Dude, it's all about AI!

[sfrydman]

Don't like Trump? How about an immortal Trump who is dictator for eternity? With AI, the potential upsides are limitless, but so are the downsides.

Geoffrey Hinton warning:
Paul Tudor Jones warning:
List of things to do: https://existentialsafety.org/

The biggest problem - researchers don't know how to make AI safe. They're black boxes that the people who grew them (that's right, grew, not programmed) don't understand. Top AI researchers are saying there's at least a 20% chance or more of AI wiping out humanity, whether on its own or due to a bad actor leading it to bad outcomes. But the U.S. wants to beat China and AI companies want to beat each other. It's bad game theory and there's a significant chance it could end in disaster. The only sensible thing is to have an international treaty with teeth to regulate and slow down the race to artificial super intelligence.

If any of this is making sense to you, please do whatever you feel is in your power to help. I've been posting videos on social media every day. I urge everyone to do the same. On LinkedIn, I contact anyone in artificial intelligence, machine learning, and national security (those people have more of a chance of understanding the threat than lawmakers) and try to nudge them in the right direction. You might also consider joining PauseAI.

Thanks for reading

 

[Greg Bernhardt]

The biggest problem - researchers don't know how to make AI safe

There is way too much pressure for AI profit and we're cutting back even loose regulations. It's a recipe for long term disaster.

 

[Borg]

I don't see this in the same way. Yes, there are concerns on the horizon but I do not see AI wiping out humanity. Assuming it reaches AGI and assuming that it has free reign across the internet, wiping out humanity would result in its own destruction as well. Where is its motivation to do that?

 

[gleem]

I don't see this in the same way. Yes, there are concerns on the horizon but I do not see AI wiping out humanity. Assuming it reaches AGI and assuming that it has free reign across the internet, wiping out humanity would result in its own destruction as well. Where is its motivation to do that?

So far, the best intelligence, Human, is on its way to destroying itself, so why can't AI accelerate the process?

 

[Borg]

So far, the best intelligence, Human, is on its way to destroying itself, so why can't AI accelerate the process?

Then that would still be the fault of humans.

 

[Greg Bernhardt]

Assuming it reaches AGI and assuming that it has free reign across the internet, wiping out humanity would result in its own destruction as well.

AGI alone is not enough, it would need to control manufacturing for robots etc.

 

[sfrydman]

I'm passing along this message from PauseAI

1. • IMPORTANT ANNOUNCEMENT: Call your Senators ASAP and save AI policy!​

     We need your help to stop a dangerous AI bill from passing the Senate.What’s going on?
     • The House Energy & Commerce Committee included a provision in its reconciliation bill that would ban AI regulation by state and local governments for the next 10 years.,
     • Several states have led the way in AI regulation while Congress has dragged its heels.,
     • Stopping state governments from regulating AI might be okay, if we could trust Congress to meaningfully regulate it instead. But we can’t. This provision would destroy state leadership on AI and pass the responsibility to a Congress that has shown little interest in seriously preventing AI danger.,
     • If this provision passes the Senate, we could see a DECADE of inaction on AI.,
     • This provision also violates the Byrd Rule, a Senate rule which is meant to prevent non-budget items from being included in the reconciliation bill.,
     (edited)
     
     
   • Find your Senators here.,
   • Here’s an example of a call:

“Hello, my name is {YOUR NAME} and I’m a resident of {YOUR STATE}. The newest budget reconciliation bill includes a 10-year ban pre-empting state AI legislation without establishing any federal guardrails. This is extremely concerning to me – leading experts warn us that AI could cause mass harm within the next few years, but this provision would prevent states from protecting their citizens from AI crises for the next decade. It also violates the Byrd Rule, since preempting state AI regulation doesn’t impact federal taxes or spending.

I’d like the Senator to speak out against this provision and raise a point of order that this provision should not be included under the Byrd Rule.”

