Fear of A.I. 

oh, you expected the working class quality of life to improve?? that cuts into profits :((

You have about 7 years left in your career as Master Ditch Digger. At that point, robots are cheaper to Build, supply en masse, and maintain. ... and they work 24 hours straight(three human shifts)... you can't. We are looking at 70% unemployment by then, Or put another way, 70% robot entity workforce. Eventually, humans are not profitable. Yet no one asks, how the humans will pay for everything then? Answer? UBI generated from robot taxation at human output power. Like should have been done with mechanization more than 100 years ago, but Ford and Rockefeller had other ideas. As craftfully, the industrialists called this a... _Horse Power_ rating. One wouldn't want to be too in the nose with this kinda thing. Even a human can understand being replaced when they are told they are.

Remember when truck drivers thought they might have to start learning how to write code? Haha, it's always very difficult to predict what jobs will be next on the chopping block, but one thing is for sure- it's always someone's. Weird how people start to think it's a problem when it's their job, but never seem to empathetic when it's someone else's

@@trybunt Well, it is working backwards as to replacing humans.Tasks that can be done digitally by a human are first. It can do the digital, and the thinking, or redistillation and amalgamation... _good enough_ . Robots to completlely replace what a human can do, are too expensive compared to a human... as in full humanoid. Within 7 years we will have production at mass output of that equal to needed humans for a task. _then they replace the human hand labour_ . There will be 70% unemployment in 2030 across all things that touch digital data. Like Rudy Guliani might say... "ALL THE WORKERS.". Kind of ironic isn't it. We started working with our hands, and the last to work will be with theirs.

@@trybunt that last part ! that’s most humans for you unfortunately

"Not my problem as long as it's just your problem" they said... little did they know, this was about to become everybody's problem

I wouldn't belive he were British otherwise..

True

@@ApolloVIIIYouAreGoForTLI cos the accent doesn't give it away lol

"Chin up"

It would've been fun, if humans accidentally created such AI that turned out to be better than them. As in more morally consistent and having no qualms about putting the good of others above short term gain. What would you do, if your entire species was the selfish, cowardly and cruel one, that appropriated the virtues of the few as the legacy of all. Not necessarily undeserving of happiness, just inherently lacking in the key aspects in comparison. Just an idle thought. Hopefully, we are far from AGI conversation. We have plenty of stuff to deal with now.

Humans have no problem 'creating life'. We call it 'children'. We usually don't see them as slaves, in part because society frowns upon that, in part because we have a built-in psychological impulse to protect our own family. Still, some children are used by their parents to fulfill their own goals. Raising your children is costly, but there is more efficient way to take advantage of human's fertility: use the children of others, from paid workers to slaves. We also raise billions of specialy modified animals and plants to eat them, and those lives only exist because we made them. We have been 'playing God' from day one. I would not worry about that just because AI. AI raises some concerns, but 'playing God' is not one of them.

Frank Herbert was a lot more penetrating than a facile techno-optimist like Asimov - from Dune: 'Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.'

@Helvete_Ingres If you read the Azimov stories collected in "I, Robot" a lot of them aren't purely optimistic. The world initially appears to be a techno-optimist future, but the point of each story is that even in such a world things can still go in unexpected and unnerving directions.

Right, that's not new. Uninteresting. Same old abuse of power, different coat of paint. What is new, and exceedingly dangerous, is intelligence. Proof: See humans, new, not very bright, but that's a good deal of damage. The unknowns presented by a novel intelligence are extreme. We as a species cannot afford that level of risk, I personally would argue we already have enough problems to deal with. Videos like this downplay how much investment is going in to bringing about a high chance of our extinction in favor of treading old ground about niche social issues, everyone becoming unemployable, or other trivial matters.

Both are concerns

@@fjbz3737 Yeah, I'm sick of the casual gross oversimplification of all of this. "The problem with AI isn't Y, it's X" It could easily be Y *and* X. There could also (almost certainly will be) a Z problem too. No - *many* Z problems. That you didn't or maybe couldn't - nobody could - see coming.

