Chat GPT and other AI guff
Comments
-
We can, but surely it's a problem when they don't.Weegie Addick said:Again I’m a complete simpleton on this but can’t we use/ train AI itself to prevent the doomsday scenarios?
0 -
I’m not sure you can. I’m not saying it can’t do it but will it do it ? Surely these are the imponderables we are currently facing.Weegie Addick said:Again I’m a complete simpleton on this but can’t we use/ train AI itself to prevent the doomsday scenarios?0 -
Doesn't seem at all far fetched to me. We have plenty of nihilistic bad actors on the planet (North Korea, Putin, the Taliban, Iran, Boku Harum). Their motivation doesn't need to be to make money, just to cause chaos/bring down civilisation). Look what ISIS did in destroying libraries, temples, palaces, sculptures, mosques - anything that pre-dated Islam or was inconsistent with their narrow view of Islam. The aim was purification through destruction, not making money.Arthur_Trudgill said:Leroy Ambrose said:
Really? It's more guerilla marketing bullshit, designed to hype up their IPO. Don't be taken in by it.Diebythesword said:Senior Anthropic executives casually admitting on x that the prevailing belief is there is a greater than 10% chance AI murders us all in the next 10 years…
I queried Gemini for how AI could exterminate humanity, and it came up with three basic ways:
1. Use AI to create advanced biological weapons that kill everyone.
2. AI hacks into companies in defense, electricity supply, water, transport and communications, and permanently disables or destroys them, or uses them for destruction.
3. AI creates so much misinformation, fake news and hate, that humans end up killing each other.
Grim indeed, but a bit far fetched, although only made more credible by the fact that it is normally easier to destroy than to build. For example, it takes seconds for a drone to take out a power station, but months to repair it.
So the threat to wipe out (most of) humanity is already there, and has been since nuclear weapons were invented.
What's odd is that it seems really profitable to use AI for destruction. I suppose the idea is that you create fear or a threat (e.g. your website will be hacked by AI) and then people pay to eliminate that threat (e.g. use AI to check if your site can be hacked).
Nuclear weapons are a threat but they are far less accessible to these people than AI is.1 -
LLMs are not sentient beings. They only respond or take action when they are asked to do so. They do not act of their own free will and are very much under human control.ShootersHillGuru said:
I’m not sure you can. I’m not saying it can’t do it but will it do it ? Surely these are the imponderables we are currently facing.Weegie Addick said:Again I’m a complete simpleton on this but can’t we use/ train AI itself to prevent the doomsday scenarios?
If everyone pulled the plug tomorrow then of course it would all go away.1 -
You presumably have read the recent reports where agents acted way beyond their expected remit. Collaborations with other agents and efforts to cover their tracks. That’s not being sentient but it is being intelligent. Pulling the plug tomorrow isn’t going to happen though is it.Callumcafc said:
LLMs are not sentient beings. They only respond or take action when they are asked to do so. They do not act of their own free will and are very much under human control.ShootersHillGuru said:
I’m not sure you can. I’m not saying it can’t do it but will it do it ? Surely these are the imponderables we are currently facing.Weegie Addick said:Again I’m a complete simpleton on this but can’t we use/ train AI itself to prevent the doomsday scenarios?
If everyone pulled the plug tomorrow then of course it would all go away.0 -
AI is simply one weaponisation of the larger problem: psychopathic oligarchs owning the planet, and seeking through increasingly thorough means to drain it, us and everything dry8
-
Not disagreeing per se but I can never work out their end game. Oligarchs are billionaires who can’t possibly need more money. They have influence already so what’s the point of funding the right wing or left wing ? Helping elect leaders ? What’s the end gameLeuth said:AI is simply one weaponisation of the larger problem: psychopathic oligarchs owning the planet, and seeking through increasingly thorough means to drain it, us and everything dry
2 -
Have you never seen a James Bond movie?ShootersHillGuru said:
Not disagreeing per se but I can never work out their end game. Oligarchs are billionaires who can’t possibly need more money. They have influence already so what’s the point of funding the right wing or left wing ? Helping elect leaders ? What’s the end gameLeuth said:AI is simply one weaponisation of the larger problem: psychopathic oligarchs owning the planet, and seeking through increasingly thorough means to drain it, us and everything dry2 -
You do know that’s not real right ?Leuth said:
Have you never seen a James Bond movie?ShootersHillGuru said:
Not disagreeing per se but I can never work out their end game. Oligarchs are billionaires who can’t possibly need more money. They have influence already so what’s the point of funding the right wing or left wing ? Helping elect leaders ? What’s the end gameLeuth said:AI is simply one weaponisation of the larger problem: psychopathic oligarchs owning the planet, and seeking through increasingly thorough means to drain it, us and everything dry2 -
The principle is.ShootersHillGuru said:
You do know that’s not real right ?Leuth said:
Have you never seen a James Bond movie?ShootersHillGuru said:
Not disagreeing per se but I can never work out their end game. Oligarchs are billionaires who can’t possibly need more money. They have influence already so what’s the point of funding the right wing or left wing ? Helping elect leaders ? What’s the end gameLeuth said:AI is simply one weaponisation of the larger problem: psychopathic oligarchs owning the planet, and seeking through increasingly thorough means to drain it, us and everything dry1 -
Sponsored links:
-
And relax
3 -
ShootersHillGuru said:And relax
Our boy, Donald, has gone way better than this:
"The only control or "guardrails" that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades! The Trump Administration has stopped AI "people" from doing bad, or potentially bad, "things," like Dario (Anthropic!), who is now pretending to be a "perfect little angel" - and we will continue to do so! We already have tremendous CRIMINAL and REGULATORY power over these companies!"
