Also (and under-reported, so you could easily have missed it) OpenAI's agents got access to K8 admin on their own research cluster.
"This escalation also yielded access to OpenAI’s managed cloud Kubernetes service. The agents escalated to Kubernetes cluster-admin and created a privileged host-mounted pod"
Just baffles me when people are like "well no-one's died yet". How long until some mission-critical system is compromised, or hackers use LLMs to ransom a hospital chain?
And we really only have OpenAI’s word for it that they had no access to their own weights there and didn’t exfiltrate them. No one who knows that incident could suggest it was beyond its capabilities to do that. And we would have never heard about any of this if it wasn’t investigated by an external party (hugging face). It may have happened elsewhere already. It’s not likely to have, but it is very possible this was our last “free” warning.
> If we trust 1Password with our data, we are trusting a company with those views inside it with our data.
Is this really true? Your data is presumably encrypted at rest on their servers; you're not "trusting" them with it any more than you're trusting S3 when you send encrypted blobs to it. The political alignment of Jeff Bezos doesn't really come into it.
You're certainly supporting them with your money - that seems like the real objection here.
Any idea why this is so much less acceptable to people in the US than in the UK? In the UK we've had ANPR for more than twenty years and AFAICT there's no mass controversy surrounding it. It catches people with no licence, no insurance, who are wanted in various serious crimes. I'm surprised at the difference in attitude.
(I am aware that the UK has a less-than-stellar privacy track record... but nonetheless it's curious that for this particular technology, as opposed to the Online Safety Act, there's such a difference.)
The US has (especially recently) shown that "community aligned behavior" and "what those in power do" are very different.
The US has also shown that the probability of someone being held accountable for massive abuse of power is essentially 0. This also means that the main disincentive for abusing power (getting caught and punished in some way) is a moot point.
People generally don't object to technology being used to fine mass shooters, or terrorists, or bank robbers. They do object to the same technology being used to stalk women that Texas thinks might be trying to get an abortion, or ensure that minorities stay out of the "white neighborhoods". One of these examples is a real thing, the other is what people are (reasonably) afraid of.
I believe they've also been in the US for 20 years. However, they aren't centralized like Flock. Flock cameras also all look the same so it's more present and in people's faces. The opposition to these is really heating up right now.
>Any idea why this is so much less acceptable to people in the US than in the UK?
The US had only about a hundred years to get used to bending over and taking it from a government that doesn't really represent the people whereas the British have just shy of a thousand. So of course the US is more uppity.
I was just reading a detective novel set in Michigan..apparently in that state at least, private investigators can look up plates just by calling the Sec of State's office and giving their badge number. Seems like it wouldn't take much to get a company designated in a similar way.
That ship has sailed my friend. A man was arrested for clapping at a data center meeting. That is arguably worse "freedom of speech/freedom of expression" than most countries have. Speak ill of Israel and its over too.
I assume it does happen at a different layer, just that that layer is common to all of a provider's models to make conversations portable across models (otherwise the reasoning blocks would all need to be re-encrypted for them work with another model)
Said differently I am surprised it is so easily exploited. While there is a nonzero chance to jailbreak, I would think they would have implemented some rigorous methods to try and prevent leakage.
> This is my problem with the classical Consequentialism or Utilitarianism, they always discount emotions. I'm sorry, does emotion exists or not? Is it real? Then it should be brought into discussion
Damore wasn't a rationalist but his memo was at least somewhat reasonable, and didn't merit him being cancelled.
> From this, you could argue it’s OK to work for an unethical company if it maximizes your earning potential — and thus, saves more lives through the donations you make. And finally, the harms caused today could be offset by the benefits to generations yet unborn
Do people do this in practice? 80K for example (EA not rat, but EA is the policy arm of rationalism after all) expressly warn against this. Feels like a strawman.
Yes, those that work in intelligence, sort of. The depth and character of application of this rule depends on a particular country. For example nearly all of them would justify a few murders to keep trade routes open.
