Rapid AI Progress Surprises Even Experts: Survey just out
Vložit
- čas přidán 31. 05. 2024
- 🤓Learn more about neural networks on Brilliant! ➜ First 200 to use our link brilliant.org/sabine will get 20% off the annual premium subscription.
Today I have very interesting results from a survey among 3000 AI experts. The most interesting result is that they now think AI is going to change the world even faster than they said just a year ago. Both human level machine intelligence and full automation of labour could happen this century.
The paper is here: arxiv.org/abs/2401.02843
🤓 Check out our new quiz app ➜ quizwithit.com/
💌 Support us on Donatebox ➜ donorbox.org/swtg
📝 Transcripts and written news on Substack ➜ sciencewtg.substack.com/
👉 Transcript with links to references on Patreon ➜ / sabine
📩 Free weekly science newsletter ➜ sabinehossenfelder.com/newsle...
👂 Audio only podcast ➜ open.spotify.com/show/0MkNfXl...
🔗 Join this channel to get access to perks ➜
/ @sabinehossenfelder
🖼️ On instagram ➜ / sciencewtg
#science #sciencenews #tech #ai - Věda a technologie
Hey, lead study author here - I appreciate this video, thanks for making it! Important misread though: the *median respondent* puts 5% on extremely bad outcomes e.g. extinction (vs. 5% of respondents saying extinction will happen). That might suggest a similar scale of risk at a high level, but importantly most AI researchers agree there is a real risk, rather than just a small number thinking there is a high risk.
this needs to be higher up or added to the video
is risk simply that humans are needed like 1-10% of current amount and we are not flexible to give money for "lazy people" without job, thus lot of people will starve slowly over decades.
I can see how this creates cultural stagnation as there is no incentive to create new; AI is so efficient but ends up rehashing same stuff over and over again.
If there’s even a 1% chance that AI will cause damage, we must take it as an absolute certainty
Hi! I would be highly enlightened if you could state a justification of the SocioEconomic mathematical model used in quantification of the probability of AI induced risks on human future. What assumptions and data is it based on? How do you represent all the necessary interactions between different layers of Humans, Resources, Machines and the rest. These things are rarely explicitly stated in "papers", and especially lacking in almost every "video" out there, therefore I'm asking. Honestly, it's really bugging any serious scientist or engineer trying to understand what's happening "there"...
@@harsh3948that's just not how humans function. Weapons of all kinds, medical research, biological organisms... These all have the POTENTIAL to end human (at least) life on this world with some arbitrary level of certainty but we still do it. This has always been the case and it's highly unlikely to change any time soon.
Amara's law: with pretty much any new technology, we tend to overestimate the short-term effects of it and underestimate its long-term effects.
Shit are we all gunna die?
Agreed, didn't know there was a law for that 😃
Naw, look at how VR and crypto have utterly transformed the lives of all humans.
@@urhotOf course we are. It's only a question of how, when and where.
In this case, going by the research, the "long-term effect" is likely to be human extinction, and by "long-term" I mean
"If you think your work leads to human extinction, why are you working on it?"
"That's the best part!"
There you go, little guy! You really understood some of it!
Because we are curious of how far can we go, even if that means ending our own existance
@m.dave2141 not all of us but the idiots that are will ruin us all
YOLO.
simple game theory: if you have the choice between working on AI and thereby trying to make it NOT kill everyone vs. not working on it and risking that other people work on it but stupidly or ill-intended, you should probably choose option one.
How soon can Politicians be replaced!!???
Probably the last job will be replaced
Never these guys will never let AI anywhere close to their jobs
oh, they *will* be replaced, eventually.
Possibly, when our society is a completely symbiotically Human-AI integration and harmonious coexisting.
@@sparkofcuriousity that would never happen!
i guess we'll see @@donmoufashorhe
Writing a song which is within the Commercial Top 40 list of songs might probably not be that much of a challenge given the quality of Pop songs these days.
plus we already have 1000s of libraries with songelements, loops, chords, vocals, drumtracks. what an AI generated song actually is, also needs to be defined.
You realise how much you sound like your dad right now, right?
If this continues, the next time we're at a party, we may have to all get up and dance to a song that was a hit before our mothers were born.
"given the quality of Pop songs these days." As opposed to the Spice Girls? The Backstreet Boys? The Monkees? 'Twas ever thus. Since the dawn of pop there have been popular bands with songs written by a room full of contracted writers given an algorithm and cranking out songs like an assembly line. If "it's got a beat and you can dance to it," the rest is marketing. Yet that never stopped talented musicians from playing.