See here for sample call + email templates.,

 

[Greg Bernhardt]

Google’s AI Boss Says Gemini’s New Abilities Point the Way to AGI​

Google’s AI Boss Says Gemini's New Abilities Point the Way to AGI

Google’s AI models are learning to reason, wield agency, and build virtual models of the real world. The company’s AI lead, Demis Hassabis, says all this—and more—will be needed for true AGI.

www.wired.com

 

[dlu]

Google’s AI Boss Says Gemini’s New Abilities Point the Way to AGI​

Google’s AI Boss Says Gemini's New Abilities Point the Way to AGI

Google’s AI models are learning to reason, wield agency, and build virtual models of the real world. The company’s AI lead, Demis Hassabis, says all this—and more—will be needed for true AGI.

www.wired.com

Should be worth another few billions from investors crazy on AI.
AGI just around the corner they say. Hype all the way to the bank.

 

[sfrydman]

AGI just around the corner they say. Hype all the way to the bank.

I disagree. Frankly, I don't understand how anyone is saying all this is "just hype". We've passed the Turing test, which used to be a crazy hypothetical. You can have a conversation with your computer - text or spoken, take your pick - on any topic in any language and get a much better response than what you would expect from most humans. AlphaEvolve is making breakthroughs in math and science that the best human minds could not make, actually creating new knowledge rather than regurgitating it. Robots are cleaning messy rooms and breakdancing. The space of things humans can do that machines can't is getting smaller and smaller. It's getting to the point where people are about to say, "Well, I'm still better at whistling so no need to worry." Can you name a specific task that you don't believe machines will be able to perform?

 

[gleem]

Recent developments from MIT, Goethe University, and IBM are allowing AI to take images and audio to learn like humans, associating the sound with the image.

MIT researcher Andrew Rouditchenko says

“We are building AI systems that can process the world like humans do, in terms of having both audio and visual information coming in at once and being able to seamlessly process both modalities. Looking forward, if we can integrate this audio-visual technology into some of the tools we use on a daily basis, like large language models, it could open up a lot of new applications,”

 

[dlu]

The space of things humans can do that machines can't is getting smaller and smalle

The hype on AI is for the investors to drop cash. The hype is 'this is what our new product can do', with minimization of product shortcomings. So what if the LLMs if it hallucinates and gives wrong answers and makes stuff up.

Humans always had a superiority complex, even though any machine can do its task better than any human. The infiltration of machines into intellectual arenas does seem to bother a lot of people, especially since the advance is less than generational in human terms.

Facts listed are what current MI can do. And it is certainly impressive.. What is really more impressive is the amalgamation of known mathematical equations into the MI model. Regression, Fourier, differential equations, matrix multiplication, and more all put together into a mish mash known as a neural net for problem solving. Done by humans.

Difficulties arise with this model. The LLM suffers from thresholds of diminishing returns with larger neural nets and the data wall. Eventually all human data will have been accessed, but the LLM's demand more.

AlphaEvolve takes advantage of reinforced learning, and optimization, so yes that is a huge breakthrough, which probably surpasses the bottleneck overcome by back propagation of a neural net. A genetic algorithm looks for an optimization. AlphaZero asks itself whether there is another minimum better than the first one obtained, so somehow Google has tweeked the neural net for this to occur.

Robots are cleaning messy rooms and breakdancing

And karate. Small competitions are already being held. I can imagine this spreading out, even to Las Vegas, Champions of AI, similar to Robot wars, and the Mice races, only bigger. Who doesn't like a good mashup.

AI sports may become a thing.

THE hype.
Companies are being led to believe, falsely I might add for a lot of them, that they best adopt AI or be left behind, adopting a product worse than the dreaded MS blue screen of death. A survey of companies revealed that they 'saved' around 2% of time on tasks using AI, not a so everwhelming considering that people have to check the LLM output for valid result

The AlphaZero achievements included solving 70% of mathematical questions posed to it, in opposition that it could not solve 30%. Another was that it decreased resources necessary for data centre by 0.7%, with no mention that humans themselves with other approaches had already solved the first 99.3%. And that it improved upon the multiplication of a 4x4 matrix by finding a new method. Considering that neural nets use matrices of greatly larger dimensions, is that new algorithm able to be carried over and scaled up, or will it suck up vast amounts of power ( half of data centres are used for AI ) and continue to lie in the unprofitable.

