Akshay Iyer
@AkshayIyer22
Top 0.01% in India's toughest exam. Avid reader, learner. Building the future w RL. Columbia University, IIT Bombay.
I built Corta, your second brain. ChatGPT does not help you think deeply, it does not leverage your own intelligence. Corta does. Looking for beta testers, ideally people who think deeply, learn, use tools like Notion. DM me if this resonates, would love to chat!
I hate that so many interesting ideas are constrained by what could turn into a good business Once AI does everything in the world I look forward to getting to freely follow my own curiosities
When we see something, we can decide if it generalises or not. We can remember to apply it elsewhere. How much we learn/generalise is in our control When llms see an example they can’t control their learning on the fly Seems like a missing degree of freedom in the transformer…
Pre internet era companies seemed way easier to disrupt than post internet era tech companies Innovators dilemma fading for ai
The gentle singularity👍
woke up early on a saturday to have a couple of hours to try using our new model for a little coding project. done in 5 minutes. it is very, very good. not sure how i feel about it...
I have met the guys at Astera (where Eli works) and seen their labs, they’re also working on terraforming Mars it’s very very cool
Congrats to the team at @MarathonFusion for identifying a plausible, scalable process for transmuting mercury into gold. This is something that drove Isaac Newton insane.
So optimistic for the future😭 If we do transmutation well, we could go to any part of space, use the background hydrogen to form every element and live there. (hydrogen from around would keep diffusing in) We could inhabit any part of space. Re : the beginning of infinity
Congrats to the team at @MarathonFusion for identifying a plausible, scalable process for transmuting mercury into gold. This is something that drove Isaac Newton insane.
What is Apple even doing they fumbling the bag Unless they release insanely good, dev trainable edge models in iPhones
Insane, RL ftw
1/N I’m excited to share that our latest @OpenAI experimental reasoning LLM has achieved a longstanding grand challenge in AI: gold medal-level performance on the world’s most prestigious math competition—the International Math Olympiad (IMO).
Sutton says with the bitter lesson - keep architectures general and throw compute The brain has areas for language, vision with different architectures, way more data efficient We can probably do better than the bitter lesson, we're just not smart/experimental enough
I do some deep thinking, reading and have ideas that I'd love to share Planning to start posting blogs here on Sundays!
Bret Taylor (OpenAI board chairman) gave a talk yesterday at SPC Said how the market for AI is different from the internet in that it's predictable, Sam and gang were almost completely accurate in forecasting which domains AI will make the most impact in
Great blog on the bets Mira Murati is making with Thinking Machines medium.com/%40adnanmasood…
Most GPT-4/Claude-4 level application layer usecases are already saturated. RL is the unlock that will allow agents in spaces where they had no chance before