Where does slop start? If you use auto complete and it is just adding a semicolon or some braces, is it slop? Is producing character by character what you would have wrote yourself slop?

How about using it for debugging?

I also want to say that Linus is still the one merging things into the kernel and he is ahm… opinionated?

For others reading it:

ChatControl 1: allow scanning on voluntary basis (voted down twice recently)

ChatControl 2: mandatory scanning

Well you say it was air printing could be: wet filament, clogged nozzle, extruder or AMS issues.

I think you are making the mistake of attributing intent to an LLM. A LLM does not have intent. It takes the context and generates the statistically most likely tokens that come next. The biography is part of the context.

The fact that it gives different answers based on context purely comes down to how it was trained and that there is no concept of “factual information”.

I’m not defending LLMs, this is just LLMs doing exactly what they were trained to do.

I mean this study literally says that poorly worded prompts give worse results. It makes sense too, imagine you are on some conspiracy Facebook group with bad grammar etc, those are the posts it will try to emulate.

You were just boycotting before it was cool.

Lots of governments want it. But its no excuse doing it in a stupid way. For example EU IDs should have a function that just verifies someone is over 18 without any other info being send. At least the German one does. But its not being used.

Depends on how important it is. Looking for a hint for a puzzle game: never. Trying to find out actually important info: always.

They make it easy though because after every statement it has these numbered annotations and you can just mouse over to read the text.

You can chose different models and they differ in quality. The default one can be a bit hit and miss.

I use kagi assistant. It does a search, summarizes, then gives references to the origin of each claim. Genuinely useful.

They wrote this without googling anything.

I’m missing a bit the fact that this is not a law yet. This is the position of the commission, which the parliament will then need to approve and has to get past the ECHR as well most likely.

It depends on your threat model or how I like to call it: the paranoia level. Since all connections go through the ISP router anyway you won’t really gain that much privacy unless you directly put a VPN on your router.

Here is what you could potentially stop leaking:

• MAC addresses of your devices
• DNS queries if you use unencrypted DNS

Also theoretically, the router could be an entry point to do attacks against your devices.

People who use pfsense mostly do that because they want more features. For example I have an IoT VLAN that cannot talk to the internet.

For privacy the simplest thing would be to try and put a custom firmware on your WiFi router, like OpenWRT.

Everything else is a bit of an overhaul. And in the end, you always have to trust that the WiFi access points manufacturers firmware does not exfiltrate data.

Also, I would just try plugging in to the modem and see what happens. Most likely you’re just wasting power right now with that ISP router.

Yes absolutely. The results are actually useful, they don’t have an incentive to keep you from finding what you are searching for. There is way less copywritten content and if there is, you can just block it.

Whenever I have to go back to “free” alternatives I am shocked by how much worse it is.

The threat model is that all communication is recorded and will be decrypted once the technology becomes available. The question then becomes for how long you want your data to be secure. If its for example 40 years, you need to chose an algorithm today that is still secure in 40 years.

Bruh we are in it.

If you have a drink that creates a nice tingling sensation in some people and make other people go crazy, the only sane thing to do is to take that drink off the market.

I’m not sure LLMs can do this. The reason is context poisoning. There would need to be an overseer system of some kind.

It’s not better than nothing - it’s worse than nothing. It is actively harmful, feeding psychosis, and your chat history will be sold at some point.

Try this, instead of asking “I am thinking xyz”, ask " my friend thinks xyz, and I believe it to be wrong". And marvel at how it will tell you the exact opposite.

Its probably talking about the UK stratospheric aerosol injection research. Like all conspiracy theories, just enough of a grain of truth.