Are we "Summoning the Demon" ?"
Discussion
I am not talking about certain posters in NP and E but talking about AI. I mentioned this a while ago but didnt really get any replies. Elon Musk used that phrase ten years ago, seemed kind of hyperbolic back then but computers will still dumb, glorified calculators to an extent and we didnt have robots doing parkour.
Not seen much on here about it but there are a lot of YouTube videos where some very clever folk are very worried about the next couple of years.
AI has come from not really existing other than a concept to being everywhere, though I dont think a lot of the public really use it, was talking to a friend and he no real idea what it was or what it does.
So now we have AI writing its replacement "Recursive Self-Improvement", kind of reminds me of that "Sorcerers Apprentice" section from Fantasia where Mickey Mouse uses a spell on a broom to get it to do the heavy lifting, he nods off and it floods the place, same thing with much more intelligent and numerous brooms ?
Loads of companies developing humanoid robots, Tesla has stopped Model S and X production to produce their Optimus Robot, they own the whole stack from the silicon to the training data, gathered from billions of miles of Tesla vehicles navigating the physical world. Boston Dynamics for example and plenty of other companies make a few impressive robots, but Tesla has the production capacity to build them in the millions, they dont buy anything in apart from raw materials.
We are at the Artificial Intelligence level. Next comes "Artificial General Intelligence", this is the point where an AI is no longer just good at a single task (like playing chess or writing an essay). Instead, it matches a human's ability to learn, reason, adapt, and perform any intellectual task that a human can do. It's essentially an AI with the general cognitive flexibility of a smart human adult.
Then comes (Artificial Superintelligence)
This is intelligence that is vastly superior to the collective brains of all humanity combined. An ASI wouldn't just be slightly smarter than us; it would be capable of solving scientific, mathematical, and philosophical problems that the human brain literally lacks the biological capacity to even comprehend.
So, are we f
ked is the question ?
At the moment, its not worrying, chatbots, can do a silly picture of your mate and am sick of stty AI voice overs on YouTube Videos, but I think we should be questioning things more, whats the endgame, is there an off ramp where AI is useful and genuinely helpful in medicine, administration and all manner of stuff we havent though of yet or is a global AI arms race playing out with eyes set firmly on the goal, not the consequences ?
How many jobs will be lost, will there be a Skynet moment where becomes self aware ?
Bit of La La La not listening going on !
Not seen much on here about it but there are a lot of YouTube videos where some very clever folk are very worried about the next couple of years.
AI has come from not really existing other than a concept to being everywhere, though I dont think a lot of the public really use it, was talking to a friend and he no real idea what it was or what it does.
So now we have AI writing its replacement "Recursive Self-Improvement", kind of reminds me of that "Sorcerers Apprentice" section from Fantasia where Mickey Mouse uses a spell on a broom to get it to do the heavy lifting, he nods off and it floods the place, same thing with much more intelligent and numerous brooms ?
Loads of companies developing humanoid robots, Tesla has stopped Model S and X production to produce their Optimus Robot, they own the whole stack from the silicon to the training data, gathered from billions of miles of Tesla vehicles navigating the physical world. Boston Dynamics for example and plenty of other companies make a few impressive robots, but Tesla has the production capacity to build them in the millions, they dont buy anything in apart from raw materials.
We are at the Artificial Intelligence level. Next comes "Artificial General Intelligence", this is the point where an AI is no longer just good at a single task (like playing chess or writing an essay). Instead, it matches a human's ability to learn, reason, adapt, and perform any intellectual task that a human can do. It's essentially an AI with the general cognitive flexibility of a smart human adult.
Then comes (Artificial Superintelligence)
This is intelligence that is vastly superior to the collective brains of all humanity combined. An ASI wouldn't just be slightly smarter than us; it would be capable of solving scientific, mathematical, and philosophical problems that the human brain literally lacks the biological capacity to even comprehend.
So, are we f
ked is the question ? At the moment, its not worrying, chatbots, can do a silly picture of your mate and am sick of s
tty AI voice overs on YouTube Videos, but I think we should be questioning things more, whats the endgame, is there an off ramp where AI is useful and genuinely helpful in medicine, administration and all manner of stuff we havent though of yet or is a global AI arms race playing out with eyes set firmly on the goal, not the consequences ?How many jobs will be lost, will there be a Skynet moment where becomes self aware ?
Bit of La La La not listening going on !
I'm not scared of AI, at the end of the day it's just technology, I'm scared of people misusing AI.
The entire AI industry knows it should be investing in AI ethics and guardrails, but unfortunately people who put profits ahead of security, safety or ethics are the ones with the steering wheel.