Can you name a specific task that you don't believe machines will be able to perform?

Yes. Learning and doing tasks with a lightbulb amount of power from a few pounds of grey biological tissue VS MW of power for a silicon based LLM model. Although the silicon ones are getting better by the way of passing what has been learned to 'clone' agents.

 

[sfrydman]

My position is that the race to superintelligence has a greater than 20% chance of leading to catastrophic outcomes including bad actors using the technology to carry out massive attacks, authoritarians using the technology for hyper-enforcement, human beings losing control, and misalignment of a highly competent AI system. For this reason, I believe efforts should be made to slow down AI arms races between the U.S. and China and between AI companies and to focus on narrow AI rather than general AI. Do you agree with this? For those wanting to get involved - visit pauseai.info or aisafety.com/map.

 

[dlu]

Pandoras box all over once again for new tech.
Disruptions will be inevitable, some good, some bad, and as usual this depends on point of view.

China uses facial recognition on its society - good for the gov't point on view on control.
Some police dept in the US have already been using facial recognition generally to catch suspicious activity , and some depts have been called out. The police, always looking for better tools, justify this use in the terms of protecting the people, yet not acknowledging that this blanket surveillance undercuts a person's right to privacy in a democratic society. An extension of camera on every street corner, to having an AI cop on every street corner. The end justifies the means is the argument.
These police depts are not bad actors per se, just misguided in their blinkered pursuit of justice for the wicked in their community, not realizing that this expansion of police powers ends up as a police state. Ask them if they support authoritarian and communist ideals and they would say no.

Small creep, as this hyper-enforcement example illustrates, is surely to continue as AI gets better, and more uses are found in all walks of life. Good intentions are, in my opinion going to be the cause of most disruptions. Good intentions to make things faster, cheaper, more productive, more options, are difficult to argue against as the downfalls are not readily apparent.

 

[dlu]

When HSI ( Human Super Intelligence ) takes over.
Time frame 1951 to present.

The following in an example of 'creep' overtaking the prime agenda of an organization from one of banal interests in support of its members, to that of a top down ideological agenda - an ideology into which the members did not sign up, into which they have minimal input, and of which can lead to member harm.

Imagine the HSI being replaced by ASI ( although in this case it was not an ASI running amok to achieve a goal ), and one can extrapolate how in the future when AGI becomes predominate and relied upon to make choices, can ( and will since it will be smarter ) cause disruption, and destruction of human inter-relationships, in an unfettered march towards a goal. I assume the HSI goal here was something vague such as "determine injustice". One can see deceit, overreach, narrow mindedness, all that of which ASI is contemplated to use to outsmart mere humans. The HSI in this case are all probably nice people, but blinkered in their thinking.

https://www.msn.com/en-ca/news/cana...AN&cvid=df72ae9a1fe140cba300499096013cde&ei=9

Terry Newman: Profs call out their association for left-wing mayhem​

Some Canadian academics are accusing the Canadian Association of University Teachers (CAUT) of straying from its core mission of advocating for academic rights and fair working conditions to pursuing a politicized agenda that undermines its fundamental purpose. Specifically, they accuse CAUT of — issuing an unsubstantiated U.S. travel advisory, producing a likely skewed academic freedom report with soon-to-be added anti-Israel rhetoric, and encouraging administrative overreach into equity-based hiring that risks faculty autonomy — betraying its founding principles

Founded in 1951, CAUT was envisioned as a national association that might help faculty members deal with “salaries and pensions, sabbatical leave and academic freedom” issues — basic, bread and butter issues for its members who now total 75,000 teachers, librarians, researchers, and other academic staff in over 130 Canadian colleges and universities across the country.

