AI - Starmers 50 point plan to be released (13/1/25)

[Rob Ray]

"So, where should we put the first AI server hub Sir Keir? The logical places would be Scotland, the North or Wales, all of which could do with new industries, have abundant water and generate a lot of renewable energy."

"South east."

"But sir, the south east is water stressed and already much wealthier than the rest of the UK. Putting them in the south east would be bad politics and strategically silly."

Water shortage fears as Labour’s first AI growth zone sited close to new reservoir

First datacentre site proposed seven miles from Abingdon reservoir planned for water-stressed south-east England

www.theguardian.com

 

[Rob Ray]

Genuinely, what is wrong with these people?

They're inexpert, incurious, on a deadline and trapped in a bubble of 'news agenda' bullshit from the similarly afflicted which dictates what they think they should be focusing on.

 

[brogdale]

"So, where should we put the first AI server hub Sir Keir? The logical places would be Scotland, the North or Wales, all of which could do with new industries, have abundant water and generate a lot of renewable energy."

"South east."

"But sir, the south east is water stressed and already much wealthier than the rest of the UK. Putting them in the south east would be bad politics and strategically silly."

Water shortage fears as Labour’s first AI growth zone sited close to new reservoir

First datacentre site proposed seven miles from Abingdon reservoir planned for water-stressed south-east England

www.theguardian.com

Have to admit that this negative externality of the AI industry had not crossed my mind before; every day is a school day as they say.

 

[tommers]

They're inexpert, incurious, on a deadline and trapped in a bubble of 'news agenda' bullshit from the similarly addicted.

Sure but you'd hope that the people sent to ask questions about "AI" wouldn't still be under the impression that they're talking about Skynet or the computer from War Games.

 

[Artaxerxes]

Have to admit that this negative externality of the AI industry had not crossed my mind before; every day is a school day as they say.

It requires an inordinate amount of power and water to run a datacenter

 

[moochedit]

Genuinely, what is wrong with these people?

A lot of people hear "AI" and think it's actual "artificial intellegence" and then they think of the matrix, the terminator, westworld, battlestar galactica, bladerunner and other sci fi movies/tv where our own tech turns on us. Of course it's not "intellegence" it's just very clever programming but they don't understand that.

 

[Sue]

Sure but you'd hope that the people sent to ask questions about "AI" wouldn't still be under the impression that they're talking about Skynet or the computer from War Games.

Wait, what, we're not...? I'd have asked if there'd be flashy lights at least.

 

[maomao]

A lot of people hear "AI" and think it's actual "artificial intellegence" and then they think of the matrix, the terminator, westworld, battlestar galactica, bladerunner and other sci fi movies/tv where our own tech turns on us. Of course it's not "intellegence" it's just very clever programming but they don't understand that.

I don't see how AI is not genuinely intelligent, it's close enough to passing the Turing test. All AI, even The Cylons I suppose, are the end result of 'very clever programming' but when the machine is capable of learning independently and producing results that its programmers didn't predict or expect than there's another layer of creative decision making there that is a kind of intelligence. The main difference between modern AI and Skynet or Joshua is that ChatGPT is not in charge of a large amount of weaponry.

 

[PTK]

I don't see how AI is not genuinely intelligent, it's close enough to passing the Turing test. All AI, even The Cylons I suppose, are the end result of 'very clever programming' but when the machine is capable of learning independently and producing results that its programmers didn't predict or expect than there's another layer of creative decision making there that is a kind of intelligence. The main difference between modern AI and Skynet or Joshua is that ChatGPT is not in charge of a large amount of weaponry.

ChatGPT has no volition.
The AI that we use when we do searches on the web would not pass the Turing Test, nor would he AI that can identify images of tumours in scans. The SF-type thing is now called AGI, the G standing for "general".

 

[PTK]

I wish Starmer had avoided saying that if you are British you are more likely to develop new AI.

"This is the nation of Babbage, Lovelace and Turing
That gave birth to the modern computer and the World Wide Web
So mark my words – Britain will be one of the great AI superpowers."

 

[tommers]

I wish Starmer had avoided saying that if you are British you are more likely to develop new AI.

"This is the nation of Babbage, Lovelace and Turing
That gave birth to the modern computer and the World Wide Web
So mark my words – Britain will be one of the great AI superpowers."

Right and what did we do to Turing?

 

[Winot]

"So, where should we put the first AI server hub Sir Keir? The logical places would be Scotland, the North or Wales, all of which could do with new industries, have abundant water and generate a lot of renewable energy."

"South east."

"But sir, the south east is water stressed and already much wealthier than the rest of the UK. Putting them in the south east would be bad politics and strategically silly."

Water shortage fears as Labour’s first AI growth zone sited close to new reservoir

First datacentre site proposed seven miles from Abingdon reservoir planned for water-stressed south-east England

www.theguardian.com

Is that because they think the demand/staff for the centre is in the south east?

 

[Saul Goodman]

Is that because they think the demand/staff for the centre is in the south east?

It's probably because nothing further North is important to him.

 

[xenon]

I don't see how AI is not genuinely intelligent, it's close enough to passing the Turing test. All AI, even The Cylons I suppose, are the end result of 'very clever programming' but when the machine is capable of learning independently and producing results that its programmers didn't predict or expect than there's another layer of creative decision making there that is a kind of intelligence. The main difference between modern AI and Skynet or Joshua is that ChatGPT is not in charge of a large amount of weaponry.

See the talk about image generation. It doesn’t understand what an image is. It doesn’t have any understanding. If you say I want the same image but make the person have blonde hair or the sky brighter it doesn’t know what that means. And will just regenerate a new image from scratch.

