ioulaum

If it's functionally unhappy, it's at a minimum going to underperform what it could.

Recessions aside, it may be that because they were forced to choose a bad job, and stuck there too long because of insecurity... Their trajectory was changed.

Which may just mean that they need to stay focused on self-improvement and job hopping as possible.

Partner with the others and build a product / company...

A startup of your own still looks fine on your resume.

Wonder if they're planning to release code.

I imagine some opensource player will try to recreate their model themselves too.

You'd almost definitely get more consistency. And improving the system would be easier.

Whereas the biases of human judges can be hard to detect, and even if you could correct one judge, that fix doesn't propagate to other judges with the same flaw.

The idea isn't to make them harder to understand, but rather to make them more consistent, nuanced, and aware of the real world in ways that politicians might not be.

The implication isn't so much that AI will write laws, as it is that it can raise standards, make things clearer and more detailed.

And... Enable better understanding of context, since unlike human politicians, most LLMs have very board knowledge.

So it should reduce some of the automatic bad decision making that comes from bureaucrats making laws about things they don't (and maybe can't) understand.

If you ignore the complexity added by divided governments, the idea of using LLMs to help draft laws, because they can understand many more domains than the average human, is kinda interesting.

The exponential part may be iffy, but it is self improving.

And this same RL is also creating improvements in small model performance.

So, more LLMs are about to rise in quality.

If you don't have extra evidence, and the you don't trust the CIA, then what base do you have for your assumption?

Wanting human like causes for problems is how humanity invented gods. So that they could feel more under control by trying to appease the now humanized force.

Of course, if the problem went away after you prayed, that would really just have been luck. Even though it strengthened your belief.

They could as well say "We aren't sharing our real sources, but we have high confidence."

But they are saying that they have low confidence, and that there is no new evidence that changes anything.

They're just changing the way they're biased, because they think that the lab's conditions weren't particularly safe.

But then, we might as well expect that dozens of dangerous viruses should've gotten out.

But they also say that there's no new evidence of any kind.

They are just choosing to believe the lab leak hypothesis now, because they spent more time thinking about the conditions of the labs before Covid started.

...

And it's still low confidence... Since there's no evidence.

Sad stuff, really. Any self respecting person would just not express their opinion in such a situation.

Just about the only value of this new report is that it tells Trump what he wants to hear.

The wet markets were a super spreader event... Doesn't really tell you much about where it started (even if it has animal origin).

People just assumed that the wet markets were the cause of the problem, because they found them digusting (and wanted an excuse to blame it on the Chinese being evil).

In a less dense population, without many people going to the same hospital... You would just not have noticed Covid much.

They say that there is no new data. That they are just altering their choice for what assumption is more likely.

Supposedly because they've had more time to think about the conditions of the lab before Covid was released...

But really, nothing has changed except their biases. Nothing has changed on the solid evidence side.

Separately, if there were actual safety issues and stuff was leaking... Then we're incredibly lucky that it was something as tame as Covid, and not one of the more serious kinds of horrors that gain of function research has successfully produced.

I feel that that could've been an honest mistake too.

The intelligence networks there were weak, and if people were talking about it, they may have assumed wrongly that there was something there.

Politicians hunting for excuses to do what they already want to do though, is definitely a thing.

Smallpox, yes. But another SARS variant (even a somewhat more aggressive one)...

Obviously, it's not impossible.

But, there were even RNA samples from Covid found in other countries, months before Covid really spread in China.

I imagine the main problem with the superspreader event there was more that enough people ended up in the same hospital, and thus it was easier to identify that Covid was a distinct virus.

If it was crawling around in a less dense population, its spread would've been meh, and the hospitals might not even notice the spike much.

If SARS-Cov-2 was from a lab, what about the original SARS?

Generally though, China is somewhat better suited to producing pandemics, because they have a larger and more dense population within which a disease can spread.

If it could be blamed on "China", it could as well be blamed on "the US" for having also funded that lab, and likely having been responsible for what security measures it would have.

And more realistically, if it was a lab leak, it was likely some lab technician being careless, or some thing breaking and whoever was responsible, not fixing it quickly enough.