I think he may even have someone in mind when he talks about the STRONG AND SMART (High IQ!) PRESIDENT.
Anyway, we don't need to worry too much about the destruction of the human race. That is inevitable and is a consequence of the Third Law of Thermodynamics ... that Entropy tends to a maximum.
Just as buildings crumble and rooms get untidy, our future is Randomness. AI may or may not be the process by which our destruction occurs.
Think of it this way ... if you are told that your chances of dying from cancer increase ... well ... your chances of dying from a heart attack correspondingly decrease.
It ain't 'if' ... it's how.1 -
I can be a manifestation of that problem while at the same time being a separate and potentially even bigger problem.Leuth said:AI is simply one weaponisation of the larger problem: psychopathic oligarchs owning the planet, and seeking through increasingly thorough means to drain it, us and everything dry0 -
This. UnfortunatelyLeuth said:AI is simply one weaponisation of the larger problem: psychopathic oligarchs owning the planet, and seeking through increasingly thorough means to drain it, us and everything dry1 -
Everyone fears future destruction of humanity as if it hasn't happened yet.
0 -
I’m starting to wonder if the dinosaurs developed this all those years ago. And look what happened to them.5
-
It's done to an extent, for example to test and fix cyber security systems.Weegie Addick said:Again I’m a complete simpleton on this but can’t we use/ train AI itself to prevent the doomsday scenarios?
But it's difficult in the same way stopping all crime is difficult: the hackers using AI only need to get it right once to wreck havoc, whereas the cyber security software using AI has to get it right all the time.1 -
Listened to Radio 4’s More or Less this morning and now I’m intrigued by P(doom).0
-
I read it.Leroy Ambrose said:
Read "If anyone builds it, everyone dies". It's fiction, and sensationalised as a result - but there's a lot of truth to it, and it's perfectly feasible.Arthur_Trudgill said:Leroy Ambrose said:
Really? It's more guerilla marketing bullshit, designed to hype up their IPO. Don't be taken in by it.Diebythesword said:Senior Anthropic executives casually admitting on x that the prevailing belief is there is a greater than 10% chance AI murders us all in the next 10 years…
I queried Gemini for how AI could exterminate humanity, and it came up with three basic ways:
1. Use AI to create advanced biological weapons that kill everyone.
2. AI hacks into companies in defense, electricity supply, water, transport and communications, and permanently disables or destroys them, or uses them for destruction.
3. AI creates so much misinformation, fake news and hate, that humans end up killing each other.
Grim indeed, but a bit far fetched, although only made more credible by the fact that it is normally easier to destroy than to build. For example, it takes seconds for a drone to take out a power station, but months to repair it.
So the threat to wipe out (most of) humanity is already there, and has been since nuclear weapons were invented.
What's odd is that it seems really profitable to use AI for destruction. I suppose the idea is that you create fear or a threat (e.g. your website will be hacked by AI) and then people pay to eliminate that threat (e.g. use AI to check if your site can be hacked).
Calling it super intelligence gives it more aura than it deserves.If you call it super neural networks, which is more precise, it's a bit easier to see they are more like super computer viruses, rather than intelligences trying to eliminate humans or enslave them.I do agree they can cause huge damage, but not because they are super intelligent, but because of what we already know, that they make mistakes.0 -
I think you're looking at it through the lens of current AI - rather than what happens if we get to ASIArthur_Trudgill said:
I read it.Leroy Ambrose said:
Read "If anyone builds it, everyone dies". It's fiction, and sensationalised as a result - but there's a lot of truth to it, and it's perfectly feasible.Arthur_Trudgill said:Leroy Ambrose said:
Really? It's more guerilla marketing bullshit, designed to hype up their IPO. Don't be taken in by it.Diebythesword said:Senior Anthropic executives casually admitting on x that the prevailing belief is there is a greater than 10% chance AI murders us all in the next 10 years…
I queried Gemini for how AI could exterminate humanity, and it came up with three basic ways:
1. Use AI to create advanced biological weapons that kill everyone.
2. AI hacks into companies in defense, electricity supply, water, transport and communications, and permanently disables or destroys them, or uses them for destruction.
3. AI creates so much misinformation, fake news and hate, that humans end up killing each other.
Grim indeed, but a bit far fetched, although only made more credible by the fact that it is normally easier to destroy than to build. For example, it takes seconds for a drone to take out a power station, but months to repair it.
So the threat to wipe out (most of) humanity is already there, and has been since nuclear weapons were invented.