Over time, CAUT’s role expanded to include other, non-controversial concerns, such as the protection of intellectual property, necessary for the digital age, advice on legal support and collective bargaining, and fair employment, which includes organizing against the increasingly precarious conditions of contract workers.

Fast forward to 2025. CAUT’s scope is now far more ambitious, political and global.

Example.
On April 15, CAUT issued a statement advising academics against non-essential travel to the U.S.
YET
CAUT has not issued any sort of travel advisory for the 100+ countries that the Government of Canada lists as higher risk for travellers, including several with significant authoritarian policies that threaten faculty members’ academic freedom

members like Rachel Altman, an associate professor of Statistics and Actuarial Science at Simon Fraser University in B.C., who has no choice but to associated with CAUT because her union is. .... are legally mandated to be members of our bargaining associations — which then use our dues, in part, to pay for membership in CAUT.

 

[Greg Bernhardt]

I disagree. Frankly, I don't understand how anyone is saying all this is "just hype". We've passed the Turing test, which used to be a crazy hypothetical. You can have a conversation with your computer - text or spoken, take your pick - on any topic in any language and get a much better response than what you would expect from most humans. AlphaEvolve is making breakthroughs in math and science that the best human minds could not make, actually creating new knowledge rather than regurgitating it. Robots are cleaning messy rooms and breakdancing. The space of things humans can do that machines can't is getting smaller and smaller. It's getting to the point where people are about to say, "Well, I'm still better at whistling so no need to worry." Can you name a specific task that you don't believe machines will be able to perform?

The utility is not hype. I work in marketing, and everything we do now has an LLM layer. The hype is AGI. The concern is AI will destroy job markets and the environment so that people can generate memes on the fly.

 

[dlu]

Influencers are getting in on the promotion on what AI can do for you, better than what human interaction can, or a simple Google search on its own could do to find fitness apps.

I replaced my fitness instructor with AI. I'm now stronger, more motivated, and saving hundreds of dollars.​

MSN

I used ChatGPT as a master app, entering my goals and progress each week and requesting new or updated workout routines.

 

[sfrydman]

The utility is not hype... The hype is AGI.

I don't think AGI is hype. I don't see any reason why it wouldn't be possible and couldn't happen soon. That's the point of this thread. AGI is coming, ASI is coming, and we're not ready. Everyone should find an organization to get involved with and get involved yesterday. Go to aisafety.com/map to get involved.

 

[gleem]

For a long time, people said that LLM were fancy autocomplete programs. But think what the probability of one word following another and the difference in those probabilities for different word combinations. Think of the probability of making an error at each word. It is remarkable LLMs can produce responses that are not only coherent but also accurate for hundreds of word sequences. I think it is amazing.

Humans learn language from birth with no formal instruction, and little training if any. Our responses to new ideas involve familiar word sequences that we do not consciously think about. That makes us autocompleters [sic], doesn't it? What is lacking is a functionality that gives LLMs the ability to review or compare new responses to a memory of previously made correct responses. When we make a comment, and it does not sound wrong, we often use it, even when it is wrong. I never considered that hallucinating. Sometimes we apologize other times we obstinately defend our misstatements.

LLMs have a great amount of utility and are useful for current applications. I believe developers are in a developmental rut because of the money they can make with current applications. Of course, we do not know what if any improvements have been made toward AGI that they have not yet disclosed.

 

[Borg]

I don't think AGI is hype. I don't see any reason why it wouldn't be possible and couldn't happen soon. That's the point of this thread. AGI is coming, ASI is coming, and we're not ready.

I personally look forward to it.

 

[sfrydman]

I personally look forward to [ASI].

You also said earlier that wiping out humanity would lead to AI's destruction. It wouldn't. Perhaps that misconception is making you overly optimistic. For those reading this thread who understand the risks, find a way to help. Visit aisafety.com/map.

