Hacker News: williamtrask

Zero-Setup Federated Learning: Train Models Across Private Datasets with GColab

williamtrask — Thu, 08 Jan 2026 10:45:46 +0000

Article URL: https://openmined.org/blog/zero-setup-federated-learning-google-colab/

Comments URL: https://news.ycombinator.com/item?id=46539543

Points: 1

# Comments: 0

New comment by williamtrask in "A website to destroy all websites"

williamtrask — Fri, 02 Jan 2026 00:34:19 +0000

...with a price :)

Leadership Lab: The Craft of Writing Effectively [video]

williamtrask — Thu, 01 Jan 2026 09:20:45 +0000

Article URL: https://www.youtube.com/watch?v=vtIzMaLkCaM

Comments URL: https://news.ycombinator.com/item?id=46452577

Points: 2

# Comments: 0

New comment by williamtrask in "Everything as code: How we manage our company in one monorepo"

williamtrask — Tue, 30 Dec 2025 21:48:57 +0000

"Conclusion Our monorepo isn't about following a trend. It's about removing friction between things that naturally belong together, something that is critical when related context is everything.

When a feature touches the backend API, the frontend component, the documentation, and the marketing site—why should that be four repositories, four PRs, four merge coordination meetings?

The monorepo isn't a constraint. It's a force multiplier."

Thank you Claude :)

New comment by williamtrask in "You can't refuse to be scanned by ICE's facial recognition app, DHS document say"

williamtrask — Sat, 01 Nov 2025 13:48:05 +0000

tried searching for "noodlesUK" and didn't find anything meaningful

New comment by williamtrask in "A History of Large Language Models"

williamtrask — Thu, 09 Oct 2025 08:48:09 +0000

Nit: regarding (2), Phil Blunsom did (same Blunsom from the article, and who was leading language modeling at DeepMind for about 7-8 years). He would often opine at Oxford (where he taught) that solving next word prediction is a viable meta path to AGI. Almost nobody agreed at the time. He also called out early that scaling and better data were the key, and they did end up being, although Google wasn’t as “risk on” as OpenAI on gathering the data for GPT-1/2. Had they been history could easily have been different. People forget the position OAI was in at the time. Elon/funding had left, key talent had left. Risk appetite was high for that kind of thing… and it paid off.

The Bitter Lesson's Bitter Lesson

williamtrask — Sun, 28 Sep 2025 23:31:54 +0000

Article URL: https://twitter.com/iamtrask/status/1972167087037808995

Comments URL: https://news.ycombinator.com/item?id=45408988

Points: 5

# Comments: 1

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Thu, 25 Sep 2025 16:01:59 +0000

"This is not the reason, the reason is that this data is private. LLMs do not just learn from data, they can often reproduce it verbatim, you cannot give medical records or bank records of real people, that will put them at a very real risk."

(OP) You make great points. I think we're actually more in agreement than might be obvious. Part of the reason you need to "give" data to an LLM is because of the way LLMs are constructed... which creates the privacy risk.

The principle of attribution-based control suggested in this article would break that principle, enabling each data owner to control which AI predictions they make more intelligent (as opposed to only controlling which IA models they help train).

So to your point... this is a very rigorous privacy protection. Another way to TLDR the article is "if we get really good at privacy... there's a LOT more data out there... so let's start really caring about privacy"

Anyway... I agree with everything in your comment. Just thought I'd drop by and try to lend clarity to how the article agrees with you (sounds like there's room for improvement on how to describe attribution-based control though).

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Thu, 25 Sep 2025 00:30:08 +0000

With you on this one. I do think ABC is a step in the right direction to improve things. <3

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 22:22:46 +0000

"The claim that humans need petabytes of data to develop their mind seems completely indefensible to me."

And yet every human you know is using petabytes of data to develop their mind. :)

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 22:21:57 +0000

I'm relatively close to publishing my PhD thesis which is broadly a survey paper of what you're describing. Will share (almost done with revisions).

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 21:14:09 +0000

I think this is the right question to ask. I think it depends on the task. For example, if you want to predict whether someone has cancer, then access to avast amounts of medical information would be important.

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 21:13:28 +0000

This article is meant for a policy audience, so that does keep the technical depth pretty thin. It's rooted in more rigorous deep learning work. Happy to send your way if interested.

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 21:12:53 +0000

I agree with you in a way - that it seems likely that new data will be incorproated in more inference-like ways. RAG is a little extreme... but i think there's going to be middle grounds betweeen full pre-training and RAG. Git-rebasin, MoE, etc.

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 21:07:25 +0000

Yeah Zama's stuff is great.

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 21:02:19 +0000

Agree with you on the nuance.

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 21:01:17 +0000

(OP) YOLO

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 21:00:55 +0000

(OP) fwiw I fully agree with the privacywashing you're describing here, and this piece is advocating for a more rigorous standard than input privacy (homomorphic encryption), which is insufficient to enable data owners to actually retain control over their data (but is a useful ingredient).

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 20:59:36 +0000

This is the right question. If full attribution-based control is achieved, then this would be impossible. And the ingredient you've suggested could be a useful way to help achieve it.

New comment by williamtrask in "Unlocking a Million Times More Data for AI"

williamtrask — Wed, 24 Sep 2025 20:58:49 +0000

This is the magic :)