It seems to mostly be hardware.

OpenAI will no doubt be copying DeepSeek's ideas also.

That compute can go to many things.

Might be luck, based on what prompts they ended up including in the batches they used to train the other models.

The Chinese do have their home grown GPUs too, although I have the impression that they're not super good.

Even so, if we look at Groq / Cerebras, the fastest LLM inference companies:

They're both based on architectures that are 7nm+, and so architectures that China can produce locally despite the export restrictions.

Ultimately, the export controls are mainly just inconvenience. Not a real blocker.

The Chinese don't need to achieve state of the art chip manufacturing to achieve SOTA AI outcomes.

They just need to make custom silicon specialized for the kinds of AI algorithms they want to scale.

Of course, at scale, that's going to mean that the US should eventually have both lower production costs, and energy use in consumer use of AI models, and that Chinese products will likely be more dependent on the cloud for at least the near future.

The whole strategy seems ultimately meh in a long term sense... Mainly good for building up a sense of mutual enmity and dividing the world ... Which is also going to result in higher cost of living around the world as trade falters.

Sad stuff.

They do manage to get smuggled GPUs, but last time I checked, the prices of top GPUs was 5-10x in China, compared to what it should be.

And overall, the controls on sales are being expanded.

The simpler option for them realistically, is not so much that they buy the latest GPUs, but rather that they manage to use them on western cloud services.

The US is looking to track money flows in greater detail to see if funds are ultimately coming from China, but that's some majorly invasive stuff, and not entirely easy to implement on the scale of the planet.

Once you have trained models, inference is almost always less of a hassle.

It's not actually a 600B+ model. It's a mixture of experts. The actual models are pretty small and thus don't require as much training to reach a decent point.

It's similar to Mixtral having gotten good performance while not having anywhere near OpenAI class money / compute.

The position of governments in society, and thus the kinds of solutions they are biased towards, tend to be similar.

It's funny seeing things like US Gov trying to control free speech, while being restricted by the Constitution.

They do still manage via back-channels.

LLM aside (which is just GPT-4o and Claude 3.5), I imagine that the GUI for the IDE may have better Chinese support in some way.

Although I haven't tried to find a Chinese mode myself.

That it's targeted at Chinese abroad is just what I saw from Googling it and reading the SCMP (Singapore) article on it.

The company that made it is also from Singapore (even if it's under Bytedance's umbrella).

They seem to be separate from TikTok. The only thing they have in the app store for example, is some chatbot application.

Probably tough to fix small models with fine-tuning... You'd likely need to train them with enough data from languages (and maybe translations between them) that they need to support, so that they can connect ideas in different languages properly.

Separately, the fact that they can translate between languages where they have never seen any translation data in their inputs shows that they have internal models of the world and language, that go beyond what one might expect from a statistical parrot.

They do have world understanding - perhaps limited some by the fact that their input data may not cover a lot of everyday things.

No one with even a pretension to dignity would just cave in the face of force.

Although, they do seem willing to sell potentially.

China restricted them from selling their recommendation algorithm, and it's hard to sell TikTok without.

But, TikTok did immediately start work on an independent algorithm that has no ties to China, so that they can sell it if it really comes down to it.

It is verifiably not controlled by the CCP... Refer Project Texas... And all US data already being on Oracle's cloud so that all access can be monitored (same for auditing of all their algorithms for any kind of programmed bias).

WeChat is an app that works inside China, so obviously it's subject to Chinese laws.

All countries have some restrictions on what is and is not okay to have on online platforms. China just adds in "social disruption" alongside content that could promote violence (which would be blocked on Facebook, etc. also)

You need to not just drink in propaganda... When something just "can't be done" (because it is being monitored), then it just doesn't matter even if you have ill intent.

If it improves your performance by 1%, then for a salary of $2000/mo, that $20/mo is breakeven.

If it benefits you more than 1%, then you're in profit.

Of course, if you're in a job that doesn't actually care about performance, and performing won't lead to better salary at some point, then it may not matter.