The last few months have been devastating for LLM dreams: • Apple reasoning paper and the ASU mirage paper and many others confirmed that LLMs still can’t solve distribution shift. • GPT-5 came late and fell short. • Turing Award winner Rich Sutton thanked me for my

Dimitar youtube.com/watch?v=-J_xL4…

It’s really irritating when I say “AI” and people hear “LLMs”.

What's really irritating is when AI experts delude themselves and others in believing they're actually working on artificial intelligence, when all they're doing is computer automation. And they've been doing it for over 70 years. GOFAI is alive and well in the AI community. 🙄😬

in fact the entire point of generative AI is to avoid the hard work of knowledge engineering. that’s not working as well as people had hoped.

New Stanford+Oxford and other top university study just dropped highlighting flaws in AI benchmarking. Benchmarking falls apart when the score on a test does not reflect the real skill people care about, and the paper calls this problem construct validity. The authors scan

Threw this in as the last slide of an otherwise near term technical talk on LLMs/LRMs (..as I was pressed to make some "predictions".. that I normally leave to "futurists"..)🤷‍♂️

"The speed and efficiency of using A.I. to write code is balanced by the reality that it often gets things wrong." A The New York Times op-ed from SCS faculty Mary Shaw and Michael Hilton describes why a CS education matters in a generative AI world. nytimes.com/2025/11/12/opi…

Join AAAI to access discounted conferences, publications, and a global network of experts. Collaborate, learn, and advance your career with resources, mentorship, and leadership opportunities. Be part of the premier community shaping the future of artificial intelligence

All the great breakthroughs in science are, at their core, compression. They take a complex mess of observations and say, "it's all just this simple rule". Symbolic compression, specifically. Because the rule is always symbolic -- usually expressed as mathematical equations. If

In an open letter addressed to von der Leyen on Monday, more than 70 scientists call out the EU’s president for hyping AI, including two members of the UN’s high-level advisory body on AI Stop overhyping AI, scientists tell von der Leyen | Euractiv euractiv.com/news/stop-over…

.Yann LeCun believes that LLMs are a dead end on the path to human-level #AI. “If you are a Ph.D. student in AI, you should absolutely not work on LLMs.” wsj.com/tech/ai/yann-l… via The Wall Street Journal

MIT research suggests 95% of GenAI pilots are failing to deliver ROI. Why? Because we are forcing LLMs to do jobs they weren't designed for. LLMs mastered language. But they don't understand the structured, relational data that businesses actually run on. My new piece in

Dileep George Yes, but to what extent are LLMs solving tasks by capturing the humanlike abstractions / understanding the tests are meant to evaluate, vs. using superficial "shortcuts"? Our latest preprint: they may do more of the latter than you might expect. arxiv.org/abs/2510.02125

LLMs will always hallucinate.

Proceedings of the AAAI 2025 Fall Symposium Series have been published. View here: ojs.aaai.org/index.php/AAAI…

"'Neurosymbolic A.I." has become the hottest buzzword in town.' Whether this mash-up with LLMs will propel the field forward is controversial @nature nature.com/articles/d4158…

Blending ‘old-fashioned’ logic systems with the neural networks that power large language models is one of the hottest trends in AI go.nature.com/48ajJ9K

Join AAAI to access discounted conferences, publications, and a global network of experts. Collaborate, learn, and advance your career with resources, mentorship, and leadership opportunities. Be part of the premier community shaping the future of artificial intelligence

The CEO of Google admits that Google search provides more accurate information that the company’s AI. He acknowledges that Google search is "more grounded in providing accurate information."