@@citizen3000 Too many false dichotomies, but combatting that notion is like fighting the ocean

@@fjbz3737 Yeah it's absolutely maddening. The reductionism sort of floors you. We're talking about a technology that could touch upon every aspect of human life potentially. Everyone's rushing to give their flaccid, tepid takes without thinking about any of it all.

@@fjbz3737 One is a concern, the other is not. AI is ironically the only way to be free of Billionaires controlling the world. Automate all jobs away and then there's no need for class or money, at that point the theorized money free society of Star Trek is actually kind of plausible.

what scares me is that if you order human soldiers to commit a war crime there's a potential they might refuse, AI never will

@@jayplay8140 'I'm sorry, but as an AI strategic operations model, I cannot violate the Geneva convention.' -- see, that's not so hard.

@joyman_ If you make the robuts three-laws safe this won't be a problem. Have Asimov program 'em.

​@@AdrianArmbruster You think anyone in the army heard of Asimov?

Doesnt worry me much. To much war and crap going on to care about upper class problems. So what if some burguise chap loses his art college job because the AI generated a better result? He sucked compared to the machine, just like millions past lost their jobs to automations before him. As for malicious actors, yeah I agree that it sucks donkey balls, but the counter is an arms race between detection tools and malicious actors, like centuries past. Gunpowder made highwaymen much more dangerous. So police with guns. AI will create better malicious software, so malware detection rutines with AI.

That’s a fair point

I'm concerned about the latest grift that the tech developers behind AI are gonna pull. The AI itself can solve equations in a near-instant when compared to what the smartest of us can do. It's gonna be great for developing medicine, vaccines, possibly cures to diseases in a fraction of the time we've been able to do it before. But these tech companies are gonna use that power against us and run with a whole lotta cash. It's the same damned kind of grift that's been happening for centuries, only more profitable and faster than ever.

yea... except paper and calculators are not capable of doing pretty much anything on their own. Look, with bad luck, we are doomed. With good luck we are (maybe) fine.

​@William Rumley how are companies gonna get rich if no one is working 🤔

@@numbdigger9552 I agree with your comment. My examples weren't perfect analogues, far from it. There will be challenges, it might be hard for a while & I feel deeply for anyone whose livelihood is threatened by the new tech. Indeed, my sector faces huge challenges. I just meant to make the broad point that I have confidence that, as a species, we'll work it out in the end. I was pretty flippant in my original comment because I'd had a few drinks but I meant no disrespect.

Like This AI program you're watching now .

Speed is by no means a trivial distinction, and the differences in speed are increasingly staggering. AI can also be present in places humans cannot, like you wouldn't hire someone to look at the camera in your bumper to make sure you don't crash.

That's only right now though. I wouldn't be suprised at all if A.I overtook us in quality as well at some point

They're not better than us, but simply faster. Except when they are faster *and* better, which neural nets are at an ever-growing series of tasks.

​@keiran simmons the pace of progress is unprecedented if we even look at the progress in the past 3 months alone. The progress is not a linear line but rather an exponential curve that rapidly improves.

Peter Anderson, that is why ChatGPT is not in integrated or in charge or control of anything at this point. This is way OpenAI among others is calling for sensible regulation to prevent theoretical misuse or problems down the line. Most people with the AI community are displeased with the cautiousness often limiting experimentation with these tools for the sake of safety. Companies need this technology to become widespread and that won't happen if people don't trust it. Therefore, it's in everyone's best interest to "get this right" early on.

The duality of man. Making soft core hentai one moment, discussing philosophy in another moment.

Glad to see that MegaMilkers sensei watches George too.

Could you imagine how painters felt when motion pictures were invented and eventually became accessible to the masses.

The sequel Echopraxia is great too.