 

[dlu]

I came across this video about GPS. Amazing how intertwined GPS is our lives. A technology that became so commonplace that no one really gibes it a second thought; it is just there.
So what if it goes down?

AI robots, I would conjecture, have GPS enabled systems.
To shut down an AI run amok, turn off GPS. The robot would loose ability to know where it is, when it should be doing what it does, and it won't be able to empty your bank account.

PS IF humanity dies then so does AI.

The US Coast Guard’s Clever Trick That Perfected GPS Accuracy​

MSN

 

[jack action]

My position is that the race to superintelligence has a greater than 20% chance of leading to catastrophic outcomes

I would love to see how someone comes up with such a number.

including bad actors using the technology to carry out massive attacks, authoritarians using the technology for hyper-enforcement,

Aren't we way past this point with nuclear energy?

AGI is coming, ASI is coming, and we're not ready.

When will we be ready? You want to "pause", so what do you need to be convinced it will be OK to continue research and development (R&D) in that field? In the meantime, how do we stop "unlawful" R&D?

Some new human invention is so bad that some people cannot imagine anything worse and fear for the future of the World. Where did I hear that before?

In 1096, Pope Urban II took a good hard look at this new “crossbow” thing and gave it all of the nopes. No Christians were to use it in any battle against a fellow Christian on the punishment of excommunication and eternal damnation of the soul. But the weapon that would act as the precursor to the rifle was simply too valuable to leave on a shelf.

[...]

And Western knights did not like it [the crossbow]. Their armor protected them from most weapons they would face with the exception of the longbow, a weapon that took years to learn and decades to master. But crossbows could slice right through the armor at greater range than even a longbow, and shooters could be trained in hours or days.

[...]

[...] For a few years, kings largely tried to follow the ban on using crossbows against Christian foes, and Pope Innocent II continued the ban in 1139 after ascending to the position of Pope.

But it couldn’t last. Kings used the weapons against pagans and Muslims, but then had to leave the men behind while fighting against each other in Europe. By the early 1200s, they were once again common in European combat. Crossbowmen played a crucial role for Holy Roman Emperor Frederick II in 1238.

In fact, just a year later, Pope Gregory IV used mounted crossbowmen against the Lombard League, an alliance of European kingdoms that were all Christian. Yeah, the allure of crossbow power was so strong that a pope employed them against Christian forces.

Crossbows would continue to play a role in combat until after the 15th Century when advances in gunpowder slowly rendered them obsolete. First, advanced cannons were able to break up their formations from further away than even the crossbowmen could fire. And muskets and rifles eventually filled the role that crossbowmen once had.

Of course, the church didn’t love firearms either. It declared all black powder weapons to be daemonic, but armies quickly embraced them anyway.

Somehow, humanity got through, and anyone can still buy a crossbow delivered right to your door. Not the armies, though: they are now playing with supersonic bombers. I wonder what Pope Urban II would have thought about that!

 

[Borg]

Perhaps that misconception is making you overly optimistic.

Try again. I build them for a living and I'm very well aware of what they can and can't do. I have also worked on Explainable AI and AI safety programs. So I don't have any 'misconceptions' as to the state of things or where the technology is headed.

IF AGI comes, it is not going to be some Skynet fantasy. It needs humans and will for a very long foreseeable future (50+ years). I fully expect that if it does anything, it will slowly subvert our way of thinking in ways that we can't comprehend - that's purely conjecture on my part but is based on how people are currently manipulated by companies like Facebook. Either way, it would be fine for me since I expect that a truly intelligent AGI would work to fix the piles of stupid that currently exist like thinking that global warming is not real while also working to fix it.

Meanwhile, Artificial Intelligence systems are providing very real benefits to society in many ways both big and small. Self-driving cars have been shown to have far fewer accidents per million miles than human driven ones. There have been a number of advancements in mathematics, chemistry and medical discoveries just in the last few weeks.

