Hacker News: madisonmay

New comment by madisonmay in "Show HN: Kreuzberg – Modern async Python library for document text extraction"

madisonmay — Sat, 15 Feb 2025 13:03:00 +0000

pypdfium2 is a great choice and a solid piece of software!

You might want to look into https://github.com/VikParuchuri/surya as an alternative to tesseract. Yes, it's associated with a commercial company, but as you long as you aren't a company with 5M in ARR or $5M in funding it's free to use.

New comment by madisonmay in "LLM4Decompile: Decompiling Binary Code with LLM"

madisonmay — Sun, 17 Mar 2024 13:21:16 +0000

This is an excellent use case for LLM fine-tuning, purely because of the ease of generating a massive dataset of input / output pairs from public C code

New comment by madisonmay in "Our next-generation model: Gemini 1.5"

madisonmay — Thu, 15 Feb 2024 16:55:58 +0000

It's more like saying "I've upgraded to 128GB of RAM, I'll never use my disk again".

New comment by madisonmay in "Llama 2"

madisonmay — Tue, 18 Jul 2023 16:57:47 +0000

See figure-2

New comment by madisonmay in "Show HN: Psychic - An open-source integration platform for unstructured data"

madisonmay — Mon, 22 May 2023 23:38:03 +0000

Why the decision to license as GPL?

New comment by madisonmay in "User In Yer Face, a worst-practise UI experiment (2018)"

madisonmay — Thu, 18 May 2023 11:49:45 +0000

Thanks, I hate it.

New comment by madisonmay in "How are you using ChatGPT internally at your company?"

madisonmay — Sun, 07 May 2023 00:52:26 +0000

Coding aid for unittests. Debugging aid for languages / frameworks I'm not particularly familiar with. Work that requires reformatting. Translating from rough drafts to more polished / professional language. Learning more about domains I don't have much expertise in where I need specific conceptual questions answered.

RLHF: Reinforcement Learning from Human Feedback

madisonmay — Wed, 03 May 2023 20:28:04 +0000

Article URL: https://huyenchip.com/2023/05/02/rlhf.html

Comments URL: https://news.ycombinator.com/item?id=35807341

Points: 4

# Comments: 1

New comment by madisonmay in "When you combine two things that are close, but not the same"

madisonmay — Fri, 14 Apr 2023 01:45:52 +0000

Whether or not to split is more a measure of whether or not these two concepts are likely to split down the road than whether or not share similarity today.

New comment by madisonmay in "Introducing Agents in Haystack: Make LLMs resolve complex tasks"

madisonmay — Tue, 04 Apr 2023 01:31:13 +0000

Imperfect systems are still useful, and any sufficiently complex system is imperfect.

New comment by madisonmay in "Petals: Run 100B+ language models at home bit-torrent style"

madisonmay — Mon, 02 Jan 2023 21:30:59 +0000

Interestingly it sounds like offloading could be made quite efficient in a batch setting if you primarily care about throughput rather than latency. Though I guess for most current LLM applications latency is quite important.

New comment by madisonmay in "Ask HN: How to get back into AI?"

madisonmay — Sun, 11 Dec 2022 02:37:34 +0000

Often it might be viable to implement prediction w/o necessarily implementing training (especially if there are published weights or a reference implementation). Not viable for papers where the key contribution is a change to the pre-training objective / training methodology / optimizer, but useful for papers where the key contribution is architectural.

New comment by madisonmay in "Show HN: Automatically fill PDF templates per API"

madisonmay — Mon, 08 Aug 2022 11:39:34 +0000

Guessing data security constraints -- I'm likely in a similar boat.

New comment by madisonmay in "Show HN: Automatically fill PDF templates per API"

madisonmay — Mon, 08 Aug 2022 11:38:51 +0000

I'm getting a variety of CORS errors in console. Maybe this helps:

``` Access to XMLHttpRequest at 'https://api.doqs.dev/v1/organization' from origin 'https://app.doqs.dev' has been blocked by CORS policy: Response to preflight request doesn't pass access control check: No 'Access-Control-Allow-Origin' header is present on the requested resource. ```

New comment by madisonmay in "Show HN: Automatically fill PDF templates per API"

madisonmay — Sun, 07 Aug 2022 21:26:06 +0000

Awesome idea, but website seems unstable. Wasn't able to login after sign-up :/

New comment by madisonmay in "A basic introduction to NumPy's einsum"

madisonmay — Sun, 10 Apr 2022 00:12:38 +0000

For more efficient einsum, see projects like https://optimized-einsum.readthedocs.io/en/stable/path_findi....

New comment by madisonmay in "PyFlow – Visual scripting framework for Python – NodeRED alternative?"

madisonmay — Sun, 16 Jan 2022 19:46:49 +0000

I suppose so, but perhaps trying to prevent the spaghettification has some positive benefits in terms of DRY + code structure.

New comment by madisonmay in "OpenAI disbands its robotics research team"

madisonmay — Sat, 17 Jul 2021 21:22:47 +0000

Wojciech stated this pretty explicitly on his Gradient Dissent podcast a few months back.

New comment by madisonmay in "Show HN: Parrot.vc – I forced a bot to read 65,000 VC tweets and it became a VC"

madisonmay — Wed, 04 Dec 2019 22:35:08 +0000

Sweet! Looking forward to it.

New comment by madisonmay in "Show HN: Parrot.vc – I forced a bot to read 65,000 VC tweets and it became a VC"

madisonmay — Wed, 04 Dec 2019 22:05:38 +0000

@nloui any chance you're willing to share your dataset? Would be fun to replicate this with GPT-2 fine-tuning instead of a Markov chain.