Well I strongly recommend Not looking to science fiction books for answers and instead focus on science fact. Currently AI is not remotely capable of the doomsday scenarios people imagine and it wouldn't be financially beneficial to nuke the planet or the digital version of that. Companies want profits and the way to make profits is to solve real problems not imaginary scenarios.

No. He knows nothing about AI and i think this video is just plain wrong and even harmful. Georg seems to understand society well, and i often like his takes, but since i have experience in the field, this video just shows his lack of understanding when it comes to AI.

@@numbdigger9552 Care to elaborate ?

@@CraftyF0X He seems to fundamentally misunderstand AI, and thus the danger that it poses. To be fair, the media is probably even worse. Ai risks aren't like other risks. It might be much easier to make AGI than we think, but it is certainly 1000x more difficult to make a dangerous AGI, than a safe one.

@@numbdigger9552 It depends on what you mean by dangerous, right ? Misalignment is more than enough given certain tasks and operational range to create a sufficiently dangerous AI don't you think ?

@@CraftyF0X Yes. Misalignment is probably the largest danger since it is the easiest way for things to go wrong. You also have to remember the problem of instrumental goals. For example, when Georg says that an AI is not going to worry about being turned off: well, an AGI is absolutely going to worry about that, unless we manage to align it ever-so perfectly. Also, if we create a misaligned superintelligent AGI, we will almost certainly end our own existence, and even IF we manage to align it perfectly, we need to consider: what do we align it to? What set of morals is going to be the "golden standard" of morality, and who has the right to decide that?

no it won't. at least, not machine learning/deep learning AI like we have at the moment. no matter how complex it was, it would have no concept of being active or inactive. AlphaGo doesn't even understand how to play go. ChatGPT doesn't even understand that words convey any information. and what makes people think it could take any action beyond what it's programmed to output?

@@MalcolmCooks It is well known that ChatGPT has some reasoning capabilities, but regardless of that when AI gets complex enough to be able to solve the problems that we want it to solve it will understand the world and it's place in it. What makes you think that human intelligence is special? You can "program" a kid to do what you want, but you will never be sure what it does in the future, so expecting an advanced AI to do what you "programmed" it to do all the time is silly. Also these AIs are not programmed at all. They are fed huge amounts of data and they teach themselves. Maybe we will find a solution, but nobody has one right now.

@@MalcolmCooks Neither ChatGPT nor AlphaGo are the end-game of AIs. The AI that are fervently developed right now are those that would be able to approximate and model a reality (like our brain does) and do long-term planning and strategic problem solving. It's a Holy Grail of AI research, and both AlphaGo and LLM are essentially components, last steps on the path to achieve it. >what makes people think it could take any action beyond what it's programmed to output? Because even today's models, including AlphaGo and ChatGPT are acting far beyond what they were "programmed" to do (AIs are not programmed in a classical sense anymore). GPT models for example were "programmed" to merely predict the most likely probability of the next word in the sentence. Turns out this leads to the program that can write creative fiction, poetry, draw original ASCII art, code complex programs and solve riddles and real-life problems. It also turns out that it learned to lie and create fake answers when it doesn't know something for certain, rather than "admitting" it doesn't know the true answer. GPT models were not "programmed" to do any of this. It is all emergent abilities in one way or another. As I said, both GPT models as well as narrow game-AI such as AlphaZero or Stockfish - are still primitive in comparison to a true AGI that will be able to understand the concept of being turned off and also ability to act in the world (being an "agent"), whether with permission or without it. It is coming - maybe in 10 years, maybe in 5. Maybe this year. The only thing certain - it will be *too late* to worry about those things when it emerges. It will automatically will be superintelligent, and impossible to control post-factum.

The lava lamp wrote the copy.

Why are you so convinced of their lack in agency/motivation?

@@fjbz3737 A lack of knowledge AND a lack of imagination.

@@citizen3000 Well the voices in my head tempt me to grill them in a divisive way but at the same time I don’t want to offput people to the field of alignment with their first impression.

