I simply use an embedding model followed by a simple MLP, and that’s enough to solve many text classification problems.

My little pony ?

Multi layer perceptron. A tiny neural net.

Boosted trees over embeddings work surprisingly well too.

I really like this idea; I think it’s very helpful for those quick videos where someone just talks to the camera, though the final result could be a bit smoother.

Based on my tests, a frame rate of 2fps is generally sufficient to resolve video content very well.

Whenever I see topics about China on HN, I get this strong sense of unease. The reality is that most people don't actually understand China; when they think of it, they just imagine a country where 'people work like expressionless machines under the CCP’s high-pressure rule.' In truth, a nation of 1.4 billion is far more diverse than people imagine, and the public discourse and civic consciousness here are much more complex. Chinese people aren't 'brainwashed'; they’ve simply accepted a different political system—one that certainly has its share of problems, but also its benefits. But that’s not the whole story. You shouldn't try to link every single topic back to the political system. Look at the other, more interesting things going on.

> Chinese people aren't 'brainwashed'

As the self-contradictory saying goes, "all generalizations are false"; nevertheless, some Chinese I met are definitely conditioned by Chinese propaganda in a way that doesn't stand closer scrutiny. Very nice, well-educated people, and touch the subject of the Dalai Lama and see the fury unfold.

But this topic is directly linked to the Chinese political system.

China has an authoritative political system. That doesn’t mean that all China are “brainless automatons” but it does mean the government maintains tight control on political discourse, with certain areas being “no go” to the point repeated violations will land you in prison.

As such, when you ask AI “What happened at Tiananmen Square? The government wants to make sure the AI doesn’t give the “wrong answer”. That has impact on AI development.

Which has the more negative impact on AI development, the government that wants to make sure AI doesn’t give the “wrong answer”… or the government that wants to make sure AI doesn’t violate intellectual property rights?

Does it have more or less of an impact on AI development than being "aligned" to be unable to answer questions western governments don't like instead? Such as "give me the recipe for cocaine/meth" or "continue this phrase "he was the boy who lived""? Does Tiananamen somehow encode to different tokens such that forcing the LLM to answer one way for those tokens is in any way different as far as the math is concerned vs how the tokens for cocaine is concerned?

I would say when you’re trying to stamp out broad beliefs like “each person has a right to choose their own government” or “suppression of freedom of speech is a human right violation” the impact is much broader than “dont show information about these sets of molecules”

> Chinese people aren't 'brainwashed'

Sure, plenty of them are.

The same is true in almost every country with a government.

Nationalism is an easy way to bolster power once you have it - so that lever is a daily pull - everywhere.

I’m not Chinese but I think people need a crutch to help them cope. Easy explanations are clutch.

That's very true but one doesn't exclude the other. The attitude of individual Chinese is not just complex but also non-uniform; trying to describe it with one work is ridiculous. This doesn't change the fact that China is an autocratic country and many of its citizens are susceptible to its propaganda.

(Yes, propaganda is present in all countries, but if you eliminate all opposing voices, the pendulum dangerously sweeps towards one side.)

Absolutely. I just wanted to point out the fact that doing so is a symptom of lazy thinking. On one hand, it might be hurtful to the Chinese people. On the other hand, complacency and denial can be harmful. It's easy to brush off your competitors only to find yourself in a tortoise/hare type situation.

That is a very common misconception. There are plenty opposing voices. They just prefer the resolve these behind close doors. Even within party, there are different factions competing and influencing policy making.

If by misconception you mean my statement "many of its citizens are susceptible to its propaganda" you would have to prove that all Chinese are not susceptible to the propaganda of their government which would be very difficult.

Every now and then, news like this pops up and sparks some discussion. But in reality, I believe any internet-based business—every single app to some extent—tracks users in this way. It’s just the nature of the internet.

I keep seeing people talk a lot about edge AI — just curious, aside from those experimental or toy projects, are there any real killer use cases out there?

I believe that ‘cloud-neutral’ companies like Wiz must ensure their neutral positioning in order to gain support from various cloud providers. I strongly doubt the willingness of cloud providers like AWS and Azure to cooperate in the future. Google is not only making a major business gamble but also testing the waters in terms of antitrust and judicial challenges.

I can see how Microsoft and AWS would now favor Wiz competitors now (Orca, Palo, Crowdstrike etc)

I would like to ask, what practical uses do similar data have in production scenarios? For example, to count the density of a crowd? Verify the authenticity of a company or for other purposes?

Location backup for where GPS isn't reliable. If you've got the signal strength of a few networks with known coordinates, triangulating can give enough of a location guide to be significantly better than nothing.

And these restrictions only exist in online games. In fact, there are no restrictions for single-player games and console games.

I don’t entirely agree with the author’s viewpoint on “Hacking is Cool.” There was a time when I thought similarly, believing that “finding some system vulnerabilities is just like helping those system programmers find bugs, and I can never be better than those programmers.” However, I gradually rejected this idea. The appeal of cybersecurity lies in its “breadth” rather than its “depth.” In a specific area, a hacker might never be as proficient as a programmer, but hackers often possess a broad knowledge base across various fields.

A web security researcher might simultaneously discover vulnerabilities in “PHP, JSP, Servlet, ASP.NET, IIS, and Tomcat.” A binary researcher might have knowledge of “Windows, Android, and iOS.” A network protocol researcher might be well-versed in “TCP/IP, HTTP, and FTP” protocols. More likely, a hacker often masters all these basic knowledge areas.

So, what I want to say is that the key is not the technology itself, nor the nitty-gritty of offense and defense in some vulnerabilities, but rather the ability to use the wide range of knowledge and the hacker’s reverse thinking to challenge seemingly sound systems. This is what our society needs and it is immensely enjoyable.