Artificial Intelligence (AI) is progressing at warp speed - A bit nerdy but I think you'll be amazed
8 watchers
May 2024
11:39pm, 20 May 2024
2,195 posts
|
Mark J 🇳🇿
run free wrote: The web says a definition of intelligence is "the ability to acquire and apply knowledge and skills" - I think it is already doing that Working in the industry, I would concur. |
May 2024
9:44am, 21 May 2024
17,094 posts
|
jda
It definitely does some things that can be useful. But it also fabricates convincingly, and if you don't have the knowledge to verify correctness of its output, that's pretty dangerous. And if you do have the knowledge, it may not have gained you very much...
|
May 2024
9:04pm, 21 May 2024
117 posts
|
RicheyJames
I would strongly disagree with the notion that any existing LLM is doing anything which could be accurately described as 'acquiring and applying knowledge and skills' on the basis that none of them have any awareness of what they are doing. It's no more intelligent than the predictive text on your mobile phone. More here for anyone who's really interested: pluralistic.net |
May 2024
9:26pm, 21 May 2024
1,468 posts
|
Sam Jelfs
"The I in LLM stands for intelligence." All these generative models can do is parrot from the corpus used for training... Interesting article on El Reg today theregister.com |
May 2024
9:59pm, 21 May 2024
434 posts
|
DaveG
Intelligence isn't about providing the right answers. It's about being able to ask the right questions. Peter Higgs didn't provide the answer to how particles have mass, but recognised that it's worth asking why particles have mass. We always hear every 10-15 about advances in AI and how it's much better than it was in the past. But it's not until it's possible for computers to produce random numbers that any of these advances will be transformative. I can ask a child to pick a number between 1 and 10 and they can arbitrarily select a number. But asking a computer to pick a number between 1 and 10 is basically asking it to retrieve a number from a list as it will select a certain digit from the microseconds of the request and report that. Once we find a way of allowing computers to arbitrarily report random numbers they might sort the process of creating things. But until then they will just mimic stuff back to us and be able to provide the right answers, but not generate any questions. |
May 2024
6:41am, 22 May 2024
19,212 posts
|
Dave W
But is the child not also just picking a number from a list of numbers from 1-10, just doing it in it's head. There is only a list of numbers from 1-10 to choose from. I'm afraid I don't understand the argument. |
|
May 2024
10:49am, 22 May 2024
435 posts
|
DaveG
It's the arbitrary way a child picks which makes the difference. They can just select a number from the list. A computer cannot. It looks up a list of numbers likely to be random and reports the one its told to. The child has the potential for creativity as it can create new things. At the moment you ask the child any of the 10 numbers could be suggested. The computer just reports something its told to look up, so at the moment you ask the computer you've defined the answer you get back (you just don't know what that answer is until it's reported back to you). Arbitrariness is essential for intelligence, as it sparks inspiration. Computers can't do arbitrary things, or have inspiration. They can process information much faster than we can, but can only do what they are programmed to do. It's similar to the difference between Stevie Wonder and a Stevie Wonder tribute act who can perform his songs perfectly - there's only one of those I'm interested in hearing new music from, even if I would enjoy a concert performed by either. |
May 2024
11:07am, 22 May 2024
3,801 posts
|
cackleberry
A child could also be cheeky and say 12. A computer would not. In the OP video, the AI knows where it is by looking out of the window. Clever... ish. Look at the route pic of the day, if we have knowledge of that place, we know where the picture was taken. The AI is only doing the same. It cannot know something it has never seen before. Can the AI look at a picture of a place it has never seen before, but use the information to have an educated guess at where the picture was taken? Like how we might recognise the picture was taken in Greece because of the colour of the buildings and the plants, but we don't know the exact town because we've never actually been there. I am trying to say (badly) that the AI has learnt from looking at and matching existing pictures. If you showed it something not already in it's database, it wouldn't know. The AI could also just be using the location data from the persons phone. (I do think the uses where it can check things 100x faster than a person can are amazing, like the piece on the news about checking mammograms faster and more accurately. Very good, very useful.) I'm going back to lurking now as I am not qualified to argue here! |
May 2024
6:59pm, 22 May 2024
5,348 posts
|
run free
Interesting comments! Sounds like we talking about "humans". I know many people simply regurgitate the information they have been fed and propaganda shapes the way they think. As for asking the right questions....there are courses to help you ask AI the right questions!!!
|
|
May 2024
7:33pm, 22 May 2024
5,349 posts
|
run free
AI Chat bots are being integrated into business environments. Am sure a UN department will start up for the monitoring and safeguarding of humans.
|
Related Threads
-
Rise of the robots :-)
Aug 2024
-
Night sky aka Star gazing aka Astronomy :-)
Jan 2025
-
What's your view on someone who believes Earth is being or has been visited by UFOs/UAP's
Dec 2024
-
Interesting study
Jul 2024
-
Science
Feb 2024
-
Units of measurements
Feb 2022
-
Mars Landing
Feb 2021
-
No limit to the benefits of exercise in reducing the risk of cardiovascular disease
Jan 2021
-
No such thing as too much exercise
Oct 2018
-
Rocketeers!
Apr 2017