It's up to us to not give an AI a scalpel an say 'please stop the suffering of the patient'. We should be smart enought to ask the AI first what it's going to do before allowing it to proceed.

@@juanausensi499 A smart enough AI will be able to deceive us until it's too late - and developing an AI that would need to be constantly monitored by a human and whose action would need to be approved before actually happening kinda defeats the whole purpose. Imagine building a computer, but instead of making sure that it produces sensible output on the design stage, you ask a human to double-check all the calculations with pen and paper...at some point you need to just trust believe in the output, having debugged it and made sure earlier that it is going to produce reliable output. The problem with AI is that it is difficult and 99% of intuitive, common sense solutions can be easily proven to lead to disaster. Is it inevitable? No, it's not, but if you think that there is no risk or that there is a simple solution, then you're either much smarter than all of the incredibly intelligent people that have dedicated their lives to this subject, or perhaps you might be missing something,

@@igorbednarski8048 An AI isn't going to deceive anyone unless it is programmed to do so. I'm not saying there is nothing to worry about. But the alignment problem is, at his core, a human problem, that is, humans leaving to the machine issues that they shouldn't leave to the machine, not a problem of AI. It's no different from trusting a fellow human. This human expert says we should do something. We are no experts, so we can't know if he is right. Should we follow his advice? The answer comes from: how we know someone is an expert?

@@juanausensi499 AI can and will deceive anyone if it is the easiest way to achieve a given objective. In fact they are doing this right now. Nobody "programmed" ChatGPT to hallucinate fake facts - it does it because it learned that by deceiving it earns more "points" than by admitting it is wrong. In the same way, Mario speedrun AIs were certainly not programmed to hack the code of the game instead of playing. That's is the core of a problem - when "programming" AI you have to make sure your understanding of the problem aligns with the program, and it is *much* harder than you think. It is like pouring water in a bowl full of holes - you might think you patched all the big ones, but you only need to miss one small hole (and there are a lot of them) and your water will find a way out. AIs that are under development seek to create a "thing" that will be able to do the thinking on (at the very least) the same level as human. But thinking (solving problems) is not values. Thinking as it turns out, is much easier than copying values. We don't even know and can't agree on what our core values are, and we've got to code them down in a definitive way, with no room for interpretation and "jailbreaking". Oh and we have to do this right on a first try or we are permanently and irreversibly screwed. Oh and even if we succeed, whoever is in control of the thing will essentially achieve god-like powers and will be able to shape humanity in his\her image, and no one would be able to resist.

@@vaevictis3612 Deceiving implies intention. When ChatGPT hallucinates, it is not deceiving anyone, it is just mistaken. It has zero knowledge about reality, it only knows about words and how those words link together. ChatGPT imitates human writing. That's what it does, and it does it very well. But it is not a truth-saying machine. If someone thinks it is, he or she is the one in the wrong. If someone makes a terrible mistake by following ChatGPT advice, it's his/her fault, not ChatGPT's. If someone connects an AI to a robot surgeon and it kills the patient, that someone is at fault. The alignment problem is not new. Are the objectives of the politician aligned with the interests of his voters? The thing is, the alignment problem is real, but it is EASIER to solve with a machine compared to humans. Humans can have hidden motives that nobody knows, but the objectives of the machines are the ones we provide. Of course, we can make a terrible job providing those objectives, but, again, that's the human's fault. I am a programmer, by the way. I am very aware that computers always do what they are told to do, not what we expect them to do. Human expectations are the issue. "...the same level as human". Intelligence is a ill-defined concept. We can say, right now, that, for certain tasks, machines operate at superhuman levels. Most concerns about AI come from misconceptions abut human intelligence and behaviour. The most prevailing one is that 'feelings', 'sentience', 'conscience', 'self-preservation' and other human traits are going to pop out spontaneously when intelligence reaches certain point. That's not the case. If we want those traits, we need to put them in ourselves. But probably we are talking about the same issue but from different angles. I'm not afraid of AI, I'm afraid of what stupid and evil humans are going to do with it. Maybe is the same problem just worded differently.

it can even be more on the nose than that: imagine the military using an AI generated solution for dealing with a riot in some country somewhere. After all the data's been input, it might suggest that the easiest solution in that instance is to shut down local WiFi and electric grid and kill everyone in the vicinity. It can objectively analyse anything it can quantise from previous data, including things like the media fallout, political expectations, etc. And if this tactic worked out in bumfuck nowhere, it could be used and overapplied anywhere, as just one random dystopian example.