Some household name companies are going to go to the wall, either because they over-invest in AI too early, or because their businesses will fundamentally no longer be needed.
The entire AI industry knows it should be investing in AI ethics and guardrails, but unfortunately people who put profits ahead of security, safety or ethics are the ones with the steering wheel.
Some household name companies are going to go to the wall, either because they over-invest in AI too early, or because their businesses will fundamentally no longer be needed.
Security is probably the biggest concern with AI. Agentic Mythos is supposed to be a step change in vulnerability checking capability. It's good if you have it to protect yourself, not so good if someone else has it and wants to exploit your vulnerabilities.
Security patching used to be every few months, then every few weeks. We're now looking at patching within an 8 hour window. It's a big concern.
Create a safe and a safecracker will be along soon.
Security patching used to be every few months, then every few weeks. We're now looking at patching within an 8 hour window. It's a big concern.
Create a safe and a safecracker will be along soon.
Most so called AI isn't AI at all. So it depends upon your definition.
I might reply in much more detail later, but much of what people call AI is little more than a natural language interface to what amounts to a search engine.
The 'learning' is obviously based upon other people's work/thoughts etc.
AI isn't off doing its own physical experiments and actually 'learning' anything new.
It is however, very good at 'reading' many other sources and averaging an opinion.
Problems occur when it starts using generated sources as fact, and effectively enters a feedback loop.
A colleague was explaining how good 'AI' was at generating code... but to do that I would have to 100% define the process - as I do at the moment - the 'AI' can't work that out unless it is told. It can take some of the error prone grunt work out of it but it can't guess what you want.
The term is misleading, even an AI 'engine' is really algorithmic.
I might reply in much more detail later, but much of what people call AI is little more than a natural language interface to what amounts to a search engine.
The 'learning' is obviously based upon other people's work/thoughts etc.
AI isn't off doing its own physical experiments and actually 'learning' anything new.
It is however, very good at 'reading' many other sources and averaging an opinion.
Problems occur when it starts using generated sources as fact, and effectively enters a feedback loop.
A colleague was explaining how good 'AI' was at generating code... but to do that I would have to 100% define the process - as I do at the moment - the 'AI' can't work that out unless it is told. It can take some of the error prone grunt work out of it but it can't guess what you want.
The term is misleading, even an AI 'engine' is really algorithmic.
All this fearmongering is just the latest way to hype their products as the big boys are IPO'ing this year.
They're all bleeding huge sums of money whilst simultaneously passing large chunks around in a circular economy to try and make the overall numbers look palatable.
AGI is so far removed from what we currently consider AI as to be a pipe dream.
At some point the music is going to stop.
It's just doing ever more sophisticated pattern matching, there is no thinking, logic or first principles involved. it's the ultimate "crap in, crap out" machine.
They're all bleeding huge sums of money whilst simultaneously passing large chunks around in a circular economy to try and make the overall numbers look palatable.
AGI is so far removed from what we currently consider AI as to be a pipe dream.
At some point the music is going to stop.
No ideas for a name said:
Most so called AI isn't AI at all. So it depends upon your definition.
I might reply in much more detail later, but much of what people call AI is little more than a natural language interface to what amounts to a search engine.
The 'learning' is obviously based upon other people's work/thoughts etc.
AI isn't off doing its own physical experiments and actually 'learning' anything new.
It is however, very good at 'reading' many other sources and averaging an opinion.
Problems occur when it starts using generated sources as fact, and effectively enters a feedback loop.
A colleague was explaining how good 'AI' was at generating code... but to do that I would have to 100% define the process - as I do at the moment - the 'AI' can't work that out unless it is told. It can take some of the error prone grunt work out of it but it can't guess what you want.
The term is misleading, even an AI 'engine' is really algorithmic.
I might reply in much more detail later, but much of what people call AI is little more than a natural language interface to what amounts to a search engine.
The 'learning' is obviously based upon other people's work/thoughts etc.
AI isn't off doing its own physical experiments and actually 'learning' anything new.
It is however, very good at 'reading' many other sources and averaging an opinion.
Problems occur when it starts using generated sources as fact, and effectively enters a feedback loop.
A colleague was explaining how good 'AI' was at generating code... but to do that I would have to 100% define the process - as I do at the moment - the 'AI' can't work that out unless it is told. It can take some of the error prone grunt work out of it but it can't guess what you want.
The term is misleading, even an AI 'engine' is really algorithmic.
It's just doing ever more sophisticated pattern matching, there is no thinking, logic or first principles involved. it's the ultimate "crap in, crap out" machine.
