Hacker News: cauliflower2718

New comment by cauliflower2718 in "Prism"

cauliflower2718 — Wed, 28 Jan 2026 00:48:58 +0000

ChatGPT lets you refuse to allow your content to be used for training (under Preferences -> Data controls), but Prism does not.

New comment by cauliflower2718 in "Reflections on AI at the End of 2025"

cauliflower2718 — Sat, 20 Dec 2025 18:50:39 +0000

Regarding medical information: medical professionals in the US, including your doctor, use uptodate.com, which is basically a medical encyclopedia that is regularly updated by experts in their field. While it's very expensive to get a year long subscription, a week long subscription (for non medical professionals) is only around $20 and you can look up anything you want.

New comment by cauliflower2718 in "There may not be a safe off-ramp for some taking GLP-1 drugs, study suggests"

cauliflower2718 — Wed, 26 Nov 2025 17:23:24 +0000

The original title is so much more informative. It might be so informative that many people didn't feel a need to read the article.

New comment by cauliflower2718 in "A definition of AGI"

cauliflower2718 — Mon, 27 Oct 2025 04:10:47 +0000

Thanks for your engagement but it would help if you read my comment the first two times.

You've personally demonstrated that humans don't have to be reasonable and cooperative, but you're not at all refuting my claim.

New comment by cauliflower2718 in "A definition of AGI"

cauliflower2718 — Mon, 27 Oct 2025 03:05:32 +0000

My example is asking for way less than what you're asking for.

Here is something I do not see with reasonable humans who are cooperative: Me: "hey friend with whom I have plans to get dinner, what are you thinking of eating?" Friend: "fried chicken?" Me: "I'm vegetarian" Friend: "steak?"

Note that this is in the context of four turns of a single conversation. I don't expect people to remember stuff across conversations or to change their habits or personalities.

Your goalpost is much further out there.

New comment by cauliflower2718 in "A definition of AGI"

cauliflower2718 — Mon, 27 Oct 2025 01:32:00 +0000

I think this depends on how you measure task.

One common kind of interaction I have with chatgpt (pro): 1. I ask for something 2. Chatgpt suggests something that doesn't actually fulfill my request 3. I tell it how its suggestion does not satisfy my request. 4. It gives me the same suggestion as before, or a similar suggestion with the same issue.

Chatgpt is pretty bad at "don't keep doing the thing I literally just asked you not to do" but most humans are pretty good at that, assuming they are reasonable and cooperative.

New comment by cauliflower2718 in "'Attention is all you need' coauthor says he's 'sick' of transformers"

cauliflower2718 — Fri, 24 Oct 2025 19:41:44 +0000

+1, I am also big user of PGMs, and also a big user of transformers, and I don't know what the parent comment talking about, beyond that for e.g. LLMs, sampling the next token can be thought of as sampling from a conditional distribution (of the next token, given previous tokens). However, this connection of using transformers to sample from conditional distributions is about autoregressive generation and training using next-token prediction loss, not about the transformer architecture itself, which mostly seems to be good because it is expressive and scalable (i.e. can be hardware-optimized).

Source: I am a PhD student, this is kinda my wheelhouse

New comment by cauliflower2718 in "Space Elevator"

cauliflower2718 — Mon, 20 Oct 2025 20:02:28 +0000

I was wondering about this, because the link shows a bunch of birds in the "death zone".

New comment by cauliflower2718 in "The Molecular Basis of Long Covid Brain Fog"

cauliflower2718 — Fri, 10 Oct 2025 16:00:44 +0000

What do you do for managing dependency / tolerance?

New comment by cauliflower2718 in "Which colours dominate movie posters and why?"

cauliflower2718 — Mon, 15 Sep 2025 03:54:07 +0000

I think the author's line between pink and purple is also not very clear. For example, teen spirit appears in both.

New comment by cauliflower2718 in "Making a font of my handwriting"

cauliflower2718 — Sat, 06 Sep 2025 16:01:02 +0000

This is so cute!! I hope you two are having a lovely life together.

New comment by cauliflower2718 in "AGI is an engineering problem, not a model training problem"

cauliflower2718 — Sun, 24 Aug 2025 00:51:47 +0000

It seems like all of the links to more of their work (e.g. "research on deterministic vs. probabilistic systems") are currently broken.

New comment by cauliflower2718 in "Series C and scale"

cauliflower2718 — Fri, 06 Jun 2025 21:29:21 +0000

I've gotten an AI generated email back when I asked for support, which did seem to understand my request at a surface level, but was not actually helpful. That's the extent of support I've received from cursor.

New comment by cauliflower2718 in "The Lost Art of Logarithms"

cauliflower2718 — Thu, 13 Mar 2025 21:46:16 +0000

This follows directly from the fact that exp(x+y)=exp(x)exp(y).

New comment by cauliflower2718 in "A secret poker game you can play on the subway"

cauliflower2718 — Thu, 20 Feb 2025 02:32:31 +0000

Agree -- especially if this was a post about something a real person thought of, but edited with AI (the author is in France, so it seems reasonable to use AI for editing; also it's a small blog that has been around for a while and not a content farm).

For some reason I think I would find it less valuable if the idea itself came from an AI, too.

New comment by cauliflower2718 in "A secret poker game you can play on the subway"

cauliflower2718 — Wed, 19 Feb 2025 17:25:39 +0000

In section "Works in every major city!", the author mentions they are from France.

Given that, it's not surprising that they used an AI to help with translation.

New comment by cauliflower2718 in "Emotional support across adulthood: A 60-year study of men’s social networks"

cauliflower2718 — Sat, 25 Jan 2025 19:21:04 +0000

Is this to be interpreted as most men have two parents, and in adulthood one of them dies?

Or, men have a parent they are close with plus a spouse, and then the parent dies?

New comment by cauliflower2718 in "TinyStories: How Small Can Language Models Be and Still Speak Coherent English? (2023)"

cauliflower2718 — Fri, 03 Jan 2025 11:55:07 +0000

I think basic definitions for LLMs are solidly within the bounds of what we would expect e.g. chatgpt to be competent at. The task (defining terms) is simple and the specific content (basic LLM stuff) is easy to check by anyone who works on the LLM.

I agree with the general sentiment that we should not just blindly trust LLMs though.

New comment by cauliflower2718 in "TinyStories: How Small Can Language Models Be and Still Speak Coherent English? (2023)"

cauliflower2718 — Thu, 02 Jan 2025 20:04:20 +0000

You can ask an LLM exactly this question and it will tell you.

(The answer is billions of parameters)

New comment by cauliflower2718 in "Explaining Large Language Models Decisions Using Shapley Values"

cauliflower2718 — Sat, 28 Dec 2024 16:15:18 +0000

There's a lazy way to submit to arxiv, which is to submit just the PDF, even if you did it in latex. Sometimes it can be annoying to organize the tex files to submit to arxiv. It's uncommon, but the font and math rendering are the standard latex font.