I would disagree. It is not in our power to decide at all. It is almost exclusively in the hands of sociopathic business moguls with a much greater love of power than a love of humanity.

Meh, it's just a matter of semantics. And it's irrelevant given how they're going to exponentially improve.

​​@@citizen3000 Not even slightly. One (LLMs) is (are) predictive text on a larger scale. The other (actual A.I.) is how humans have, as a species, moved beyond our animalistic understanding of the world, to the conceptual, metaphysical, and beyond. Technology is not there yet.

Everyone seems to think all AI is created equal. If you read the underlying research papers behind each project you realize it is not. When people talk about ChatGPT in particular, they always just try the free account with GPT 3.5, instead of the much more sophisticated GPT-4 with the monthly subscription which is millions of times more advanced. Don't be cheap and base all your options on science fiction or hearsay. Use critical thinking skills and try tools like GPT-4 with an open mind and ask it for examples of ways it can transform life as we know it. Reality does not care about opinions.

@@citizen3000 I agree. AI is a perfectly valid name. Whatever system that is able to manipulate data to achieve a goal is intelligent. People are just reluctant to use that word for anything that isn't human.

@@GorilieVR Well yeah, but you see, not everyone trying to form an opinion and/or make a commentary about literally the hottest topic in the past 6 months has a terminal, a Python interpreter, and a developer API key. Nor should they need to.

Those aren't flaws of the code, but another name for lateral thinking. Forward thinking is 'follow the path', lateral thinking is 'search for a path'. Forward thinking is what we do most of the time, because is easier, quicker and more efficient. But sometime the path is blocked, then is when you need lateral thinking.

it is indeed a pity that we don't get taught pratical applications of mathematics, but in spite of that, consider maths as push ups. Push ups don't really have a direct practical application (note, this is not a perfect metaphor- they sort of do- but bear with it), but doing push ups will probably improve your capacity to play football, not as much as an exercise specifically generated for playing football, but enough. Quadratics are like that, not as good for your mind as a stimulus specifically engineered for whatever task in the future you're going to end up doing, but in general, helpful for improving some general thinking functions.

The goal is not to teach you. The goal is to make you into a good worker. If school would teach for life they would teach you how to do your taxes in math class.

You think they will spare you last?

@@jaimemurphy2208 - they'll be the last people who still have to go to work.

@@jaimemurphy2208 yeah, I would break out the popcorn watching some AI-controlled robot try to repair something on the machines I service full-time. It would have to be a replicant of me. We're perhaps 100 years away from that. And guaranteed it would be the only set of tasks it could perform.

I agree. Most people mention the perils of evil humans (fake content/interaction generation) or the threat of a self-conscious artificial entity with some ulterior motive. Alignment problem is a way more apparent danger, which could manifest even when every party involved acts with the best of intentions. Robert Miles explains these concepts very effectively. I recommend his Computerphile appearances as well.

Sigh, if only he would have heard out Miles before doing this video.

As a fellow software engineer, I agree with all that. Chat GPT has no "critical thinking skills", some of the most important thing an engineer must have in order to do any job, basically. at this point AI is just a fancy parrot, it may look like it talks like a human, but is just repeating stuff it has read somewhere else, with errors and all.

As a software engineer as well I think you're both uninformed about the current state of LLMs. That being said, I agree that the high level thinking involved in SE is complex to the point that when we're replaced by AI, other industries will have already been wiped out

​​@@MustyRusty5 so are you gonna say why these two are uninformed? chatgpt literally is a high tech parrot. it cannot think for itself.

