pnt12

In a war scenario, Russia would could destroy Ukranian phones for example.

For the lulz.

The distinction is interesting, but for a different concern. One thing is to have rogue models breaking stuff, other is to have them following orders, and breaking stuff with success.

This means cloud models can have some guardrails and follow them most of the time (we could model their authorizations like a benign employee with permission to break glass during incidents). But self hosted models can avoid that and help malicious people do advanced attacks, which is a shame.

In principle I'm all for self hosted models, to avoid risks of AI centralization, but they also have other types of risks.

It's a bit of a lazy take - if we don't work, we don't hang out with other people? Why? Humans are social beings and will hang out for many reasons besides work.

The more interesting question is, if this happens, how will society organize. Would companies subsidize our new bohemian way of life? Why? Would the state need to create incentives for them to focus in the right things? Would there be a need for companies and governments? Would we give up all our autonomy?

You're not lying.

"The wait is finally over—the Walmart Museum is back and better than ever!"

"This video series is intended to be updated as time goes on, so go ahead and come back on your next visit to experience more messages!"

Hear hear! It's very harmful to cherry pick one side, if the other is almost objectivibly terrible.

Because some can have critical conversations, with pros and cons, right and wrong doings. But others will lie, cheat and steal, and degrade democracy, with long term impacts.

No need to code in Assembly, just the same high level language you're using agents with.

I don't think llms are an abstraction - they're a delegation.

Completely agree. When more people ignore politics, the more politicians can ignore them.

In my country, there have been pacific protests about housing. They were faced with police brutality, but this only increased the people protesting and making demands. Some politicians initially said they wouldn't change policies, but seeing so many people involved, now some are changing their stance.

Politics have gotten really divisive, but ignoring them is not the answer.

Their stated key goal is avoiding it, sure. But in practice?

My idea is that through first principles and made up probabilities, they can reach any conclusion they want. But there's a veil of authority: "I'm a rationalist, therefore I arrived this conclusion, surely not from rationalization, but from first principles, probabilities and logic! It doesn't matter if it sounds insane, because it uses first principles, probabilities and logic!".

The rationalist community seems great at rationalizing their bias, not at rational or critical thought. They will argue their way into anything - but that makes them good debaters or writers, not good thinkers.

Some arguments about future predictions seem like a sci-fi fan base nerding about their favorite stories, but without any self awareness.

You get a nerds wet dream: nerdy stuff and ego boost. But very detached from reality.

The goal is not to train you, but to limit their liability. They can say they have a track record of ethical training, and then off the track have incentives to do unethical stuff.

There's so many differences, this comparison is quite naive. Out of the top of my head:

* Determinism: you can create mental models with compilers, not so much with agents

* Autonomy: of you understand the code, you can always jump in to do fixed yourself if needed

* costs: most compilers are free

* availability: you may lose access to models for misc reasons (peak usage, price hike, depreciation, etc)

I guess OP was using the user centric approach of measuring success, however naive or ir realistic it is nowadays.

That's the problem: ad based apps want to keep you there, not solve your problems.

I thought about a dating app where you pay for entry and that's it. Then the incentive is that you don't need it anymore ASAP, because the more time you spend there, the more costs.

Theres a network problem, where it must be high quality, successful, and with lots of people before people are willing to pay. Maybe it would need to start with a free trial in a small region to get some traction before, and rolled out slowly.

Nitpick: I think that's called skimpflation (worse quality), a friend of shrinkflation (less quantity) or inflation (higher prices).

I agree with the article, but I like to write, while lots of people just see it as a chore.

To people who don't like to write a ticket or PR description, the AI text is better than nothing.

In Portugal, you can get a good wine for 5 euros, but we are not great at marketing ourselves.