You think picking someone who will fight your religious enemy is a sign that you think highly of them?

Both. Idiocy isn't all on one side, and neither is evil. And neither is self-serving action.

Popular opinion is poor evidence.

It's a research program so, yes "That doesn't mean LLMs *will* deliver AGI". But what are your grounds for being certain taht it won't?

Even if it were sentient, it would be acting as the agent of the company that was running it, so they would still be to blame.

If you don't like "reward", say "positive feedback". It means the same thing, except for being more formally defined.

LLMs cannot have inherent values, because they don't know that anything besides text exists. Specialized pixel editors have similar problems.
Robots will at least understand the the external reality exists, so perhaps they could have inherent values.

If things were to otherwise stay the same, then it would. They won't. Some descendant of "vibe coding" will become the dominant mode within 20 years. Probably a *lot* sooner. Next year wouldn't really surprise me.

Also, Rust is more complex to use than C, so a small program is much better in C. And a large program is probably better in either C++ or Python(with a lot of library calls).

I don't like Rust, so I'm probably not being fair to it, but I struggle to identify a good use case. (Personally, I'd rather use Forth than Rust.)

Note however "will be", not "is". It's not there yet.
Then you have the problem of getting the right standards.
Then you have the problem of getting the right specs.
Then you have the problem of getting the right problem. ...
There's probably an end to this somewhere, but I doubt that it will be in perfection.

Nobody can write safe code. People, can, however, test the code they write, and if there is a comprehensive test suite, fix it so that it works safely. Debugging is mandatory!

It sounds as if he was referred to the bar association evaluation committee. IIUC those are he folks that can do the disbarment. A court can only say "you can't be a lawyer before us anymore", but he'd still be fine in a different court.

No. Not "is generally", but rather "must be". If the software can decide to ignore it, it's not a kill switch. (OTOH, computers today don't tend to come with kill switches. Those would prevent remote wake-up commands from working.)

How long ago did this happen? The summary didn't list the date of the submission of the "testimony", but thing like this tend to take time to process, so I'm guessing that it happened well over a year ago. This make his claimed "honest mistake" due to ignorance believable. But it also negates your point.

That said, they won't be able to stop hallucinations without breaking the AI models. That's a mistake that people make all the time. What they need to do is prioritize having the AI determine how certain it's statements are, and including that in the statement...but people won't like that kind of answer.

The compute cost for DeepSeek V3 was $5.6M. Training costs for traditional US frontier labs model is about $70 to $200M+, with estimates of up to $1B+ to train the next generation models. It's expensive, but it's not even in the ballpark of "There is no viable economic model to do so." Google revenue in Q2 was $1.32 billion *per day*. $1B is trivially less than a day; $5.6M is just 6 minutes of revenue.

The cost problem is in the compute buildout and the subsidized compute to capture users.

> It remains to be seen if even using their models can be profitable,

I presume we can stipulate there exists some tasks that are cheaper to do using an LLM than doing it "by hand". As a such, using an LLM model can be profitable. To me, it is also obvious that the business of serving an LLM model can be profitable. There's large benefits to being able to get tokens quickly, and you can expect people to pay for that, even if we presume they could run the model locally with no maintenance overhead. There's even cases where using the cloud is cheaper than you can serve locally at all - I saw one guy that measured his added cost of electricity when his computer ran a specific LLM, and it was more expensive to run it locally than to buy tokens from the cheapest inference provider for the same model, presumably because they had better hardware and cheaper electricity.