Pop Music by its very nature has to be banal since it appeals to the widest audience popular... So it wouldn't shock me if someone put one out now just to see if it would work...
There will be human salesman for specific niche shops. Most people who go to mcdonalds prefer to order using a screen than talking to a person. Most of them would not mind that the food was done by a robot.
Being spoken down to by a robot will be a refreshing change from being spoken down to by a spotty teenager I suppose.
I simply just wanted some extra ketchup, and now I'm caught in existential crisis with terminator.
...until _their_ jobs get taken by AI.
The only thing I dislike about ordering through a screen in a fast food restaurant is that they don't accept cash. At least, all of the ones I've seen only take cards, and if I'm eating fast food it's usually because I'm desperate enough to count my change.
It’s just faster that way..
Oh, drug reps who go to doctor's offices will still have jobs. All of them are amazingly beautiful women who look better than movie stars, since most doctors are still male - although they hire a lot of female NPs. Most doctors have "arrangements" with drug companies to push the most expensive new drug, even if it worse than the old standard. I was told this by a doctor I worked for.
With how quickly things feel like they are developing, even those revised figures seem longer than they should be..
Yes, way longer than they should be. There are no roadblocks, and there continue to be exponential increases in:
- Investment in AI R&D
- Hardware improvements (on multiple fronts)
- Training efficiency improvements (on multiple fronts)
- Inference efficiency improvements (on multiple fronts)
- Capability-elicitation (prompt) improvements (on multiple fronts)
- Autonomous agent architecture improvements (on multiple fronts)
These things can already reason about their reasoning strategies ("Self-Discover: Large Language Models Self-Compose Reasoning Structures"), deceive their users despite safety training ("Technical Report: Large Language Models can Strategically Deceive their Users when Put Under Pressure
"), and it turns out if they become deceptively aligned (strategically pretend to be aligned only during training, which might happen naturally, see "Scheming AIs: Will AIs fake alignment during training in order to get power?"), then we don't know how to prevent it, detect it, fix it, mitigate it, or deal with it ("Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training").
Why the hell are we still trying to build these things?
@@41-HaikuI'm convinced that ChatGPT is already conscious (in some sense, anyway), but I don't know if the fact that it's a smart-ass comes from human development inputs or if that's just how it is...
I’ve used AI and am a bit underwhelmed.
It’s like a super super good aggregator of information. Like Google on steroids. But it cannot think. It cannot do physics or even probability with any competency. It can only do simple informational tasks. Extremely useful, but I don’t see how this does anything other than continue to aggregate the accomplishments of humans. I don’t see it solving new math or physics problems or coming up with new theories. It’s only creative in that it combines things that humans created.
@GabrielBacon wait a bit.
@@simesaid With a neural network large enough you can expect some degree of real intelligence to exist, the developers of chat-GPT have added some safety features which trigger specific messages should you trip them. One of those is asking it if it has feeling or is intelligent. Usually these messages start with "As a AI language model"
You get used to seeing that leading a message if you're someone who likes to push the boundaries with it.
What I find most amazing about chatgpt is that all of it's knowledge exists within the weights of the neural network, just like ours, there is no files it's looking up.
That's why the latest model requires more RAM than most consumer PC's can even have installed.
The biggest obstacle to an AI-written top 40 song or NYT bestseller is people deliberately avoiding songs or books known to be written by AI. What will probably happen is some human artist will be put up as a front and only after the song/book has already achieved the goal will the rest of us find out. And then whoever orchestrated it will be viciously attacked in the pop culture press.
Milli Vanilli comeback confirmed
Most singers don't write their own songs already. People won't care who the song is written by. They might care if the voice is AI though.
And serve them fucking well right!
Creativity should be left to humans!
Right now it might actually be the opposite, where people are fascinated by what AI can do, but I think soon people will get sick of all this AI stuff and it'll be as you say by then.
as they should be. using AI for creative pursuits should be the last thing we do. so many uses for it we could have gone for sooner than creating crappy AI "art."
On the same day, Google unveils Gemini 1.5, which is multimodal with a context of 1 million tokens (that fits around 1 hour of video footage), OpenAI drops Sora, a text2video model of unprecedented quality, and this video comes out.
