andyk

8 ms·

31.
SlimPajama: A 627B token cleaned and deduplicated version of RedPajama(cerebras.net)
60 points by andyk 3y ago|7 comments
32.
by andyk 3y ago
Saw this mentioned by @dragonwriter in the discussion at https://news.ycombinator.com/item?id=36095032 but it seemed buried/easy to miss there. Also the article that thread is discussing is from 2021, whereas this was
33.
The HTTP QUERY Method specification(httpwg.org)
165 points by andyk 3y ago|95 comments
34.
by andyk 4y ago
Why hypothesize what our grandmas would say when we can ask GPT: you are a grandma sucking eggs recursively. Let's think step by step: 1. Take an egg in your mouth. 2. Suck the egg, swirling your tongue around it to extract
35.
by andyk 4y ago
Thanks for the pointer! I hadn't read this before. I enjoyed it and yeah it's definitely relevant. I knew many folks have been thinking about this stuff, and it is great to accumulate more pointers to any related work. I added a s
36.
by andyk 4y ago
yep.
37.
by andyk 4y ago
Yeah, I cite the ReAct paper in the README in the repo.
38.
by andyk 4y ago
Done! Well, the first suggestion you made anyway :-) https://github.com/andyk/recursive_llm/blob/main/run_recursi... def recursively_prompt_llm(prompt, n=1): if prompt.startswith("You
39.
by andyk 4y ago
> What's the actual goal here? I tried to expand on my goals and paths I want to explore in a comment below [1], but basically I wonder if we can use this sort of technique as a more powerful version of CoT where prompts can break d
40.
by andyk 4y ago
...and to reply to your second question, one thing I find interesting and want to explore further is how (and when) to best leverage what the LLM has memorized. The way humans do math in our heads is an interesting analog: our brain (mind?)
41.
by andyk 4y ago
Yeah, I see the similarities! I like the idea of the prompt containing context that the resulting prompt is going to be executed at the terminal.
42.
by andyk 4y ago
My personal point was just to document my exploration of using prompts to generate new prompts, and more specifically the case where the prompts contain state and each recursively generated prompt updates that state to be closer to an end g
43.
by andyk 4y ago
It is quite unstable and frequently generates incorrect results. E.g., with the Fibonacci sequence prompt, sometimes it skips a number entirely, sometimes it produces a number that is off-by-one but then gets the following number(s) correct
44.
Show HN: Recursive LLM Prompts(github.com)
97 points by andyk 4y ago|66 comments
45.
Show HN: Natural language Twitter search using Codex(perplexity.ai)
5 points by andyk 4y ago|0 comments
46.
by andyk 4y ago
Good question. We are betting that the cost per LLM inference will continue to go down for given level of performance as the market matures over the next few years.
47.
by andyk 4y ago
Yeah, exactly "no code first", which I hadn't heard of before. Cool.
48.
by andyk 4y ago
hal9 is cool, and in a similar spirit. One thing we've observed with many approaches being explored lately is that they generate code into an IDE like experience. We are more of the camp of making the code gen and code run happen under
49.
by andyk 4y ago
Good ideas. Business model wise, beyond this interface we have a demo of natural language to data-code (SQL and Python pandas/matplotlib/etc. APIs).
50.
by andyk 4y ago
Cool project. I feel like the punchline got buried a little in the article. "It is being used by more than 10 organizations for a diverse set of use cases, including: model training on GPU/TPU (3x cost savings), distributed hyperp
51.
Tips for self-study and independent research(andykonwinski.com)
3 points by andyk 6y ago|0 comments
52.
Free live stream of 1st Apache Spark Big Data Summit Dec 2-3(spark-summit.org)
9 points by andyk 13y ago|0 comments
53.
1st Apache Spark Summit agenda posted, 18 talks by edu & biz(spark-summit.org)
4 points by andyk 13y ago|0 comments
54.
The UC Berkeley Big Data AMP Camp, August 29-30(ampcamp.berkeley.edu)
37 points by andyk 13y ago|5 comments
55.
AMP Camp 3 Big Data Bootcamp(ampcamp.berkeley.edu)
12 points by andyk 13y ago|0 comments
56.
by andyk 14y ago
Awesome.