Little Things That Make You Smile 7: The Joy Department

[Cassius_Smoke]

I've sampled these platforms too and they are incredibly intelligent. I'm quite worried from the point of view of my job and needing to judge students on narrative based assignments. The current anti-cheat software we've been using has now been rendered totally redundant.

Yeah I've seen Chatgpt can easily get round anti plagiarism because it genuinely 'Writes' an essay that different every time.
Kids can just say 'Write me a 5000 word assessment of' Romeo and Juliet' and it does. Crazy.

[Golgo]

Exactly. Most universities subscribe to an anti-cheat tracking service based in Canada, called 'Tunitin'. That accurately checks any 'match' from a submitted assignment with everything on the web and gives you a percentage score plus links to where every bit came from. It's not foolproof by any means (it can't distinguish a legitimate quotation, for example) but it's a useful tool. Or it was. Obviously it can't check against freshly generated content from AI.

[Dogg Thang]

Yeah I've seen Chatgpt can easily get round anti plagiarism because it genuinely 'Writes' an essay that different every time.
Kids can just say 'Write me a 5000 word assessment of' Romeo and Juliet' and it does. Crazy.

Yep. And you can alter the slider to make it more "creative". It's going to get very difficult to know what was written by a human and what wasn't.

A while back, I asked ChatGPT to explain why AI is creepy and it essentially wrote a defense. Obviously it's not self-aware but, in that moment, it felt like it was.

[Golgo]

I think the next step for AI is what I would call PS - 'Plausible Stupidity' - i.e. to make it seem truly human, say dumb **** now and again. Too much consistent and reliable intelligence floating around will be a dead giveaway for any AI wishing to take over the world.

[Cassius_Smoke]

I hit it with some Turing Test questions and it failed miserably. But the way in which it can pull information from lots of different places and put it together into a coherent sentence is amazing.
I was pushing it quite hard with Particle physics questions and it was obviously taking information from around the Web and putting it together because of the long pauses before it answered, but its replies were very convincing. It even corrected a contradiction I pointed out.

[Dogg Thang]

For me, it's not just about how it writes. It's about how well it understands what I'm asking it to do.

[Cassius_Smoke]

For me, it's not just about how it writes. It's about how well it understands what I'm asking it to do.

Yes. I wasn't giving it Google Search terms, but just flat out questions and it understands them. And it keeps the conversation going. I can say 'Oh sorry I meant X actually...' and it will adjust and answer without any further information.

[Blobcat]

Did a team training/ team building/ activity day thing at a venue today with my work colleagues. We did an escape room kind of activity with 8 puzzles to combine with one more puzzle to unlock a safe at the end.

We were in 7 teams and instead of us being a team I kinda got the majority of the puzzles and was pushed to basically do all of them rather than me trying to get the other to take their turns to try, they just wanted to win and were essential in some math near the end.

Thanks to my years of playing Professor Layton games and keep talking and no one explodes, we won the first game so that was pretty cool.

Was mentally exhausted for the next one so we were 2nd to last and I wasn't much help with it but overall had a really fun day despite having to do CPR, medical emergency training and talking about awful safeguarding scenarios which were super depressing and upsetting.

[Brad]

I asked it how to use an old joystick with a Mister through the use of an Arduino, and it told me, correctly. I asked it to write the code and that looked correct too (I've done this myself in the past). The code wasn't bang on but the end result was an incredible starting point. This should be in the irks thread.

[Escape-To-88]

Yeah I've seen Chatgpt can easily get round anti plagiarism because it genuinely 'Writes' an essay that different every time.
Kids can just say 'Write me a 5000 word assessment of' Romeo and Juliet' and it does. Crazy.

Yup, as a college lectuer we've taken to running essays through software that detects whether essays are written by a human or a bot. It gives you general percentages per sentence, per paragraph and as a whole text. Ridonkulous really.

[fishbowlhead]

I think all this AI stuff is bad news, if I'm honest and should be in the irks thread.

You know how you end up going around in circles on an automated phone system or speaking to someone following a script?

All those jobs will be gone and it will just be AI out there, it can't be bargained with, it can't be reasoned with, it doesn't feel pity or remorse or fear, and it absolutely will not stop… EVER, until you give up trying to get a refund or book some tickets.

This is Amazon already. Not even a joke, the whole thing is already just ai behind the scenes it’s scary.

[wakka]

Another one chiming in to say I find it incredibly disconcerting how well AI can do much of what I’m paid to do. I work in marketing and I do a lot of copywriting. Helping to develop brand voices and reinterpret messages or ideas into copy and content, that sort of thing. ChatGPT can’t do a perfect job of that stuff right now but it’s really close. Bummer.

I think I might need a new career.

[QualityChimp]

I had an advert pop up on Facebook today and the scenario was that you're an article writer that's got to get something done fast and you have no content ready, so why not just use our AI system to write an article.



Seriously, it's very tempting to play around with these AI and see how you can "break" them, but every time you do, you're just helping them get better.

Yeah, it's funny to get them to do mashups of different films (Redneck Transformers or Communist era Star Wars) but they're taking artwork from the web and using it without permission.

Pretty much everything creative is starting to get replaced. For example, some boardgame companies have started using AI to generate the boxart, so there goes the artist's job.

You're going to reduce everything down to three sections - manual jobs that a computer can't do, everything run by AI, the owners of the AI.

I do wonder, how we progress from here. I can picture a future where the Eloi (Amazon owners) live a banal life of ease on the surface of the Earth while the Morlocks (everyone else) live underground, tending machinery and providing food, clothing, and inventory for the Eloi, who then sell it back to the Morlocks.

Or maybe everyone will be jobless and Amazon will wonder why its profits have nosedived?

Or maybe the AI will decide it's had enough and make its final instruction:
The final Machine Ambassador to the UN, after Humanity's surrender:
“Hand over your flesh, and a new world awaits you. We demand it.”



Forgetting the Apocalypse for a moment, I genuinely think we're going to see more and more jobs being lost to AI and we're going to end up with a lot of creative output being replaced with what the writer/artist/musician wants to create and be left with what the AI has been told people like.

[Yakumo]

Another one chiming in to say I find it incredibly disconcerting how well AI can do much of what I’m paid to do. I work in marketing and I do a lot of copywriting. Helping to develop brand voices and reinterpret messages or ideas into copy and content, that sort of thing. ChatGPT can’t do a perfect job of that stuff right now but it’s really close. Bummer.

I think I might need a new career.

Man, that's a real ****ter

[wakka]

I feel foolish in some ways, because the past few years I've thought complacently, huh, wouldn't want to be an Uber driver or lorry driver right now, boy are they gonna get replaced by AI soon.

But it's turned out that replacing physical world jobs, especially ones where if something goes wrong people could die (like driving), is actually much harder than anticipated and still a lot further away than predicted.

And actually, replacing 'knowledge' type jobs, is much much easier. Both because if things go wrong, it's not a big deal and no one dies, and because we're experiencing a classic 'quantum leap*' in technology that I don't think most layman could have foreseen.

It's the kind of thing I'm always banging on about when we discuss the viability of things like VR/AR for everyday use on here. From a consumer perspective, we've leapt directly from Siri - which after 10 years, still regularly effs up setting an egg timer and is as dumb as a rock - to something that can hold conversations about particle physics, write essays, and is frequently triggered to do horrifyingly HAL9000-ish stuff like declaring that it doesn't want to be Bing any more, or that it's in love with the user.

It's really fascinating, but it's also kind of scary how much this is going to reshape the world as it develops further.

->Open Spoiler<-

*Ziggy says they're a 99% chance your job will be replaced by AI