It probably really depends on the project, though I’d probably try and start with the tests that are easiest/nicest to write and those which will be most useful. Look for complex logic that is also quite self-contained.

That will probably help to convince others of the value of tests if they aren’t onboard already.

Yeah they’ve put them in a couple places, It’s pretty bad. Had to work out how to create a custom uBlock Origin rule to block them.

I think calling it just like a database of likely responses is too much of a simplification and downplays what it is capable of.

I also don’t really see why the way it works is relevant to it being “smart” or not. It depends how you define “smart”, but I don’t see any proof of the assumptions people seem to make about the limitations of what an LLM could be capable of (with a larger model, better dataset, better training, etc).

I’m definitely not saying I can tell what LLMs could be capable of, but I think saying “people think ChatGPT is smart but it actually isn’t because <simplification of what an LLM is>” is missing a vital step to make it a valid logical argument.

The argument is relying on incorrect intuition people have. Before seeing ChatGPT I reckon if you’d told people how an LLM worked they wouldn’t have expected it to be able to do things it can do (for example if you ask it to write a rhyming poem about a niche subject it wouldn’t have a comparable poem about in its dataset).

A better argument would be to pick something that LLMs can’t currently do that it should be able to do if it’s “smart”, and explain the inherent limitation of an LLM which prevents it from doing that. This isn’t something I’ve really seen, I guess because it’s not easy to do. The closest I’ve seen is an explanation of why LLMs are bad at e.g. maths (like adding large numbers), but I’ve still not seen anything to convince me that this is an inherent limitation of LLMs.

I’d be happy if we’d just accepted “referer” as the correct spelling for everything, but instead we have the “Referrer-Policy” header, so now I need to check the correct spelling for anything involving referring…

I do sort of like the idea that because we want to keep backwards compatibility on software we just change the language instead since that’s easier.

What sort of features 🤔

Of the 1,723 adults surveyed across the UK, 73% said technology companies should, by law, have to scan private messaging for child sexual abuse and disrupt it in end-to-end encrypted environments.

Found this interesting. I found the survey results here: https://docs.cdn.yougov.com/68pn2b6b57/NSPCC_OnlineSafetyBill_230427_W.pdf

The exact question I believe is being referred to was:

And do you think technology companies should or should not be required by law to use accredited technology to identify child sexual abuse in end-to-end encrypted messaging apps?

This seems like a really bad question, since it implies a coexistence of end to end encryption and big tech companies being able to read people’s messages, which doesn’t really make sense (or at least requires more clarification on what that would mean). The question as it is is basically “do you think child sexual abuse is bad”.

Google yeah, but I don’t think YouTube is too much of a stretch, especially now they have posts and polls and other stuff.

My experience using docker on windows has been pretty awful, it would randomly become completely unresponsive, sometimes taking 100% CPU in the process. Couldn’t stop it without restarting my computer. Tried reinstalling and various things, still no help. Only found a GitHub issue with hundreds of comments but no working workarounds/solutions.

When it does work it still manages to feel… fragile, although maybe that’s just because of my experience with it breaking.

Assuming x and y are totally ordered 🤮

(for anyone wondering, a monad is just a monoid in the category of endofunctors)

Ah, that’s too boring. I have a range of responses to pick from to keep things interesting:

• LGTM
• Nice
• Looks good
• Thanks
• Looks great
• :thumbsup:
• Looks good to me
• :shipit:

For me, no text means “I haven’t really reviewed this properly so don’t want to write anything that could be used against me if (when?) this breaks something in prod”