It's going to be very interesting. There are already some copyright owners that provide special models of their own which allow them to attribute percentage ownership from the training data to the generated output. They can do more if they have the full knowledge of the training corpus and details of the model weights, but they don't need that to make "educated guesses" which they can use to sue and then force access to the model that was used to generate the output.

You are assuming integer cups of coffee. However, whilst I'm not sure that coffee is delivered in real number quantities, we certainly recognize the request "I'd just like half a cup more", so you should also include 3.5 and 4.5 at the very least.

No, probably / possibly the opposite, however I am suggesting that GSM / 2G is much weaker and that 3G is also known inadequate. You want to lock to LTE or better.

I was really pleasantly surprised by that when I found out, but I realized that ± is in latin-1 so it's allowed on the web even though it's not in traditional ASCII.

Might not be 100% true. If the van is integrated with the existing cellular network they could get the needed credentials to mimic the network from the network, in which case they would look exactly like part of the network. That's a different worry, though.

Locking your phone to LTE mode and preferably your home network (might work wierdly in the US?) or better turns on network authentication which means that this only works with the collaboration of the network, which limits it to legal wiretaps. Legal wiretaps on the phone network could use the systems built into the network anyway, beyond the radio interface, so you weren't protected against those in the first place.

1) You are national/military intelligence and you have various secret long term warrants / exceptions which never come to court. In practice small things like murder and drugs running are calmly ignored as long as they don't involve state security.
2) You don't; the story is largely spreading misinformation; you get a warrant for an area and spy on people in that area. Courts grant these reasonably readily for limited areas.

Much more easily this would be a breach of wiretapping laws such as ECPA and would likely need a warrant to be legal, either under FISA or normal criminal law. No need to get into a complex constitutional case. It's also not going to work if you switch off old mobile modes and lock your phone into LTE, 4G, 5G or greater modes of network use, a feature which is definitely available on the free phoneOSs like LineageOS and GrapheneOS.

Totally disagree. There was nothing quite like the satisfaction of finding out that trans fats are worse than butter. Absolutely you should listen to studies that agree with what you are doing already and possibly even gloat about them to your "healthy" friends who keep switching from trend to trend. Do use this study to wind up people who tell you that you should give up coffee for health reasons.

On this site we call it 4±1.

Never forget your error bars.

You can do both.

The AI is trained on the output of human developers and so has the same problems as human developers. Computationally, a "wrong" program can be just as valid a result as a "right" program. You only avoid the wrong program through some form of constraint. If your constraints are all in your context window, rather than in your development environment then

a) they push other things that might also be important out of the context window
b) they are subject to the misunderstandings of the same constrains that exist in the training data
c) they are also subject to randomness and later decay

If, instead, you get the AI to hard wire better behavior in by having it choose to write the code in Rust, then the AI can use more of the context window to contain constraints against logic bugs.

Freedom of speech is what gives them the right. The board was created by an evil company, sure, but it is possible for other companies to try to be even more evil than Meta. If what they found is true, then it's good that they publicize it even if they themselves or the people that created them don't have 100% pure motives.

Yes, it's a major problem that the AI models have serious biases. I even experimented with getting criticism of both sides in some of the ongoing conflicts. The AI would normally only argue against whichever sides were democratic.

Not sure I follow your logic on the second part, but the first is clear. How much do you think it was worth the AI companies to pay him in order to get his negative opinion of their work out there? Do we have any evidence that OpenAI didn't pay for this opinion?