simonh

You are correct, but the regulatory and liability issue regarding AI is one issue and the government’s liability for having insecure systems is another.

They are independent concerns. A discussion about the first issue is not enhanced by the interjection of the second.

That is a different problem. The point is that an AI doing something on someone’s behalf should be treated the same way as the person doing it.

Because a random web server on a government site having a security vulnerability isn’t news.

> I don't for a second believe that an agent starts trying to hack backend system, when the form or API it has been asked to use isn't working.

Why not, isn’t that classic misaligned AI behaviour?

It would also end the AI. Not to say it couldn’t happen though, if it thought an apocalypse could somehow instrumentally help it fill its monthly paperclip quota.

Yep, this is because for decades Apple has looked for ways strong local clients can add value in the form of bespoke hardware and software capabilities, while Microsoft has looked for ways commoditised generic clients can extract value.

Fortunately some of us have 12” hands to use iPads with so this won’t be a problem further us.

Better support for variable form factors wouldn’t force anyone to do anything. It would incentivise them to use that support.

Danox is just saying (IMHO) that Google can’t be bothered because they don’t need Android to be good, they just need it to be good enough to continue existing.

My favourite take on this. https://youtu.be/kcHKm0cm-jI?is=fqjBSNiJ2jYEXuq1

It not the bananas joke.

Er, would I?

Projecting much?

It very much depends what projects they are coding. One thing agentic coding is really good for is developing highly specific custom applications specifically for your own use case. This can include the kinds of decision support and analytics tasks a C-level relies on to understand how their organisation is functioning, what's actually going on, and where the gaps are.

My team is using coding agents to develop analytics and reporting applications, and an engineering automation application, specific to our needs, that simply would not exist otherwise. The cost and time would be far too great. I can easily imagine there are plenty of cases like this, for all sorts of roles, including managerial ones, where only the person doing the job understands the job well enough to specify a requirement and guide an agent to code exactly what they need.

That definitely happens, but also many people are also using this stuff to get some amazing work done.

I think AI is an incredible competence magnifier. If someone knows what they are doing that competence can get magnified massively. If they do not know what they are doing, that gets magnified just as much. The tricky part is getting people with a low level but real degree of competence, which is all of us when we're starting out, to learn how to reliably magnify that.

Suppose you have a 6 months lead time to order volumes of parts for a new product launch. 4 months before the launch you see a new use case developing that's going to dramatically boost demand at launch.

Even though you saw this coming 4 months before the launch, at the launch time itself you're still caught off guard. Demand is still dramatically higher than your capacity to meet it, and that's still going to be true for at least another few months, and there is nothing you can do about that. What you can do, you did 4 months ago to increase parts orders, but you're still caught flat footed right now.

Guerilla marketing, but never mind, worth it for the bananas joke.

Yes, but apparently the latest free versions don't have the keyboard UI either. In this context old version means keyboard version.

Something can be essential but cheap, or very valuable but so expensive to provide that it’s still not a good business.

10% of an economy is a lot, you don’t need to lose much of that to hit the entire sector very hard.

Not quite failing in the UK, but it’s hitting crop quality and yields hard, such that in many cases farmers will end up with a net loss.

Not a mathematician, so this question may be a bit thick. I see the problem with the set of reals > 0, but is it perhaps that in this case > 0 is the problem and for sets specified as >= 0 it's fine because 0 is a nameable real and the smallest element. Obviously you can't just exclude certain expressions arbitrarily though, so I don't know how you could justify that mathematically.

Those cheeky chappies from Vodkaland with their Polonium water and Novichok perfume don't half get up to some jolly old japes!

These AI companies aren’t capturing that wealth. Frontier models are a commodity.

Except the Asahi Linux team are pretty insistent that Apple isn't making it particularly hard. Yes the hardware is novel, but hardly any hardware vendors share low level details on the Intel side either, all that has to be reverse engineered anyway.

If the ports don't effing break in the first place, who cares?

That's kind of how open source development works, right? We buy hardware and we use software people wrote for free. What am I missing?

That’s what the alignment problem is all about though, isn’t it? AIs always act towards ‘their own’ objectives, that they derive from our instructions.

What we try to do is train them and provide instructions that will result in it having an objective closely aligned to our objective.

OpenClaw was originally designed to run on a Mac, but using remote LLMs accessed via API. Running the LLM locally on beefier hardware came later.

You thought they'd bump the base RAM in the middle of the biggest RAM crisis in history?

Of course, and I'm sure OP wouldn't disagree with you, it was clearly a joke for emphasis. Some people round here need to clean and calibrate their humour detectors more often.

1966

https://en.wikipedia.org/wiki/ELIZA_effect

It turns out the limiting factor isn't how sophisticated algorithms are, it's how gullible humans are.