huac

Joined 12 years ago

Karma

3,901

Submissions

1,271

Member since

Mar 30

View on Hacker News

Recent submissions

Comment · 1 year ago

that comment refers to the test time inference, i.e. what the model is prompted with, not to what it is trained on. this is, of course, also a tricky problem (esp over long context, needle in a haystack), but it should be much easier than memorization.

anyways, another interpretation is that the model needs to also make a decision on if the code in the issue is a reliable fix or not too

Comment · 1 year ago

> 32.67% of the successful patches involve cheating as the solutions were directly provided in the issue report or the comments.

Looking at the benchmark, https://www.swebench.com/, about half of scored submissions score under 1/3 correct? So they're either not cheating, or not cheating effectively?

Comment · 1 year ago

> Coding assistants based on o1 and Sonnet are pretty great at coding with <50k context, but degrade rapidly beyond that.

One framing is that effective context window (i.e. the length that the model is able to effectively reason over) determines how useful the model is. A human new grad programmer might effectively reason over 100s or 1000s of tokens but not millions - which is why we carefully scope the work and explain where to look for relevant context only. But a principal engineer might reason over many many millions of context - code yes, but also organizational and business context.

Trying to carefully select those 50k tokens is extremely difficult for LLMs/RAG today. I expect models to get much longer effective context windows but there are hardware / cost constraints which make this more difficult.

Are longer context windows all you need for AI codegen? hua.substack.com

1 huac 1 year ago 0 comments

Comment · 1 year ago

from an AI research perspective -- it's pretty straightforward to mitigate this attack

1. perplexity filtering - small LLM looks at how in-distribution the data is to the LLM's distribution. if it's too high (gibberish like this) or too low (likely already LLM generated at low temperature or already memorized), toss it out.

2. models can learn to prioritize/deprioritize data just based on the domain name of where it came from. essentially they can learn 'wikipedia good, your random website bad' without any other explicit labels. https://arxiv.org/abs/2404.05405 and also another recent paper that I don't recall...

Comment · 1 year ago

real-time full duplex like OpenAI GPT-4o is pretty expensive. cascaded approaches (usually about 800ms - 1 second delay) are slower and worse, but very very cheap. when I built this a year ago, I estimated the LLM + TTS + other serving costs to be less than the Twilio costs.

Comment · 2 years ago

nice work! I wrote a similar library (https://github.com/stillmatic/gollum/blob/main/packages/vect...) and similarly found that exact search (w/the same simple heap + SIMD optimizations) is quite fast. with 100k objects, retrieval queries complete in <200ms on an M1 Mac. no need for a fancy vector DB :)

that library used `viterin/vek` for SIMD math: https://github.com/viterin/vek/

Comment · 2 years ago

reminds me a lot of rmarkdown - which allows you to run many languages in a similar fashion https://rmarkdown.rstudio.com/

Comment · 2 years ago

shouldn't there be more clouds over ocean, as that is where the clouds tend to form?

Comment · 2 years ago

> there needs to be a tool/function calling step before a reply

I built that almost exactly a year ago :) it was good but not fast enough - hence building the joint model.

Comment · 2 years ago

> Current AI (even GPT-4o) simply isn't capable enough to do useful stuff. You need to augment it somehow - either modularize it, or add RAG, or similar

I am sympathetic to this view but strongly disagree that you need a transcript. Think about it a bit more!!

Comment · 2 years ago