HNHacker News
TopNewBestAskShowJobs

stevenhuang

1,424 karma · joined August 30, 2013

submissionscomments
stevenhuang··on Billion-Parameter Theories
The author is not wrong. You seem unaware of how nascent the field of LLM interpretability research is.

See this thread and article from earlier today showing what we're still able to learn from these interpretability experiments.

https://news.ycombinator.com/item?id=47322887

stevenhuang··on LLMs work best when the user defines their acceptance criteria first
At least it's an easy way for those who don't know that they're talking about to out themselves.

If they'd bother to see how modern neuroscience tries to explain human cognition they'd see it explained in terms that parallel modern ML. https://en.wikipedia.org/wiki/Predictive_coding

We only have theories for what intelligence even means, I wouldn't be surprised there are more similarities than differences between human minds and LLMs, fundamentally (prediction and error minimization)

stevenhuang··on AI Error May Have Contributed to Girl's School Bombing in Iran
What's up with sites detecting adblock and popping up modals so you can't even interact with the page anymore?

Firefox and Chrome on Android.

Guess that's hint enough that this outfit is garbage and not reputable. Flagged and added to domain block lists.

stevenhuang··on LLMs work best when the user defines their acceptance criteria first
LLMs can execute code and validate it too so the assertions you've made in your argument are incorrect.

What a shame your human reasoning and "true understanding" led you astray here.

stevenhuang··on We might all be AI engineers now
Environment impact is overstated. If you've ever looked at the numbers vs your daily carbon impact, you'd realize this.
stevenhuang··on Labor market impacts of AI: A new measure and early evidence
Whether investors will see returns soon enough to service their debt loads is entirely another matter. I do agree the likely course of action is we get a crash of sorts, since the only way their investments pan out is if labour is replaced entirely which of course sounds unlikely in near term.

My point is the cat is out of the bag. It doesn't take massive investments to achieve iterative improvements on SOTA. As long as the technology does not plateau, smaller labs have shown it's possible to advance the frontiers independent of large companies/investments. And as these frontiers advance, more and more of economical knowledge work will be subsumed by AI. I don't see a way out of this, which is why I am a strong proponent of wealth distribution eg UBI.

stevenhuang··on Labor market impacts of AI: A new measure and early evidence
12 months I won't be surprised if there's not much change. But in 5 years? 10? Anything can happen. It is presumptuous to think you can project the future capabilities of this technology and confidently state that labour markets will never be affected.
stevenhuang··on Labor market impacts of AI: A new measure and early evidence
It's telling you feel the need to create a throw away to voice this opinion.
stevenhuang··on You are going to get priced out of the best AI coding tools (2025)
It will never reach $1000/month for the simple fact you can run very capable open source models on your own hardware for next to nothing.
stevenhuang··on How to talk to anyone and why you should
Thank you for explaining. Could be he's a misanthrope, through life experiences or such.

I share the same sentiments as you, it'd be a tragic loss. But saying they'd "end" is well, unlikely. The countries will shrink. Japan population could reach 60M by 2100 if nothing is done. That's still a lot of people and by then other factors will dominate and fertility may rise again.

Humans are adaptive and a lot can change in half a century, so I would not overly index on what projections say. Everything would need to stay static for the projections to matter, which given the rate of technological changes and geopolitical tension, sounds likely.

stevenhuang··on We do not think Anthropic should be designated as a supply chain risk
That is not how Gotham works. The data you're talking about is most definitely not "just there" for anyone to have. The data is provided by the military and IC, Gotham is the data viz layer to make sense of it all. It does nothing on its own.

Again, palantir isn't what you think it is.

stevenhuang··on How to talk to anyone and why you should
Why so quick to moralize? What makes you think your perspective on world population is justified and his isn't?

This could have been an opportunity for both of you to understand each other's perspective. That's why you asked their thoughts on the matter right? It's a shame you let that pass you by.

stevenhuang··on We do not think Anthropic should be designated as a supply chain risk
In the same way operating systems are used on occasion to kill people. My point stands, you don't understand what palantir does. Palantir (Gotham) is a glorified data analytics platform. Gotham does nothing without the privileged data feeds that the IC provides. All Gotham does is process the collected data provided by the military and IC. Palantir is only a data vis platform. What part about this don't you get?

You are tilting at windmills and directing your criticisms at the wrong level. If you want change, change the laws and institutions that allow for this sort of data collection in the first place (of which palantir and orgs like it are not involved with).

stevenhuang··on Ghostty – Terminal Emulator
Tmux does it better
stevenhuang··on BuildKit: Docker's Hidden Gem That Can Build Almost Anything
Container is standard terminology to refer to a running instance of an image. Yes I was being imprecise, substitute container for oci image. But you seem hung up on frivolity and not getting what I'm saying. We are agreeing with each other and just talking in circles. I can see that you don't see that but that's ok. All of this was because I misunderstood what you said initially when you referred to build artifact as the oci image when I thought you were talking about other sorts of build artifacts.

