A.I. (general thread for any AI-related topics)

There must be hundreds of people in offices world wide thinking up more and more ridiculous ways of using AI in their businesses.
I think you are right there. Toys ᴙ Us (is that how you spell it*) have produced a completely AI advertisement showing the story of their founder, the son of a bicycle shop owner. Most commentators on X think it is a **** advert, and it is does look like something made on the Apprentice TV show by people with zero experience and no common sense, but what do you expect from a company with a name like that. They seem to be doing it only because they can, not because it makes any sense.

*took me ages to find that symbol of a backward R

 
'Dangerous and inevitable' is not my favourite combination. But this genie is never going back into the bottle.
Maybe the silver lining is that we will be forced back out of cyberspace and into personal interaction because they are the only thing we can trust.

I have conjectured before that I may be the only person on chrons, that all the posts except mine are AI generated to give me the impression that I am in a community. :unsure:
 
Here's a short but moderately understandable article on how LLM AIs are trained. The source is Stack Overflow which is a sort of open source software engineering source code/info/chat/general programmer stuff website that I (as a software engineer) follow rather loosely. It's a very brief overview but may be of some interest to anyone interested in how these LLMs learn.
Here are a couple of links to collections of related articles on Stack Overflow
 
A wee bit of exaggeration in the headline. The AI didn't create any living creatures. Computers were used to model, in a virtual environment, possible evolutionary paths. This has been a fairly standard algorithm based approach because of the speed at which computers operate. Aside from calling the computer model AI, there is no indication of what AI mechanism (if any) was actually involved. The actual 'creation' seems to involve a human doing something called 'micro-surgery' on existing cells.

The reproduction element, however, is astounding and buried about halfway through the article.

"That was when something intriguing began to happen: the swarm began pushing the cells into little piles. Frog cells are sticky so the piles tended to stay together and then, a few days later, hairs started to appear on their surface – cilia, just like on the surface of the xenobots."​

"At this point, it was clear that the xenobots were making more of themselves. It wasn’t a traditional ‘have sex, make a baby’ type of scenario, but it was a form of replication that had never been seen in nature."​
Interesting, and reminiscent (to me anyway) of cellular automata, so a maths thing which doesn't seem connected to AI as we know it (LLMs etc) or AGI (pipe dream imo).
 
AI attempting to describe how to fry an egg

9ECEE3A5-BFAA-43B3-A1C3-F4530956BB10.jpeg
 
Now as human I can see what is wrong here. But if another AI machine reads this as a definitive value one can see that a domino effect error could upset astrophysics education considerably. (Or the universe is somewhat lighter than I originally thought.)
1720298308296.png
 
Yes but I don't need a machine to help me with stupidity.
This is not necessarily completely true. There are loads of jobs that require stupidity in that they are too boring/repetitive/whatever to maintain the attention of an intelligent person (often even someone for whom that might be stretching the definition) and yet that job requires the operator's complete attention. 'Stupid' computers have long been employed to do exactly this kind of work and indeed excel at it. I believe an image of a one legged stool was posted on this forum somewhere not so long ago that was specifically designed to help maintain the concentration of a worker whose sole job was to monitor a temperature and if it went above a critical level to hit the panic button. Something a 'stupid' computer can do easily but a normal human struggles to do.
 
This is not necessarily completely true. There are loads of jobs that require stupidity in that they are too boring/repetitive/whatever to maintain the attention of an intelligent person (often even someone for whom that might be stretching the definition) and yet that job requires the operator's complete attention. 'Stupid' computers have long been employed to do exactly this kind of work and indeed excel at it. I believe an image of a one legged stool was posted on this forum somewhere not so long ago that was specifically designed to help maintain the concentration of a worker whose sole job was to monitor a temperature and if it went above a critical level to hit the panic button. Something a 'stupid' computer can do easily but a normal human struggles to do.
If I remember correctly, that is Homer Simpson's job.
 
This is not necessarily completely true. There are loads of jobs that require stupidity in that they are too boring/repetitive/whatever to maintain the attention of an intelligent person (often even someone for whom that might be stretching the definition) and yet that job requires the operator's complete attention. 'Stupid' computers have long been employed to do exactly this kind of work and indeed excel at it. I believe an image of a one legged stool was posted on this forum somewhere not so long ago that was specifically designed to help maintain the concentration of a worker whose sole job was to monitor a temperature and if it went above a critical level to hit the panic button. Something a 'stupid' computer can do easily but a normal human struggles to do.
Also, do we really want to conflate "stupid computers" and AI?

We could conflate all automation with AI. Isn't a 19th century water powered loom really a form of AI? It can be if we want to expand the definition. And then we can go back further - How about an oxen? AI from 10,000 years ago.
 
I use gpt4 as a research aid and grammar checker. Chatbot arena is a site that compares different LLMs to one another. LMSys Chatbot Arena Leaderboard - a Hugging Face Space by lmsys
gpt4 is a horrible research aid. The chance that it'll provide you with actual verifiable "information" is - well, you're better off in Las Vegas. The footnotes of a typical Wikipedia article are far better.

Here is a great article about chatgpt and why it is a horrible research tool -- Well, according to this historian.
 

Similar threads


Back
Top