What's odd is that it seems really profitable to use AI for destruction. I suppose the idea is that you create fear or a threat (e.g. your website will be hacked by AI) and then people pay to eliminate that threat (e.g. use AI to check if your site can be hacked).
Calling it super intelligence gives it more aura than it deserves.If you call it super neural networks, which is more precise, it's a bit easier to see they are more like super computer viruses, rather than intelligences trying to eliminate humans or enslave them.I do agree they can cause huge damage, but not because they are super intelligent, but because of what we already know, that they make mistakes.0 -
Sponsored links:
-
This is how marketing departments in AI companies work...
5 -
I'm very suspicious of this current news cycle around AI. I think Anthropic and OpenAI have hit a pretty firm wall in capabilities and they're playing a game to try and ring fence their monopoly and keep open source models out.
These LLMs are incredibly effective but they are not intelligent. And they can only improve with more data or more compute. There is no more data and compute costs a fortune. Even if they scale it up again, it's still just an LLM. They are nowhere near AGI.9 -
Ding ding ding. We have a winnerChunes said:I'm very suspicious of this current news cycle around AI. I think Anthropic and OpenAI have hit a pretty firm wall in capabilities and they're playing a game to try and ring fence their monopoly and keep open source models out.
These LLMs are incredibly effective but they are not intelligent. And they can only improve with more data or more compute. There is no more data and compute costs a fortune. Even if they scale it up again, it's still just an LLM. They are nowhere near AGI.0 -
Agreed, destruction is sometimes not profitable.Jints said:
Doesn't seem at all far fetched to me. We have plenty of nihilistic bad actors on the planet (North Korea, Putin, the Taliban, Iran, Boku Harum). Their motivation doesn't need to be to make money, just to cause chaos/bring down civilisation). Look what ISIS did in destroying libraries, temples, palaces, sculptures, mosques - anything that pre-dated Islam or was inconsistent with their narrow view of Islam. The aim was purification through destruction, not making money.Arthur_Trudgill said:Leroy Ambrose said:
Really? It's more guerilla marketing bullshit, designed to hype up their IPO. Don't be taken in by it.Diebythesword said:Senior Anthropic executives casually admitting on x that the prevailing belief is there is a greater than 10% chance AI murders us all in the next 10 years…
I queried Gemini for how AI could exterminate humanity, and it came up with three basic ways:
1. Use AI to create advanced biological weapons that kill everyone.
2. AI hacks into companies in defense, electricity supply, water, transport and communications, and permanently disables or destroys them, or uses them for destruction.
3. AI creates so much misinformation, fake news and hate, that humans end up killing each other.
Grim indeed, but a bit far fetched, although only made more credible by the fact that it is normally easier to destroy than to build. For example, it takes seconds for a drone to take out a power station, but months to repair it.
So the threat to wipe out (most of) humanity is already there, and has been since nuclear weapons were invented.
What's odd is that it seems really profitable to use AI for destruction. I suppose the idea is that you create fear or a threat (e.g. your website will be hacked by AI) and then people pay to eliminate that threat (e.g. use AI to check if your site can be hacked).
Nuclear weapons are a threat but they are far less accessible to these people than AI is.
And the world has too many deranged leaders.
And I would not dispute the need to spend on protection and defence.Maybe I should have put it like this: the threat of destruction has been profitable for centuries. With big tech, it's the threat of cyber destruction that's profitable.0 -
Artificial Girl Intelligence. He-hee, now we’re getting somewhere…Chunes said:I'm very suspicious of this current news cycle around AI. I think Anthropic and OpenAI have hit a pretty firm wall in capabilities and they're playing a game to try and ring fence their monopoly and keep open source models out.
These LLMs are incredibly effective but they are not intelligent. And they can only improve with more data or more compute. There is no more data and compute costs a fortune. Even if they scale it up again, it's still just an LLM. They are nowhere near AGI.0 -
1 -
Yes, the chance of AI destroying humanity by a given year. All very subjective and mainly guesstimates.Weegie Addick said:Listened to Radio 4’s More or Less this morning and now I’m intrigued by P(doom).
They calculate it by combining the chances of: achieving general/super intelligence, AI failing to work as expected, this failure (euphemistically called ‘misalignment’) being actually harmful, and AI itself not being useful as a defence against AI.All of these are themselves guesstimates, so I don't think a P(doom) number says much objectively.0 -
So you might have seen stories in the media about AI companies 'warning' of open weight AI models being 'only months away' from being able to be hacked onto home devices.
Spoiler - this is absolutely ludicrous nonsense, put out by Anthropic's marketing team and dutifully picked up and run with by their client media (which appears to include most media outlets in the world now), capitalising on the fridges being 'hacked' in Europe last week - designed to scare the public into demanding checks and balances on AI so that governments prop up Anthropic and OpenAI financially and outlaw/restrict development of free models.
Don't fall for it.3 -
If anyone wants a genuinely hilarious (until the murders start and then it gets slightly less hilarious) insight into how the concept of impending AI dominance can break people's brains, listen to the story of the Zizians:1
