot be the only one using peer review to keep up to date with research, right?25 Nov 2024 00:57 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lbqcw4b6hc2ihttps://bsky.app/profile/eaduenez.bsky.social/post/3lbl3qi7utf2dIt's great to see so many fantastic people here, but why does it still feel like doomscrolling? 🫠22 Nov 2024 23:05 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lbl3qi7utf2dhttps://bsky.app/profile/eaduenez.bsky.social/post/3lbl3iv2a4y2dFantastic take on the role of power mediating corporate and user interests: https://pluralistic.net/2024/11/07/usurpers-helpmeets/#disreintermediation22 Nov 2024 23:01 +0000at://did:plc:qpugzrh64et3fjlbd2cbdtxq/app.bsky.feed.post/3lbl3iv2a4y2d