Yeah, I guess I didn't appreciate the power and generality of text generation, like at all. You can sense it in my blog post I think; I write about char-rnn as this neat gimmick to generate hallucinated linux source code etc. It didn't occur to me at all that a text generation
@karpathy
-
Timeline Compression: From Char-RNN to Modern AI Capabilities
By
–
Hmm putting aside the specifics of 2030 etc, and just talking about "timeline compression", I think around the time of char-rnn (~2015), if someone said something like this to me, or if I read it anywhere: "It's quite possible that if you make char-rnn bigger and then finetune
-
Claude 4 Opus Solves Problem After 4 Hints, o3 Shows Improvement
By
–
fyi for anyone interested later, e.g. the new Claude 4 Opus gets there after 4 hints https://
claude.ai/share/33072dd0
-a760-4b0f-8b73-6c0bb4a883ad
…
Other LLMs do similar except – o3 didn't get it yesterday but when I tried this morning it did and now I can't tell if that's just due to the new conversation memory -
SOTA LLMs Struggle with Complex Explanations
By
–
It’s ok all sota LLMs don’t get it either and give terrible “explanations” I think it’s too coded. Felt cute might delete later
-
Rethinking Software: From Professional-Built Apps to User Liberation
By
–
I missed this post but love it & the term! Definitely, we currently think of software as something professionals write and maintain for large cost, and as a user you go out searching for 1-of-k app for your need. You're constrained to what exists. To fully "free your mind" Matrix
-
AI-generated content becoming harder to distinguish from real
By
–
actually i was in on the joke (there's too much long-term coherence), but it took me enough time, scrutiny and thought that i enjoyed it as a demonstration of hard it is now to tell. though i don't super enjoy this genre of slop posting more generally 🙂
-
Alignment as Computer Security Problem with Neural Networks
By
–
When people say Alignment I just hear Computer Security (now with neural nets) and it makes more sense.
-
Questioning the Fundamental Definition of AI Agents
By
–
Yeah I don’t think “agents” is used in this way but … it feels a bit wrong. What’s some actual fundamental distinguishing property wrt what already exists? “It happens to use an LLM somewhere” is I think the current usage but imo it’s kind of lame.
-
Digital Agents and Intelligent Systems Already Surround Us Daily
By
–
Hmm disagree. Mac OS is a highly intelligent agent with lots of background tasks. Gmail is. X is. Businesses run many on your behalf, eg anytime you swipe a credit card. There’s lots of highly sophisticated, highly intelligent digital entities we use/dispatch all the time.