The real danger is the people who would use it nefariously or unscrupulously (i.e., scammers, bioweapon terrorists and all-for-profit companies that don't care about social harms if it affects the bottom line).

 

[dlu]

Self-driving cars have been shown to have far fewer accidents per million miles than human driven ones.

Have you seen the research of Waymo in the field. They are seen to act more like human drivers, doing such things as rolling stops, and starting up again to continue driving before a pedestrian has reached the opposite end of the crosswalk, even though it has been 'drilled' into them to obey all traffic laws to a T. The Waymo is becoming an impatient driver, bent on schedule as well as safety.

A Unifor robot was recorded to go beserk flailing its arms about wildly as if having a seizure. It calmed down after 5 sec. Explanation given, not by Unifor, but by others that it looked as if it felt it was falling and used arms violently attempting to steady itself ( kind of when a human does windmills with arms to avoid falling ).

 

[sfrydman]

I would love to see how someone comes up with [a 20% chance of catastrophic outcomes].

When people give these numbers, they are usually not based on rigorous calculations but rather the individual's best attempt at estimating the trajectory we're on. The point is, these aren't random people but experts in the know, including the CEOs of the companies building frontier AI, and we would be crazy not to heed these warnings. https://en.wikipedia.org/wiki/P(doom)

Aren't we way past [bad actors being able to carry out massive attacks and hyper-enforcement of authoritarian regimes] with nuclear energy?

No. Only countries can carry out massive attacks with nuclear weapons but with AI, where all this is going, companies could carry them out - perhaps even individuals. And no, authoritarians do not use nuclear weapons to enforce domestic policies, but they will use AI.

You want to "pause", so what do you need to be convinced it will be OK to continue research and development (R&D) in that field?

The research should be in narrow AI, think AlphaFold, not AI that knows everything and can deceive, blackmail, and refuse to be shut down.

In the meantime, how do we stop "unlawful" R&D?

We do the best we can with regulation and international treaties. The U.S. is definitely moving in the wrong direction on both of these fronts:

https://www.npr.org/2024/09/20/nx-s1-5119792/newsom-ai-bill-california-sb1047-tech

https://www.theguardian.com/technol...ai-summit-artificial-intelligence-declaration

Somehow, humanity got through, and anyone can still buy a crossbow delivered right to your door. Not the armies, though: they are now playing with supersonic bombers. I wonder what Pope Urban II would have thought about that!

So if not every single person dies then it's okay? This is a version of history being written by the victors. Our ancestors somehow managed to pass on their genes despite hellish events and therefore we laugh at those who dare to urge caution in the face of "progress". I'm worried about people living now whose lives will be disrupted or worse both economically and militarily. If anyone reading this feels the same way, you can check out aisafety.com/map and https://existentialsafety.org/.

 

[sfrydman]

I build [AI systems] for a living and I'm very well aware of what they can and can't do... So I don't have any 'misconceptions' as to the state of things or where the technology is headed.

Geoffrey Hinton built AI for a living and he had misconceptions about what it could do. He thought something like GPT-4 was decades away. I'm curious to know what specific tasks you believe humans can do that you feel very confident no AI / robotic system will be able to do 5 years from now. I've asked other people versions of this question and so far have not gotten an answer. Maybe you'll be the first.

I have also worked on Explainable AI and AI safety programs.

Thank you. We need more people doing that.

IF AGI comes, it is not going to be some Skynet fantasy. It needs humans and will for a very long foreseeable future (50+ years).

1. I think 50 years is much too high.
2. Even if we assume 50 years, that's still less than one human lifetime. We need to prepare now.
3. The number of humans the system needs will get less and less as its own capabilities improve. Even if there is something we can do that no AI system can duplicate, the system won't need billions of humans to do that thing. It could need as few as 1. I think this is another important point that gets lost. As these systems improve, the economic value, and therefore the power, of regular people becomes less and less and we could wind up with social disaster even before we have a fully autonomous AI system.

