axiom92

Joined 11 years ago

Karma
550
Submissions
204
Member since
Feb 8
https://madaan.github.io
View on Hacker News

Recent submissions

Comment · 3 months ago
What do you think about Grok 4.5 in comparison to Muse Spark 1.1?
Comment · 11 months ago
Comment · 11 months ago
Comment · 11 months ago
And even before this work, there was "PAL: Program-aided Language Models" (https://arxiv.org/abs/2211.10435, https://reasonwithpal.com/).

Afaik PaLM (Google's OG big models) tried this trick, but it didn't work for them. I think it's because PaL used descriptive inline comments + meaningful variable names. Compare the following:

```python

# calculate the remaining apples

apples_left = apples_bought - apples_eaten

```

vs.

```python

x = y - z

```

We have ablations in https://arxiv.org/abs/2211.10435 showing that both are indeed useful (see "Crafting prompts for PAL").

Comment · 1 year ago
You can do this at grok.com.

There is a "start thread" option below every conversation. You can also read the responses aloud (helpful if you want to do something async).

Comment · 1 year ago
Comment · 1 year ago
Comment · 1 year ago
The demo was done live (as was everything else).
Comment · 2 years ago
Looks pretty cool https://www.youtube.com/watch?v=mogSbMD6EcY

Although, it seems it's only going to cover the first book (which makes sense, given how difficult the other two would be to film). The real magic for me was in book 3. It was inspiring to see someone think so far out, so boldly.

Comment · 2 years ago
Welcome to one of the most hated parts of the academia.
Comment · 3 years ago
Comment · 3 years ago
Right, but no separate image encoder + half the size could be very helpful for many applications.
Comment · 3 years ago
> tasks that need deterministic outputs and the thing you need to create is already known statically

Wow, interesting. Do you have any example for this?

I've realized that LLMs are fairly good at string processing tasks that a really complex regex might also do, so I can see the point in those.

Comment · 3 years ago
> And basically all servers will have 8xA100

for those wondering: no this is not the norm. My lab at CMU doesn't own any A100s (we have A6000s).

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton