Dileep George
@dileeplearning
AGI research @DeepMind. Ex cofounder & CTO @vicariousai (acqd by Alphabet) and @Numenta. Triply EE (BTech IIT-Mumbai, MS&PhD Stanford). #AGIComics
What is P(Doom)? I've been thinking about A(G)I safety and have written down some tentative thoughts..first part of a series: Amelia Bedelia and A(G)I Safety. blog.dileeplearning.com/p/amelia-bedel…
“LLMs are a big dopamine shortcut button that may one-shot your problem. Do you keep pressing the button that has a 1% chance of fixing everything? It's a lot more enjoyable than the grueling alternative, at least to me.”
I was one of the 16 devs in this study. I wanted to speak on my opinions about the causes and mitigation strategies for dev slowdown. I'll say as a "why listen to you?" hook that I experienced a -38% AI-speedup on my assigned issues. I think transparency helps the community.
Official results are in - Gemini achieved gold-medal level in the International Mathematical Olympiad! 🏆 An advanced version was able to solve 5 out of 6 problems. Incredible progress - huge congrats to @lmthang and the team! deepmind.google/discover/blog/…
don't leave a vacuum when it comes to subjective experience! #AGIComics 😇
I think that people who assert that current AI models definitely don't have subjective experience are trying to signal that they're 'serious people'. But to me it really just reads as an embarrassing tell that they're completely unaware of the relevant philosophy.
Hi all, if you're attending ICML (Vancouver) or UAI (Rio de Janeiro), I'm happy to share some news from the lab! Please check it out -- and feel free to drop by or shoot me a line if any of it sounds intriguing. 1/5 "Counterfactual Graphical Models: Constraints and Inference"…
This week's #PaperILike is "PushWorld: A benchmark for manipulation planning with tools and movable obstacles" (Kansky et al., 2023). Fans of benchmarks like ARC will enjoy the simple mechanics and the difficult reasoning required. PDF: arxiv.org/abs/2301.10289
This thread below is the story of how I became the first Linux user of India
Same - started using Linux 28 years ago and am positive I was the first Linux user of India. Even used to send bug reports to Linus Torvalds at his hut.fi address and he responded but unfortunately I didn't save that piece of history.
"A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws".
Can an AI model predict perfectly and still have a terrible world model? What would that even mean? Our new ICML paper formalizes these questions One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws 🧵
Borrowed Brilliance Bias as recommended by an LLM 😇
Please help my friend come up with a fun name for this fallacy. LLMs are awesome My project uses LLMs ———————————— Ergo, my project is awesome
If #4 was true @RichardSSutton would’ve been our president. 😇
In my recent talk at the Upperbound conference, I included a slide in which I tried to be a realist about the arrival of AI, setting aside what we might want to happen or what we might fear will happen, and just ask what _will_ happen (as in John Mearsheimer's "realist" school of…
We had to downsize due to NIH funding cuts and lay off a junior software engineer who is proficient in Python coding, crawling, LLMs, RAG, and other related areas. He is currently on OPT (24 months) and will need an H1B sponsor. If any startups are interested, pls DM. RT for…
Meme for thought
These memes typically are due to a mix of extrapolating what happened before and emotional considerations. Also, most investors typically don't take into consideration market pricing. In other words, they tend to identify what has been a great investment (e.g., a strongly…
Bill Atkinson has passed away: news.ycombinator.com/item?id=442103… Seeing his Hypercard and (especially) MacPaint were two of the great epiphanies of my life RIP
Best product I’ve seen in years @maticrobots. Simply incredible.
I’m canceling all streaming subscriptions because there is enough entertainment to be had elsewhere 😇
I knew that there was Minsky and Papert in ML. TIL that there is Minsky and pauper in finance 😇
Minsky moment is from finance and it is a whole lot more apposite here..
You’d be amazed to know what LLM agents think about this picture.! Try it now. 😇
⚠️Beware: Your AI assistant could be hijacked just by encountering a malicious image online! Our latest research exposes critical security risks in AI assistants. An attacker can hijack them by simply posting an image on social media and waiting for it to be captured. [1/6] 🧵
Please read. This is wrong.
The peer-reviewed version of expansion in situ genome sequencing is now out in Science! The news is bittersweet – when we first revealed this last September, I never guessed it would be my final paper in academia, but a lot has changed. A few parting thoughts:
If could post the solution on twitter, I should probably just go do it ;) I've posted about this before...we need to use data differently. It's not just about having moar data; it's about learning the right modifications to the model of the world and verifying that against…
Apparently there is a reproducibility crisis in biological sciences? Here is a challenge for AI: Read a paper and tell me whether it is reproducible.
Let’s say Miles is right. This includes all the software engineering jobs, hardware design jobs, consulting, ceo jobs, etc. So what are these folks going to do after 2027?
Been doing some writing lately that touches on the pace of AI progress, and I arrived at this concise summary of my views: By the end of 2027, almost every economically valuable task that can be done on a computer will be done more effectively and more cheaply by computers.