I mean using the CI system to pass in keys or creds. Yes, it's better to build the image with dependencies, but sometimes you can't do that.

stevenhuang··on A Chinese official’s use of ChatGPT revealed an intimidation operation
If you ever allowed the app access to your camera then next time it will be able to access your camera.

No one is suggesting the app need have circumvented standard android permissions. I'm saying it is not surprising the app would try to open the camera on violation of rules.

You might be surprised to know that I do not disagree with what you said. When it comes to data privacy your information is safe from other private companies but not the government.

> You say I'm censored by gov.

When it comes to political inquiry? Without question this is more sensitive in China. My point is that in China apps are coerced to reveal information about their users to a degree where actively trying to take a picture would not be surprising, or much of an escalation. Not that it would be needed anyways since Internet usage is tied to ID.

stevenhuang··on A Chinese official’s use of ChatGPT revealed an intimidation operation
That you think this is made up shows how little you know about what the Chinese netizen deals with day to day.

It is very real and I am not surprised at all something exactly like what op said has happened.

stevenhuang··on The United States and Israel have launched a major attack on Iran
You need to learn your history. The Iranian people have lived under oppressive authoritarianism for 40 years since the Iranian revolution.

Putting it lightly, many Iranian (within Iran) do not support the current regime. Especially not the educated.

stevenhuang··on The Eternal Promise: A History of Attempts to Eliminate Programmers
The thing being automated in this case is human intelligence. If you've been paying attention more and more of economical knowledge work is threatened by advancement of AI capabilities. This is a credible threat. Deny this and you are the one in denial with reality.

Fundamentally unhinged? How presumptive of you to declare with confidence AI will never become more capable than humans.

But I suppose it's fitting. If after all that has happened your priors still have not budged then I'm sorry to say you will probably never understand this.

> I should be extremely skeptical about excitable tech guys predicting big things in short time frames.

Edit: I read your other comment. I don't disagree with you here.

stevenhuang··on BuildKit: Docker's Hidden Gem That Can Build Almost Anything
It was confusing of you to say build artifact to refer to the container itself in this context. Sure you're not wrong because the container is also a build artifact, but in context of CI, build artifacts is the output of running the build using the container.

Hence my confusion of what you meant -- no one's saying ssh keys are in the CI build artifacts. But obviously they can be in the container as layers if people do it wrong, which is bad.

We're talking about the same thing basically. Yes fully defining your inputs to the container by passing in the keys is a good solution.

stevenhuang··on We do not think Anthropic should be designated as a supply chain risk
You don't understand what palantir does.
stevenhuang··on BuildKit: Docker's Hidden Gem That Can Build Almost Anything
Keys leaking into the build artifact was never the concern.

It's about not having the private keys stored unknowingly in intermediate layers of a build container.

stevenhuang··on Sub-$200 Lidar could reshuffle auto sensor economics
People get into accidents not because they don't know with great accuracy how far away an object is.

They get into accidents because they make bad decisions and get distracted.

If AI makes better decisions and don't get distracted, the amount of accidents will already be greatly reduced compared to humans.

Having lidar in addition to cameras will be of marginal benefit (but a benefit to be sure) when you realize what is actually important: proper modeling of the environment. And for this, cameras are better at providing than lidar, so you still will want cameras anyways.

The focus on lidar is really a red herring. You merely push the computational budget you have to understanding a point cloud instead of vision. You're back to square 1 of "how can I properly model the environment given this sensory modality". This is the part that essentially needs human level understanding of the world that you're missing.

As the other commenter says, you deeply misunderstand the problem.

stevenhuang··on Sub-$200 Lidar could reshuffle auto sensor economics
I would bet a large portion of fatalities is from distracted/bad driving, not that human sight was insufficient.

Phrased a different way, I would expect lidar to help marginally, but it is safer driving in general that will bring down fatalities. This could be done with cameras.

stevenhuang··on Child's Play: Tech's new generation and the end of thinking
I don't need to presume; your guess was incorrect :)

It's just not a great heuristic to assume all long comments are AI generated.

stevenhuang··on Child's Play: Tech's new generation and the end of thinking
It must be disorienting for you to be unable to distinguish between AI and human written comments.
stevenhuang··on Nvidia's 10-year effort to make the Shield TV the most updated Android device
Contracts can be broken and resolved with money. Happens all the time.
stevenhuang··on YouTube blocks background video playback on Brave and other browsers
Tubular is great. Though they sometimes take a while to fix playback issues, like the one where videos only played for 10 seconds. That caused a bunch of forks to come out and I used those for a period of time.

Lately I discovered Firefox with background play and sponsor block extensions still work. If this stops working hopefully tubular will be a good backup.

stevenhuang··on AGENTS.md outperforms skills in our agent evals
It's a new technology under active development so people are simply sharing what works for them in the given moment.

> If these aren't signs of a bubble, I don't know what is.

This conclusion is incoherent and doesn't follow from any of your premises.

stevenhuang··on A few random notes from Claude coding quite a bit last few weeks
Alternatively, others just want to get to their destination.
← PreviousPage 5 of 34Next →