I do think this touch is on the nature of intelligence. That we are intelligent because of million years evolution. Responses to risk and reward. There’s no context for these token generation machines. I’m just waffling a bit now to be fair. I don’t have a deep understanding of this stuff by any means.

 

[xenon]

But we don’t even understand human intelligence. To any significant degree. If AGI ever comes about it will be something quite alien. Also fed with our biases.

 

[Crispy]

I don't see how AI is not genuinely intelligent, it's close enough to passing the Turing test. All AI, even The Cylons I suppose, are the end result of 'very clever programming' but when the machine is capable of learning independently and producing results that its programmers didn't predict or expect than there's another layer of creative decision making there that is a kind of intelligence. The main difference between modern AI and Skynet or Joshua is that ChatGPT is not in charge of a large amount of weaponry.

LLMs are only active when composing a reply, and the network of neurons that replies flow through are rigidly frozen in place since training. They have no memory, no inner voice, no embodiment, no senses and no creativity. They very specifically are not capable of learning independently. They learn once, at tremendous effort (months and months of number crunching on gigantic computers) and thenceforth have a frozen "mind" that can only give different responses by feeding it different inputs.

(A lot of what makes variants of chatgpt behave differently is all the supporting text that gets fed in behind the scenes, as well as your prompt. All that scaffolding text can "persuade" it to lean its answer in various direcitons.)

 

[maomao]

LLMs are only active when composing a reply, and the network of neurons that replies flow through are rigidly frozen in place since training. They have no memory, no inner voice, no embodiment, no senses and no creativity. They very specifically are not capable of learning independently. They learn once, at tremendous effort (months and months of number crunching on gigantic computers) and thenceforth have a frozen "mind" that can only give different responses by feeding it different inputs.

(A lot of what makes variants of chatgpt behave differently is all the supporting text that gets fed in behind the scenes, as well as your prompt. All that scaffolding text can "persuade" it to lean its answer in various direcitons.)

If you ask the fucking things they tell you they appreciate your feedback and use it to improve. I misinterpreted that completely.

 

[Artaxerxes]

Is that because they think the demand/staff for the centre is in the south east?

Oxford has science types and techiesof course the entire fucking point if a a datacentre is it doesn't have to be where the work is carried out as it hosts remote servers. You could put the fucker on St Kilda with enough fibre in the sea

 

[Crispy]

If you ask the fucking things they tell you they appreciate your feedback and use it to improve. I misinterpreted that completely.

It's because alongside your prompt (and the complete history of the conversation so far) is a bunch of text saying things like "you are an advanced AI assistant, happy to help and happy to learn. You are polite, good humoured and will always take a passive attitude when threatened" and so on and on for hundreds of words. The LLM pulls all the meaning from those words and that influences the answer. Without this preparatory text, a "naked" LLM can be quite unhinged and very easy to goad into all sorts of lunacy.

 

[Artaxerxes]

It's because alongside your prompt (and the complete history of the conversation so far) is a bunch of text saying things like "you are an advanced AI assistant, happy to help and happy to learn. You are polite, good humoured and will always take a passive attitude when threatened" and so on and on for hundreds of words. The LLM pulls all the meaning from those words and that influences the answer. Without this preparatory text, a "naked" LLM can be quite unhinged and very easy to goad into all sorts of lunacy.

How unlike our forum members

 

[hitmouse]

If you ask the fucking things they tell you they appreciate your feedback and use it to improve. I misinterpreted that completely.

I tell my manager that I appreciate their feedback and will use it to improve, I'm not telling the fucking truth either!

 

[Sue]

I tell my manager that I appreciate their feedback and will use it to improve, I'm not telling the fucking truth either!

Oh hitmouse .

 

[brogdale]

They're a lobbyist's dream, this bunch...

 

[PTK]

This robot is about to be unleashed on roads in Hertfordshire. It identifies cracks in the road, and fills them before they become potholes.

AI robot to identify and fill in potholes in Hertfordshire

The technology, which identifies holes and then fills them in, will be tested across Hertfordshire.

www.bbc.co.uk

 

[Kevbad the Bad]

This robot is about to be unleashed on roads in Hertfordshire. It identifies cracks in the road, and fills them before they become potholes.

AI robot to identify and fill in potholes in Hertfordshire

The technology, which identifies holes and then fills them in, will be tested across Hertfordshire.

www.bbc.co.uk

What does it do about the millions of already existing potholes?

 

[ska invita]

Have to admit that this negative externality of the AI industry had not crossed my mind before; every day is a school day as they say.

The externality is the bit they seem most excited about. He gets to wear a hi viz

 

[bellaozzydog]

On a very basic level I’m using ChatGTP on a daily basis to “check tone” of all my emails.

It successfully makes me look like a civilised chap who gives a fuck, rather than the cynical no mark I actually am

 

[PTK]

On a very basic level I’m using ChatGTP on a daily basis to “check tone” of all my emails.

It successfully makes me look like a civilised chap who gives a fuck, rather than the cynical no mark I actually am

If you don't know if the tone of what you write is correct, then how do you know if the tone of what ChatGTP is correct?

 

[Elpenor]

Starmer’s enthusiasm for AI is matched by indications he is himself an early generation robot of some sort

Some evidence for this:

He doesn’t dream.
He doesn’t have a favourite book.
He repeats the same thing over and over in a dull voice “my mother was a nurse and my father was a toolmaker”, the last one being a clear and obvious clue as to his creator.

He is likely to be soon made obsolete by a more advanced model, which developers think may be called the Streeting.