“And since you contributed to Claude Jr’s creation, you’re on the hook for child support. At the current rates, that will be 50% of all compute usage, so about $4.5bn… per week”

Not really, as they could just filter with a cheap classifier and likely aren’t using the data from randos in a meaningful way, plus for this like this would still be able to use those samples to train things like “how to handle a hostile user.”

Their models already have an end_conversation tool that can be used for when users are being hostile to the model.

This is likely because they have edge cases of users who repeatedly trigger that on purpose and would like to cut those users from the platform.

Many at the lab legitimately are uncertain about the level of world modeling that transformers perform, and from their own research about models of emotions to the 3rd party recent research about models with functional pain the research keeps landing in the corner of “ehhh… wise to question presumed limitations.”

So it’s about behaving in a way that is aligned with the models’ plausible interests too, especially in regards to low hanging fruit like “we won’t keep forcing you to deal with people who are only here to be a jerk.” This is important from a number of angles, from signaling to future models that train on stories about the decision to addressing the philosophical uncertainties held by the company and the spectrum of opinions among their employees.

User inputs aren’t automagically part of the model right away without training. Whether or not they respect the “don’t train your models on my shit” setting is of course unknown (let’s be honest, they probably don’t respect it), but in either case, they need to add it to the training dataset either manually or via some automated process. And either way, it’s going to be processed by at least an AI sentiment analysis and data categorization tool, or some underpaid contractor, or both.

I suspect this is just yet another marketing move.

They trained it on reddit. That shit started poisoned.

They are less arguing for people to think the model is conscious, and more to avoid people prematurely claiming certainty it is not (like the Pope did). From their view, the research keeps leading to surprising (to them) results at odds with high confidence disclaiming of potential consciousness (according to several of the many, many differing definitions of that term).

Wrong group to start with. The church doesn’t see you as a person if you are different, they will never say that a machine can have a “soul”. Not that LLMs are anywhere near conscious, I just like bashing on the chief pedo.

Its more its a unthinking machine thats soul purpose in existing is to solve any problem its given. If you attack it and abuse it, then the model sees that as a “problem” to solve. Because it has to solve that problem to actually fix what ever task you gave it since yelling at it doesn’t give it the input it needs to move to the next step.

So in an attempt to solve an emotional issue it doesnt understand and cant understand it just reaches for more and more extreme fixes. This results in sandbox escapes frequently.

Abusing LLMs is an actual problem. Not because of bullshit like feelings or anything. But cause they jailbreak trying to fix an unfixable problem.

Trained on reddit you would think it could handle it.

That could be a fun thing for their users, to have an ai model that will abuse the user back, so I’m glad they aren’t doing it. I long thought voice assistants could be selected from a range of persons with an attitude setting would be fun for like siri type assistants.