QT does a dual-licencing system, you have to pay if you don't want to share back via the GPL. And there is also the AGPL - which could work well in dual licencing for services?

Not quite the same, but there's some overlap in what it solves.

You can just use AI to write a device driver now! It works perfe$Q%"£$*"£$)%) CARRIER LOST

What about a bluetooth receiver that outputs on an audio jack?
something like this? https://www.ebay.co.uk/itm/196...

No matter how much hype there is around LLMs, It's clearly an impressive technical achievement.

It's hard to predict how much better they'll get - or systems that use LLMs as a component.

One simple possibility of such a system could be along the lines of a system that monitors it's main output, has another LLM critique it and does RL training based on that, with some occasional trusted human feedback training both networks (I assume this already exists)

There a good deal of smart people working in this area - I'd be more surprised if we don't get many more leaps forwards, in unexpected ways.

Genuine question - is the data on google drive being used as training data? or just as context while viewing the file?
If it's the latter, it could still be an issue, but more from the chance of private files bouncing around extra servers.
If it's the former, we have much more cause for concern, but I don't get that impression. Perhaps there's something I'm missing either way?

This is on android phones, google already have root, good speech recognition, closed source play services, and everything they need to listen in to audio or transcript already.

I'd rather scam calls didn't happen in the first place, but spotting them with on-device processing could actually help with a real problem

Hopefully we get a decent free software alternative soon.

Not an expert, but there seem to be more details here
Skimming the page, it looks to be things like using a hash in a distributed database, and redirecting if a resource isn't found.
There could be some interesting details I didn't catch, but it seems like the kind of things a competent software engineer would come up with if tasked with scaling up some servers.

ChatGPT (without extensions) won't search the internet, it's a LLM that gives you a direct response "from memory". Good chance of getting something right if it's fairly common knowledge, bad chance otherwise.

If you want to look up something more factual, try perplexity.ai, or phind.com - the way they work is along the lines of taking what you say, reworking it (perhaps with several variations), doing some internet searches, then answering your question with the result of the searches as context.
They still can, and will, get things wrong - but much less often, but sources are cited to check on the origin/accuracy/communication of various bits of information

stable cascade looks as open as sd/sdxl https://huggingface.co/stabili...

These books may well be written by AI, but the detectors have about the same accuracy as a coin flip. They usually work by passing text to a LLM, and asking if a LLM generated the text

Next - do power tool batteries. Would be great to have a few standard shapes for different companies drills/drivers/saws/strimmers/etc

The other way of looking at it is that even just one EU country is in the top 10 GDP

So Every EU country combined would be a very hard to ignore market indeed. Population is currently something like 450m

Come up with an algorithm for your passwords. some kind of hash-like function to be combined with a password
e.g. converting the website name to ascii numbers, multiply them, add 12, squareroot the result, add 6 - stick that on the end of a "regular" password.

It can be much more complex or simpler than the example above - ideally something you could do in your head, but something tough to reverse engineer even if someone got hold of quite a few of your passwords.
This way you don't need an external service to store your passwords, or to have them written down in any form.
Main problem is when different sites have weird requirements, like must or must not have a special character

Just looked at a couple of the patients... Seems to be things like setting the volume on a group of devices over a network.
We an only hope this will motivate google to help push patent law back in a more sane direction - but more likley they will just increase their own portfolio to counter-sue.

Theres a chance it upgrades from a ~4/5 chance of input being processed this frame to ~40/41 chance of input being processed this frame.
Not a big issue to me, but it may have a mesurable impact.