I expect that a truly intelligent AGI would work to fix the piles of stupid that currently exist like thinking that global warming is not real while also working to fix it.

Intelligence and motive are two completely different things.

There have been a number of advancements in mathematics, chemistry and medical discoveries just in the last few weeks.

Many advancements, yes, including self-improving AI. So it begins. https://www.geeky-gadgets.com/alpha-evolve-ai-recursive-learning/

The real danger is the people who would use it nefariously or unscrupulously (i.e., scammers, bioweapon terrorists and all-for-profit companies that don't care about social harms if it affects the bottom line).

I agree these are the more pressing short-term threats and are reasons we need regulation and international treaties which unfortunately this administration and the Republican party in general will fight every step of the way (although, to be fair, it was Democratic governor Gavin Newsom that vetoed SB-1047). If anyone reading this wants to advance AI safety, you can check out aisafety.com/map and https://existentialsafety.org/.

 

[jack action]

not AI that knows everything and can deceive, blackmail, and refuse to be shut down.

Nobody is doing that on purpose, and if there were one, it would be done underground, regardless of the laws. Just look at all the anti-drug laws and their effectiveness. These laws only hurt insignificant people while never achieving anything meaningful with the big guys.

We do the best we can with regulation and international treaties.

Again, studying AI is so easy to do underground, the "best you can" is just meaningless. Unless, maybe, you create a worldwide authoritarian state, thus creating the world you try to avoid.

Best-case scenario: You will create a false sense of security for everyone.

So if not every single person dies then it's okay?

That is pretty much the definition of life.

I'm worried about people living now whose lives will be disrupted or worse both economically and militarily.

There is plenty of that going around right now, no AI required. It has been like that since politics entered our societies. Are you only worried because you think it will now affect you?

We need to prepare now.

Prepare for what exactly? And how do we prepare for something that doesn't even exist at this moment? Especially by people (lawmakers) who have no idea how it works.

Take the consent management mess that laws are making for anyone who wants to make a website. The end results of these laws (like any regulations, really) will be that only a few big companies will be able to make websites because the legal stuff will be too complicated and might cost a lot for those who don't comply correctly. Websites are the basis of information, so just a few companies controlling information? This is what regulations always do: they concentrate power in the hands of a few. Always. Very anti-democratic.

Where is the threat with a 20% unemployment (more realistically, lay-offs)? This is the whole concept of a machine: simplifying your work, such that you have less to do. If there weren't any lay-offs, it would be ridiculous to pursue it.

I say lay-offs rather than unemployment because the idea is to do something else instead, not just sitting around doing nothing.

This fear is all about the concept that you need to work full-time for someone else to be able to eat and have a roof over your head. Not only is it ridiculous on its own, but doubly so with AI and machines: The idea is that the job is already done, you are already eating and already have a house built for you by the machines, so you don't need to "work" anymore.

That being said, I think this utopian goal is a stupid one for AI and machines in general, most likely unrealistic, thus unemployment will never be a problem. Based on history, it never has been.

If anyone reading this wants to advance AI safety, you can check out aisafety.com/map and https://existentialsafety.org/.

There are better places to educate yourself about AI, like, basically any reputable university: MIT, Harvard, McGill, ... You know, people who specialize in sharing knowledge rather than instilling fear.

 

[Borg]

There are better places to educate yourself about AI, like, basically any reputable university: MIT, Harvard, McGill, ... You know, people who specialize in sharing knowledge rather than instilling fear.

Exactly. The aisafety site starts from the premise that AI is an existential threat. That isn't a site that wants to work to work with the tools at hand - it just wants everyone to not have the tools in the first place.

 

[dlu]

Take the consent management mess

enlighten me.

Or do you mean content management mess.
If that is the case, it will not work out as the smarty living in the past law-makers envision to help the lowly artist.
They are basing it on the on its way out 1950's model of bands on tour, albums, hit singles, and marketing. That model has long passed. How many new artists in the music industry are worldly known? Like Prince, The Beatles, or Michael J.