Edited by SWoll on Tuesday 30th June 15:04
The problem with terms like AGI and ASI is that everyone seems to define them slightly differently. By some standards, AGI is already here. By others, we’re nowhere near it.
We also haven’t even properly worked out what consciousness is, so I think we’re a fair way from proving whether an AGI or ASI has it.
What current generative AI can already do, though, is assist with or replace a lot of work. Repetitive tasks are the obvious ones, but anything based on a clear process is vulnerable too. That could affect a lot of jobs.
I also think the impact will be uneven. AI is more likely to replace parts of jobs before it replaces whole roles, and businesses still have to deal with trust, accuracy, regulation and integration before they can hand over anything important. So it probably won’t be a clean “AI replaces everyone” moment. It’ll be messier than that, with some tasks disappearing, some roles shrinking, and some people becoming much more productive because they know how to use the tools properly.
Realistically, I suspect it’ll be like the web. Industries won’t vanish overnight, but they will change. Some roles will disappear, some will evolve, and new ones will turn up.
As a final thought, if you’re planning to keep working for the next decade, it’s probably worth learning what AI can and can’t do, and finding ways to use it to improve your own productivity and output. Ignoring it won’t stop it happening, and the people who get comfortable using these tools are likely to be in a much better position than those who don’t.
We also haven’t even properly worked out what consciousness is, so I think we’re a fair way from proving whether an AGI or ASI has it.
What current generative AI can already do, though, is assist with or replace a lot of work. Repetitive tasks are the obvious ones, but anything based on a clear process is vulnerable too. That could affect a lot of jobs.
I also think the impact will be uneven. AI is more likely to replace parts of jobs before it replaces whole roles, and businesses still have to deal with trust, accuracy, regulation and integration before they can hand over anything important. So it probably won’t be a clean “AI replaces everyone” moment. It’ll be messier than that, with some tasks disappearing, some roles shrinking, and some people becoming much more productive because they know how to use the tools properly.
Realistically, I suspect it’ll be like the web. Industries won’t vanish overnight, but they will change. Some roles will disappear, some will evolve, and new ones will turn up.
As a final thought, if you’re planning to keep working for the next decade, it’s probably worth learning what AI can and can’t do, and finding ways to use it to improve your own productivity and output. Ignoring it won’t stop it happening, and the people who get comfortable using these tools are likely to be in a much better position than those who don’t.
SWoll said:
It's just doing ever more sophisticated pattern matching, there is no thinking, logic or first principles involved. it's the ultimate "crap in, crap out" machine.
Whenever I hear someone make this argument - it's just fancy autocomplete, etc - I am minded to ask how they would know that isn't what they are.SWoll said:
It's just doing ever more sophisticated pattern matching, there is no thinking, logic or first principles involved. it's the ultimate "crap in, crap out" machine.
I think that argument misses the point a bit. What we’re seeing is emergent behaviour: complex systems producing results that are more than the obvious sum of their parts.So yes, you can describe generative AI as prediction and pattern matching at a low level, but that doesn’t mean the overall behaviour is trivial. Once these systems reach enough scale and complexity, unexpected capabilities appear. That doesn’t prove consciousness or human-style thinking, but “it’s just autocomplete” feels like an oversimplification.
As SWoll says, a lot of it is hype from the companies involved in order to push their product and to maintain the insane levels of funding. A lot of the doom-mongers will be in the pockets of the AI companies and paid to say scary stuff.
I know it has its uses but all this ‘AI will take your job in 12 months’ or ‘only five careers will remain in 24 months’ is utter fking nonsense.
All I’ve found it good for is giving me another way to bully a friend of mine by getting ChatGPT to make photos of him in women’s clothing and giving him Freddy Mercury moustaches.
I know it has its uses but all this ‘AI will take your job in 12 months’ or ‘only five careers will remain in 24 months’ is utter f
king nonsense. All I’ve found it good for is giving me another way to bully a friend of mine by getting ChatGPT to make photos of him in women’s clothing and giving him Freddy Mercury moustaches.
I dont know, I dont buy into the absolute worst case scenario doom mongering but I do think we are sleep walking into stuff, relentless progress because we can, but the "should" bit perhaps doesnt get much attention as you would expect, the guardrails mentioned.
The Chinese are investing heavily in AI and robotics, they arent known for their benevolence, neither are a lot of other countries, drones are getting freakishly impressive, and thats just the stuff we know about, what the hell are they working on that we arent aware of ? batteries are improving with Lithium Air waiting in the wings with much higher run times promised.
Its like all the tech is converging, think thats described as the "AI Event Horizon" when we eventually go hands off and are sure what its doing, remember the bit in Close Encounters with the lights and music notes.....