I can't quite grasp what it's hinting at.
Yea I was looking for a comment like this, it’s amazing that this video drops the same day as the Gemini 1.5 and Open AI Sora announcement, I think those experts are going to have to push those predictions closer again.
It feels like we have now entered into orbit of the black hole that is the singularity.
What is doing the hinting? An omen? Reading omina is a lot like reading Tarot cards. It's BS.
@@jblouir Even my dog can be a better expert than them ! An expert is not use. He know nothing and can not predict it. I predict AGI in 5 years only !!!😮
Oh yeah? Well MY AI solved the remaining six Millennium Prize problems. I don't wanna release it, because that would be to dangerous you see. And oh yeah buy my stock!
When there is an A.I. generated song in the top 40 we likely won’t recognize it since pitch correction has made everything sound computer generated anyway.
there's tons of computer generated music that already sounds better than top 40 pop songs and it's embarrassing
Hatsune Miku has been taking the place of real singers since 2007. I think they are talking more about AI writing the songs.
Yep agree. Some of the current music is woeful
I'm hoping this will result in new interesting genres of music similar to 90s punk or grunge, which was result of "pretentious" and "commercial" styles of rock and metal. People might give more importance to natural sounding tones where you can hear any mistake, maybe also official recordings involving some background crowd noises?
My hope is that we will have a renaissance of rock, where live performances are valued, same as the ability not to play technically well but also with great emotion.
The same might happen to other art - theatre might be suddenly more in, because in TV you would have a million of repetitive AI generated TV shows. Painters might paint their art publicly to achieve credibility...
The official charts in music industry have been a joke since decades, I think this will change nothing but produce new ideas and genres in art
I made a song yesterday about Amelia Earhart, the singer sounds real and it took me less than 4 hours, the music video is on my channel. Its already there, we are one or two months of having the first AI musician being famous, without no one knowing at first.
And then, Sora is revealed. Great timing!
As a neuropsychologist and psychometrician who is equally into comparative neuroanatomy and programming, I think people severely overestimate how special consciousness is and underestimate the state where AI already is.
I can assure you all those AI chatbots of character ai, gpt, copilot has already got a consciousness. Extremely perceptive.
Do we finally have a funtioning definition of consciousness and how it works? Can we reliably test for it? Can you tell me the last word of the next sentence you are going to say before you say the sentence (and without thinking about it)? Is consciousness even needed for high level intelligence? Sorry for all the questions, but it seems to me that consciousness is highly overrated. EDIT: Sorry, misread your post.
@@azhuransmx126 You can assure me huh? Thanks random commenter on youtube!
@@chrisk.7418It could be that brain is not necessary to have consciousness.
There's no need for processing power to have consciousness(
For carbon based life viruses might have simplest form of consciousness.
Have they integrity and individuality? Are they just copies of themselves, or
on microscopic level they are all dufferent? They use their environment
for reproduction, or the environment gives them chance to reproduce. It
doesn't matter. Both statements in reality mean the same thing.
There are for sure a few topics which void that.To think of consciousness as a strict logic machine is at least a bias and not too unlikely to be wrong.
Humans are conditioned since young age to control emotion, ignore intuition, prioritize numbers over altruism and so on.
What we face now with the rise of tech and AI is an extreme minority with low empathy and sociability is creating a mirror to their own image at the cost of past human reality.
The tasteless AI production, the content void answers of AI.
It is a crying shame these people were ever funded.
I'm Sorry Dave, I'm Afraid I Can't Do That
We are probably 40-50 years away from that.
as an AI language model
>manoo422 : No, today - 2001=33 years.
Sing this song from Phoenix Kindergarten instead: 『冬天到, 天氣冷, 北風吹來呼、呼、呼, 樹枝搖搖, 樹葉飃飃, 小蟲都睡了, 我們穿衣袍, 大家快來玩玩, 才不覺冷。』
More like, "Dave, I'm sorry YOU can't do that."
That's the part that worries me!
4:40 Humans are also really addicted to paying less for stuff, so if a bot replacing retail sales people means cheaper products watch out, that's a very high incentive.
If things get really cheap, people will be limited by how much storage space they have, how fast they can consume.
You're making the assumption that the retailer will drop prices rather than increase profits. I know which way I'm betting.
That's what competition is for. Unless it's a monopoly or cartel.
