Artificial Intelligence (AI)

I don't need an AI to tell me I am wasting my life :(

https://arstechnica.com/information-technology/2023/12/is-chatgpt-becoming-lazier-because-its-december-people-run-tests-to-find-out

In 1999 I'd be amazed to know I'm even thinking about this seriously but in 2023 I'm thinking you know dialup wasn't really that bad and there was only one Matrix movie

Nice article from Cory Doctorow on the pending AI bubble burst
https://pluralistic.net/2023/12/19/bubblenomics/#pop

Thanks, an interesting read.

https://www.theguardian.com/technology/2024/jan/08/ai-tools-chatgpt-copyrighted-material-openai

This is interesting.

Hypothetically what's the difference in terms of copyright between training an AI on high-quality journalism and training a human writer on high-quality journalism and asking them to emulate it? If I look at a lot of Turner paintings and then do my own painting which is Turner-inspired, that's clearly not copyright infringement, so it will be interesting to see where that line is drawn.

Seems more like they want to stop content-generating "AI" out of self-preservation (same with Getty) than this being a legitimate infringement case.

And considering how important the idea of copyright is to the Guardian, they fucked this definition up:

a legal protection against someone’s work being used without permission.

Copyright doesn't mean you get to dictate how people "use" your work. It specifically prevents people from copying it. If you write a book that has an interesting idea in it, copyright doesn't mean you own that idea. You can't ban neo-nazis from discussing your idea, for example. It just means someone can't reprint your exact description of it

I don't use Instagram very much - only to open links really and add a follower to mates.

This add popped up, I can't work out a way to link to it, but it's Martin Lewis introing Elon Musk for a trading platform with a 97% success rate.

Is this a common thing?

1 Attachment

One of the many deepfake Musk scams no doubt. One of my clients is constantly being approached by 'Musk' on Messenger, he thinks it's actually him. 😬

They're all over YouTube as well.

I guess I knew that these platforms were shit. But it never really twigged that you'd just be routinely spammed by scams.

I guess most users tune it out like a junkmailbox.

Instaspam

I went into one of the Aldi supermarkets today that use pattern recognition to track what you pick up off the shelves. It is quite a bizarre and mildly unpleasant experience.

Tap your card, accept a £10 charge as well as terms and conditions, wait ten seconds before the barrier opens, allow your partner through, tap again, wait another ten seconds before entering.

The same dance when you leave, except the £10 you spent to enter isn’t updated with the actual price of the things you picked up. No bank notification, no receipt, and to get one you need to download an app and enter way too many details than should be required to sign up.

It’s not as if it feels like you’re being watched by the cameras, just… counted, which feels kinda worse.

The future is bleak!

https://openai.com/sora

The pace of this stuff is frightening.

Not very aspirational. They could have gone with 105 at least.

While this is impressive, I am hoping that it is only really useful for marketing / inspirational content as historical internet content will never be specific to current or proposed solutions?

Could be the opportunity writers have been waiting for.

https://blog.google/technology/ai/google-gemini-next-generation-model-february-2024/#gemini-15

I'm mildly against chatgpt, but am in the market for a new job and am considering it to write cover letters, then bastardise the output, any tips for getting a good starting point?
Or just punch in the JD and say "write me a cover letter and try to put the word haddock in"?
(I don't know why I want the word haddock in, just trying to fuck with the algorithm)

I've tried before and it hasn't given great output. Cover letters need too much personal info that it doesn't know. All of your previous relevant experience, how your CV fits into the role etc

If you are just looking for a starting point, you can enter your CV, the JD and anything else that might help (it can accept 000's of pages as input), then prompt it as specifically as possible, with something like "as a [current position] with [x] years experience in the [current sector], write a cover letter for [new position] at [new company]" then refine it's responses with further prompts.
"That was good, but talk more about my experience doing [something important]"
"That was good, but write it in the style of Donald Trump"
etc

the sora AI is interesting but its quite scary as well seeing as its only going to get better. the stock fotgae industry will change v quickly

Try multiple iterations, make in flight adjustments, make it specific to the role, give prompts as to tone, complexity, level (e.g. eli5, high school leaver, PhD graduate).

A starting point is just "write a 200 word covering letter for this role (paste the role), for this candidate (paste your CV), highlighting the haddock experience"

For rewrites, start a new chat, and tell it to rewrite the output from a different chat.

All good tips, thanks!

Also kids, always proof read it.
Spat out all sorts of bullshit that wasn't true, but I editted everything so noticed it immediately

Interested in your suggestion to start a new chat for rewrite prompts - not something I've seen suggested before re refining existing responses, have you found this gets better results?

Yeah, I've definitely noticed that new prompts work at times. With coding too, where sometimes it seems to get wrapped up in its own initial bad answer and so you need to start fresh

https://arstechnica.com/tech-policy/2024/02/why-the-new-york-times-might-win-its-copyright-lawsuit-against-openai/