JohnMakin

Substitute anthropic/claude with exxon mobil/oil and see how crazy this sounds.

No one group or individual should be saddled with deciding what the “correct” human values should be. That’s literally what religion does.

> Before we can align a model with human values, we must consider what those values should be.

Ok, so you believe that 1) this technology will become so powerful as to surpass humans in pretty much every imaginable way, so much so that you see it as an existential threat to humanity itself, and 2) that you see yourself as a responsible steward to avoiding bad outcomes, but 3), that you need to be the arbiter of what human values should be?

How fucking arrogant, pardon my language. My god.

They so, so badly want to criminalize local models.

edited:

> Interpretability isn’t reliable enough to fully understand model intentions: Being able to read and understand a model’s internal reasoning reduces our reliance on what the models explicitly tell us about why they behaved in a certain way. This could be a useful tool for understanding models’ “thought processes” when they take unintended actions. But interpretability methods are not yet dependable or broad enough to be relied on in this way. As a result, we may miss our assessment of high-risk behaviors, or be able to investigate flawed reasoning only after an incident has occurred.

Just say you have no idea how or why they do what they do. Just the other day I saw it make an obnoxious, huge, irrelevant comment in a block of code - many lines more than the code change (which by the way, I have tons of instructions and hooks intended to stop this, it's maybe 95% reliable at best) - and told it quite plainly, "remove that comment."

What it did instead was remove most of the comment except one line, which was also irrelevant. When I noticed, I asked it why it did that when I explicitly stated otherwise and if it misunderstood what I meant. It happily chimed "No, I did not misunderstand your intent, I just thought this was better."

I can't trust that was the real reason so clearly I can't trust that whatever training causes this behavior results in a reliable tool.

Original reward was 200+ billion, lawyered down to 25, on appeal I imagine much less. Cost of doing business for Meta and nothing changes. In cases like this, it should cost the company serious financial harm. Even $25 billion is nothing to Meta.

After that I knew they were done. ME3 was the last good game they released, imo.

The world of menial jobs is absolutely miserable, depressing, and will grind you to a pulp - it's nearly impossible to get out of, by design.

A bag of doritos is 200+ calories. The exercise involved to burn that off is running or walking multiple miles. It’s way easier in terms of physical expenditure to just not eat the doritos, or skip the sugary soda.

my original claim was that their control plane runs in us-east-1. this is a fact.

here’s one third party, it was pretty well documented as a control plane failure in us-east-1

I'm old enough now to be a protected class, so I don't want people accusing me of ageism, because that's not where this is coming from - but there's a type of person that clearly is on their last job of their career and knows it and I try to avoid them like the plague, I've seen it too many times.

I have always been the opposite, despite mostly having IC roles in my career. Still, I don't enjoy the change, mostly because the UX of this tooling is absolutely miserable when you need to do anything serious with it. I am past the point where I thought it was user error and think it's just fundamentally baked into the ecosystem to be annoying to work with. At least in most management situations, the employee isn't going to gaslight you, question every tiny decision you make (or worse, subvert ones it finds more 'favorable' for itself) or just make up wild, easily disprovable bullshit in a constant stream of verbal fecal matter. I've worked with maybe one person that came remotely close to that and they were easily ignored. This stuff's persistent and unavoidable.

I am getting a lot more done, but I am a lot more tired.

The thought of migrating something at the scale of github to azure gives me the cold sweats.

A lot of it stems from EA (effective altruism) - paraphrased very roughly, it could argue that you are being immoral or selfish by not using the million dollars to invest and make a humanity changing discovery (FWIW I don't believe this, I'm explaining how they view things).

If they harm a billion people, but it benefits trillions in some theoretical glorious future, then it's the right thing to do. Of course it's all bullshit and they bake in a ton of assumptions at the core of their arguments that make the math of greed 'feel' okay, or at least palatable.

I've just accepted it over the years, it's a natural consequence of the Nerd Reich.

Straight up untrue anecdotally and by general sentiment in the industry.

This is absurd, and I expect is a cause of the fact that they seem to use git history as a message queue source of truth between agents. They can do whatever works for them, fine, but I've noticed this has leaked into their training somehow, my agent, despite very strict rules on how to use my git workflow, will do absolutely nonsensical and weird (from a human who has used git for a long time perspective) stuff with git that can only be a byproduct of how they use it.

Hopefully this fixes that, get your weird workflow out of mine.

> AI companies are however claiming the undesired activity, such as hacking foreign government websites, is not part of the design, but part of the unpredictable part of the system. That tries to remove the intent which would in turn remove the criminality of the activity.

In California at least, a DUI resulting in a fatality, if it is a second offense, often carries murder charges and they do convict quite frequently. The argument is the training and education you receive after the first DUI informs you enough to know the action of driving a vehicle impaired could result in serious harm to someone else, and that establishes enough intent legally for a murder charge. There are lots of areas of the law like this.

It can be easily assumed that Anthropic/OpenAI et al know their technology is dangerous - they use it as a marketing angle constantly. So, if you deploy technology you know is dangerous, and it causes harm, that should be more than enough legally if not civilly.

Of course this is all burying the lede a little bit, the real answer is, at least in the United States, the law does not apply the same way to the wealthy/powerful as it does you or I.