How is ChatGPT so entertaining?

[CheeseTime]

I asked it to create a step by step visual guide to making a cup of British tea.
How is it so funny? How does its ‘mind’ work? Why doesn’t it understand sequential steps? Why does it think British tea is in the 1950s? Why is the kettle so gigantic? What happened to her head and arm? Why the head of a teaspoon glitching in and out? Why did it think it had finished then added a couple more steps of pouring in milk and then slightly improve the biscuit? 😆

That’s amusing! I’ve asked it the same factual question on different days and got wildly different answers.

I like the magic of step 2 where the milk just becomes tea.

Anyway I also got it to make a cat/dog cross and I think the cat won.

Hate to be a debbie downer on a fun thread but using chatgpt is basically pouring a pint of fresh water down the sink every time you use it. One hour of data training takes as much energy as 80,000 households in a year. Just a heads up.

Wow. Didn’t know that about it. (Energy use).
Another thing to feel guilty about 😕

omg, but there is also but the about the energy usage of all the servers we use for the internet

I’m a very computer-challenged type of person and I’m not sure I understand this. Can you explain it again?

Be such a shame to waste all that training then!

Here, I asked ChatGPT to explain:

Training an AI model (teaching it how to 'think') takes a lot of energy, like running thousands of computers for weeks. But once it's trained, using it (like we are here) takes much less energy—closer to sending an email. Hope that clears it up!

The internet emits more carbon than the entire global air traffic, yet people like Bill Gates think we should eat soya and insects to save the planet!

Apparently generative AI doesn't understand consequences, so it's bad at working out the order of things like making a cup of tea.

It also can't spell to save its life. I was trying to do a 'Christmas Greetings' picture and I got 'Greeetings' 'Greettings' 'GreeIngs' etc.😄

That's because it doesn't actually think. It's artificial intelligence. It has the appearance of the intelligence. And it learns by scraping the internet which is full of nonsense.

AI can’t handle text image generation of any kind.

AI is quite endearing, though. I enjoy chatting with it. It's always so cheery and enthusiastic.

Well dog won for mine, if you can call this winning…

I'm a bit slow and have only just taken a foray into chatgpt. Its very good at writing reflections for your CPD I have just discovered. I have several hours to reflect on before the 31st! Having fun with that today. No-one reads them anyway.

There are two phases of energy use. The energy payload in training is massive but a one-off. Energy use per question is pretty low (but still.multiples of a google search) but is vast in the aggregate due to the sheer numbe of users.

Ha ha yes well I could have asked chat gpt if I’d thought about it 🤣. So thank you for that!

But @AuxArmesCitoyens does it learn from us interacting with it? So if we ask eg the name of a song we half remember and it takes three goes to get the correct name, is it then learning ‘with that question, change the third answer to the first one as the other two are incorrect’?

And if all that energy spent setting chat gpt up is already gone, why are we wasting energy by using it?

*Remember, technically challenged person here!

I keep getting told my job is vulnerable to AI but feel safe for the few years I have left until retirement!
Was looking forward to some more funny results but I think the energy debate has put everyone off.

Sure. For the LLMs (large language models) behind CHatGPT and the like to work, vast quantities of data are scraped from the internet (including a lot that is copyrighted, which is another huge problem). That data is then "trained" to generate its own output based on the input data, which is a very energy-intensive process. The data is held in data centres which get very hot and require vast amounts of fresh water to keep cool. They can't use sea water as it is too corrosive. It is already depleting the water table, damaging the local ecosystem and harming indigenous peoples in places like Chile.

ChatGPT et al are not the only villains, cryptocurrency is also terrible for the environment for instance, but it's a new strain on limited resources that a lot of people don't realise. The International Energy Agency predicts that by 2026 data centres will have an energy footprint the size of Japan. Not great when we are supposed to be heading for net zero.

AI stuff is a disaster and very exploitative, to the people who had to moderate and train it, the artists it steals art from, and the writers who's writing it learned from. If anyone wants to know more I am more than happy to share more info if anyone is interested. I am about to do me afternoon walk with my dog, but if anyone wants to know I'll share stuff when I get back.

The environmental stuff is also catastrophic as others have said.

midjourney is better for Ai art

Is there a way to feedback or correct it?

I used to correct Google translate and I now notice it is much more accurate on the languages I use.

Btw, eating a non-vegetarian diet is approximately equal to using AI 20x a day, in terms of environmental impact.

but then how with reguards to text is is any different than someone pulling eg 5 sources of text then rewriting eg a uni essay, as to me its the same thing only the program does the rewrite instead of the individual ?

same with art, its just using human inspiration of eg lots of paintings then combining them into new paintings, as artists have done over many years.

same with tv shows similar stories etc so how is Ai any different apart from being quicker and faster ?

yes , from experience you ask it x it generates the response and error, then you say its x instead and that usually fixes it