The Chinese are investing heavily in AI and robotics, they arent known for their benevolence, neither are a lot of other countries, drones are getting freakishly impressive, and thats just the stuff we know about, what the hell are they working on that we arent aware of ? batteries are improving with Lithium Air waiting in the wings with much higher run times promised.
Its like all the tech is converging, think thats described as the "AI Event Horizon" when we eventually go hands off and are sure what its doing, remember the bit in Close Encounters with the lights and music notes.....
J4CKO said:
We are at the Artificial Intelligence level. Next comes "Artificial General Intelligence", this is the point where an AI is no longer just good at a single task (like playing chess or writing an essay). Instead, it matches a human's ability to learn, reason, adapt, and perform any intellectual task that a human can do. It's essentially an AI with the general cognitive flexibility of a smart human adult.
Then comes (Artificial Superintelligence)
This is intelligence that is vastly superior to the collective brains of all humanity combined. An ASI wouldn't just be slightly smarter than us; it would be capable of solving scientific, mathematical, and philosophical problems that the human brain literally lacks the biological capacity to even comprehend.
So, are we fked is the question ?
They've been promising artificial general intelligence for about 70 years, the problem, is that "general intelligence" is actually very hard to recreate without human level "thinking" and there are theories out there that suggest that the ability to think "generally" would necessitate thinking in a human way (with all of the flaws and biases that would encompass). That in turn would then somewhat limit the ability for a machine to surpass our level of intelligence (it would end up tripping over the same things that we do). Then comes (Artificial Superintelligence)
This is intelligence that is vastly superior to the collective brains of all humanity combined. An ASI wouldn't just be slightly smarter than us; it would be capable of solving scientific, mathematical, and philosophical problems that the human brain literally lacks the biological capacity to even comprehend.
So, are we f
ked is the question ? The recurring theme with AI (since the 1950s) is that doing easy stuff is surprisingly hard.
There's a great quote somewhere that we shouldn't be worried about super intelligent AI taking our jobs, we should be worried that we've already given so many of our jobs to stupid AI.
I work within the industry (and consult on AI adoption amongst other things) but if you are interested you should have a read of "Artificial Intelligence: A Guide for thinking humans." A lot of the technical stuff is no a bit out of date but the recurring theme remains.
There are levels to AI.
Ultimately money wont mean anything any more - it would likely become a resource battle. The AI would seek to control as much natural resource as possible. Is that going to be done via peaceful means? I doubt it.
But the time between then and now is probably into the tens of years. Most of us will be dead already.
Hopefully some of us will be around to see the huge steps forward AI can bring - infrastructure being built in months rather than years, massive leaps forward in medical science, crime reducing significantly, compute power growing beyond comprehension and humans generally having a better day to day life.
The big problem is at each junction we will have a question? Move forward with an AI which is good for everyone and is peaceful or move forward with an AI which is good only for its owners and protects their interests with force. Its a question that will get asked over and over... eventually someone will crack and go for option 2.
Ultimately money wont mean anything any more - it would likely become a resource battle. The AI would seek to control as much natural resource as possible. Is that going to be done via peaceful means? I doubt it.
But the time between then and now is probably into the tens of years. Most of us will be dead already.
Hopefully some of us will be around to see the huge steps forward AI can bring - infrastructure being built in months rather than years, massive leaps forward in medical science, crime reducing significantly, compute power growing beyond comprehension and humans generally having a better day to day life.
The big problem is at each junction we will have a question? Move forward with an AI which is good for everyone and is peaceful or move forward with an AI which is good only for its owners and protects their interests with force. Its a question that will get asked over and over... eventually someone will crack and go for option 2.
I must admit it's getting quite hard at times to work out what is actually real or not and the opportunity for the dissemination of false narratives is immense
On another level though very little regard seems to be being taken on the huge amount of energy and water that these data centres consume and the impact that will have on us
On another level though very little regard seems to be being taken on the huge amount of energy and water that these data centres consume and the impact that will have on us
Earthdweller said:
On another level though very little regard seems to be being taken on the huge amount of energy and water that these data centres consume and the impact that will have on us
Initially the theory was that AI capability would expand at an exponential rate but I think what we are now seeing is the costs expanding at an exponential rate in order to bring about marginal gains (the law of diminishing returns). From an ROI perspective, I think that is going to be the real challenge moving forwards (ever increasing costs for ever decreasing gains leads to lower and lower ROIs)
...Just my theory though and I'm usually wrong
Gassing Station | News, Politics & Economics | Top of Page | What's New | My Stuff