• Security incident: ISF was recently accessed by intruders. Please change your password, and change it anywhere else you used it. Read more

Merged Artificial Intelligence

dirtywick said:
These companies, including openai, are all running out of private investment and want to go public.
Click to expand...

The size of the bubble being created just keeps going.

There is clearly not enough income to support the current valuation of AI companies and those who supply them and like all previous bubbles, will burst.

There's probably still a couple of years to go before appetite for illusory future returns bites back.
 
in short: they have to raise prices, causing customers to leave, which necessitates raising prices, causing customers to leave...

It's Scaling in the other direction.
 
arthwollipot said:
If only more people would take that second step.


If only more people knew about this.
Click to expand...
The first bit won't work. It's become clear that LLMs don't retain anything from previous conversations (unless you personally paste in the previous excange yourself) and will continue to make the exact same mistakes over and over again.

They do not learn. They cannot learn.
 
Darat said:
Yes and no. They are brilliant at what they were initially created to do, which is to mimic human behaviour. Problem is that human behaviour is often not "intelligent", humans make mistakes, human memory is very fallible, humans make stuff up, humans sound confident even when they haven't a clue, humans try to trick one another, humans bluster when being pressed. All of which they are excellent at, they have shown that many of the criterias and behaviours we used to think required "intelligence" in humans do not need intelligence to mimic. (They definitely do not achieve their results in the same way humans do.) I've been saying for a long time that we don't want human intelligence from our artificial intelligence systems. We want emotionless, reality based AI, not AIs that merely mimic what we have in the past assumed were indicators of intelligence.
Click to expand...
They may have a larger language library* but at heart they're still Eliza.

Computer scientists have been writing bots that can fool people for decades now.

*mostly stolen.
 
Gulliver Foyle said:
The first bit won't work. It's become clear that LLMs don't retain anything from previous conversations (unless you personally paste in the previous excange yourself) and will continue to make the exact same mistakes over and over again.

They do not learn. They cannot learn.
Click to expand...
Of course they can learn. The don't on purpose. You can't have the model be changed by user input. Even if you offered it as a service, the memory requirements would be huge.

These days most public LLMs have text based long term memory, basically a text notepad, which the LLM can decide to store things to and then look-up (and you can ask it to do it). It's not really a "learning", but at least it can remember some things without repeating them all the time.

For specialised use it's common to train a local model on your knowledge base, or even to have databases of preprocessed binary modules trained on different stuff, which you can activate on demand. You can have 1 module for company policy, and then 1 module for every costumer documentation, without need to process it all as text again and again, as context length is always an issue.

There is however still huge ability gap in learning. AI can learn, it can learn very complex stuff .. but not very fast. It needs tens of thousands of repetitions typically. It seems biological brains do it lot better. There might be some mechanism to reinforce information by internal repetition even with biological brains .. but they are certainly several orders of magnitude better at this. Even if our artificial neural networks were very directly inspired by the real ones, and even if they work, and we can mathematically prove they can adapt any function .. the still clearly don't work anything like biological brain.
On the other hand they don't have to work as biological brains to be usefull.
 
I think AI has some good potential... But I remain highly skeptical at the moment. This was just reinforced by a consulting firm trying to sell my company some AI capabilities. They used a presentation that was clearly developed by AI and included several misspelled words, as well as having numbered a list
1
2
3
5
4
5
6
8
9
7
11
 
a company claims to have solved a problem that would make ai a lot cheaper to run but nobody is really sure if they're telling the truth or not
 
Okay, turns out I care a little more than that.

ETA: But not enough to struggle through the multiple popups on that website. What the hell?
 
Tech stocks tumble on concerns over AI spending

Financial markets received a sharp wake-up call on Tuesday following a sudden wave of selling in major technology shares, triggering widespread doubt over the sustainability of the AI boom.

The tech-focused Nasdaq index fell about 2% alongside international chipmakers, reigniting fears that dizzying market valuations have finally run out of momentum after a relentless three-month climb.

www.bbc.co.uk

Tech stocks tumble on concerns over AI spending

A sudden wave of selling in major tech shares triggers doubt over the sustainability of the AI boom.
www.bbc.co.uk www.bbc.co.uk
 
You must log in or register to reply here.

Share this page

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom