Nick Aceves
@thenickestnick.bsky.social
Software Engineer @ Lambda Labs. Former Googler. AI/ML hobbyist and aspiring researcher.
I think many people aren't really being exposed to uses beyond one-shot content generation.
I use LLMs daily as tutors. They are extremely helpful if, e.g., you need a quick refresh on an ML topic while reading a paper.
But I think most people are stuck in "write an email to my landlord" mode.
I use LLMs daily as tutors. They are extremely helpful if, e.g., you need a quick refresh on an ML topic while reading a paper.
But I think most people are stuck in "write an email to my landlord" mode.
November 28, 2024 at 3:22 AM
I think many people aren't really being exposed to uses beyond one-shot content generation.
I use LLMs daily as tutors. They are extremely helpful if, e.g., you need a quick refresh on an ML topic while reading a paper.
But I think most people are stuck in "write an email to my landlord" mode.
I use LLMs daily as tutors. They are extremely helpful if, e.g., you need a quick refresh on an ML topic while reading a paper.
But I think most people are stuck in "write an email to my landlord" mode.
Maybe I'm just dumb, but which axis is social and which is economic?
The natural interpretation of the axes as going in the same direction as the text labels does not comport with your claim.
The natural interpretation of the axes as going in the same direction as the text labels does not comport with your claim.
November 28, 2024 at 2:51 AM
Maybe I'm just dumb, but which axis is social and which is economic?
The natural interpretation of the axes as going in the same direction as the text labels does not comport with your claim.
The natural interpretation of the axes as going in the same direction as the text labels does not comport with your claim.
I've been reading the code, and I just want to say this is a cut above most research code. Very organized, well documented, type annotations on everything, assertions where they make sense... Makes it a pleasure to read.
November 23, 2024 at 11:07 PM
I've been reading the code, and I just want to say this is a cut above most research code. Very organized, well documented, type annotations on everything, assertions where they make sense... Makes it a pleasure to read.
This feels like it's actually a pretty important result.
I've been thinking a lot about methods like this for ARC. I am very skeptical of the LLM based methods that have been so popular recently. LPN is more along the lines of what I envision a truly general solution would look like.
I've been thinking a lot about methods like this for ARC. I am very skeptical of the LLM based methods that have been so popular recently. LPN is more along the lines of what I envision a truly general solution would look like.
November 23, 2024 at 10:30 AM
This feels like it's actually a pretty important result.
I've been thinking a lot about methods like this for ARC. I am very skeptical of the LLM based methods that have been so popular recently. LPN is more along the lines of what I envision a truly general solution would look like.
I've been thinking a lot about methods like this for ARC. I am very skeptical of the LLM based methods that have been so popular recently. LPN is more along the lines of what I envision a truly general solution would look like.