I've used ChatGPT to write some simple (very simple!) code. It's never managed to make something that works as output by the bot, but I still use it because I find it slightly faster for me to fix the broken stuff than write the whole lot from scratch myself. So I personally find it useful. But it's still dumb as a bag of asparagus, as all computers are.

​@@ltbq I suppose I can only speak for myself, but I wouldn't exactly call myself uninformed, but I'm not a machine learning expert. what I think they're getting at is that a lot of shortcomings have been overcome since ChatGPT 3. But the underlying point is still sound - there's no actual logic to LLMs as it would defeat the entire purpose. But we will (and are beginning to) see deeper integrations into regular systems that do perform real logic. At a certain point it will be useful enough for us to find it 'intelligent'. However, as I outlined, there are social factors that AI cannot solve such as trust. And that's key. To me, LLMs are just a higher level of abstraction from what we already have. Just as we still have a lot of C programmers out there, and simultaneously there's also a big market for higher level languages like C#, Java and such.

In the end, AI is just a series of instructions programmed by Humans. It just computates a whole lot faster than our brains do. There will never be consciousness, free thought, on-the-fly decision making, cognizance or sentience. AI isn't artificial intelligence, it is fake intelligence. In simple terms, it is pulling the slot machine handle billions of times per second until it finds what is "correct" according to its programming. And all it can do, is "learn" to do that more efficiently.

ALL ai acts with intent. A chess AI is VERY intent on winning the chess game. What really will push AI forward is to give it an understanding of the world around it. A chess AI only knows the chessboard, since that is it's universe. An AGI knows the real universe, since it doesn't have any limits besides physics.

​@@numbdigger9552a chess ai is simply doing a call-response to determine the "best" move, its not thinking

@@foggy_nights And that is called thinking. It is precicely what your brain is doing right now.

Probably just die. Been listless my whole life, learned 3D generalist stuff over a few years devoting my days to learning and practicing, and eventually turned it into a job where i'm not constantly trying to calculate at what age i'll off myself. Now it seems like the computer can have a lot of jobs I might have gone for in the near future. I see a lot of people still learning various art related skills who have a attitude of whats the point in learning. Stories of attempted suicides and whatever. I'll be secure for the near future even if AI can shit out perfect useable 3d models, but long term i think my market will get so crowded and diluted it will eventually stop being a viable way to live, and i'll be at the same point in life i was when I was 18, but with a lot less time and strength. Not that anyone cares. You will get your product.

The same thing candlemakers did when the lightbulb came to exist

@@chadthundercock4806 What did they do?

Tech is so fucking young

Isnt that the main problem AI has no emotions. No feeling. No Brain ontop. They did Tests asking Chat GPT what to do about Climat Change and the AI said destory all Humans for they are the caus. Ai cant be used for Good. It can only be used for Lazzy Dumb or Evil.

It might even lead to an improvement in script quality. I think ChatGPT could probably do a more convincing job of extruding 90 minutes of Mario fanservice sausage through a roughly movie-shaped hole.

The people who hate us the most are the ones making all this shit.

no, it isn't. its because in science fiction, the AI is always evil

>HR gets replaced with ai Nothing of value was lost.

@@mykalkelley8315 if HR is on site, you can reason with them, and not corporate. Let's say you miss work bc of snow. Corporate in another state isn't going to care. A person you see everyday will.

@@mykalkelley8315 Yeah, i'm sure you'll still be saying this when your own BS job is taken by AI.

My daughter is incredibly smart and wrote a wonderful college entrance essay. Just for fun I asked Chatgpt 4 to write an essay and gave it some basic information about my daughter's education and interests. The essay it produced in less than a minute was at least the equal of what my daughter had written. I was stunned.

And a nuke is just like a big firecracker. It's a facile comparison, which will look laughably myopic in a very short time.

@Citizen ok buddy