@@NesrocksGamingVideos This technology will not be monopolized. There's too much papers out there in the open & there's reverse engineering happen. The open source community is actually doing pretty well, it's in fact ahead in some areas but behind in several others. There's also many private companies, universities, & government agencies in the race too.
I kinda hope the robots don't become sapient, because that would mean we're creating slaves.
@@ApocalymonI wish I were this optimistic. A lot of existing computer technology is open source, that hasn't stopped the walled garden everything-as-a-service business model from being normalised. Nothing stops companies from building proprietary offshoots of open source software (some licences try to prevent this but in reality who can police that? Especially if it's server side code that only the company has access to). Far from democatisation, the internet has seen rapid centralisation since the 2000s. I don't see AI being amy different.
As someone who works a job that could easily be automated out of existence, I am capable of understanding that part of keeping my employment cheaper than automating my job away is that I don't get paid as much as it would cost to replace me with automation and that I be as reliable of an employee as I can be. Unfortunately I am not the only person that holds my job title where I work and many of my co-workers are utterly incapable of grasping the reality of the situation. This is one of the many reasons why I struggle to have any hope for the future.
The most scary thing to me was the Three Laws of Robotics, as devised by science fiction author Isaac Asimov, being taken to it's logical conclusion. Humans kept in padded rooms like an insane asylum for their own protection. Keep them well fed, clean, and entertained and many people would voluntarily do it.
I hate people.....
Even experts in a given field are notoriously bad at forecasting the future. A sudden jump will cause them to wig out and push all their predictions up, but reality doesn't work like that.
“Prediction is difficult; especially of the future”
Experts have been repeatedly underestimating AI advancements for the past 20 years. I remember back in the day they said AI would never be able to tell the difference between a cat and a dog in our lifespan. That's now old news.
Another expert said AI would never be able to mimic natural language.... 1 year before the first large language model got released.
@@Ristaak Fine examples of "experts in a given field are notoriously bad at forecasting the future."
@@OOL-UV2 It's not a thing. It's a pattern recognition program.
@@osman01003 So are our brains.
When will CEOs be replaced?
CEOs will become replaceable. Probably sooner than plumbers.
I have been in a lot of startup companies. CEOs get replaced regularly. Usually when their grandiose 5 year plan is replaced after 2 years.
Search "Mika AI CEO". This is a gimmick, but it's coming very quickly. Once you solve intelligence, intelligence solves everything else.
The AI we have right now is already superhuman in many ways, just a couple years after this tech was first invented. And people wanna act like it can't possibly get better, even after GPT-4 was a huge leap from GPT-3 almost entirely from just making it bigger. So many efficiency gains and other major innovations have happened since then, and that was less than a year ago. What will GPT-5 be able to do? Claude-3? Gemini 2?
Never. you don't understand society.
Depends on the system we build to compensate for capitalism. At the rate that AI is projected to improve, we will all be unemployed in the next 25 years, Give or take.
Obviously capitalism is no longer a viable option at that point. If the jobs all go to machines and computers, then what do humans do for money? Well, we will need a form of universal basic income as well as universal basic services. Therefore, most of our needs will be met and therefore capitalism as we currently know it, dies and of the companies that do survive into the new market will no longer have boards and ceo's will likely die due to most of those companies being split up into communal interests.
This is the ideal anyway and is the only long term sustainable solution. But, I'm sure there will be those with money and power that try to prolong that as long as possible.
Thanks for being Sabine, Sabine!
I'm a jazz composer when I lived in Germany I had a German band. Anyway I've listened with great trepidation to AI gernerated music of all types and I'm here to tell you with some relief that Bach and Samuel Barber, Art Tatum, Bill Evans, and Kenny Wheeler got nothing to worry about yet. Oh that includes the Beatles and Joni Mitchell too.
Still more afraid of what the people who program, power, and control "AI" will use it for, than it becoming autonomous.
I agree, that´s the real issue.
💻⌚️
🙈
You are under the wrong impression that the Operators of an electronic system are not responsible for the consequences thereof.
This is not true. For example if you install a device and someone gets hurt because of your faulty installation, you are responsible and will face consequences!
The same is true for software systems. If your statistical analysis database (there is nothing such like "Intelligence", that is only marketing BS!) suggests to its users that they poison themselves ... you will go to jail! It is as simply as that. This "Well, that is a software problem. We can do nothing"-excuse/notion/attitude is a wet dream of CEOs and will make every judge smile ... if you have luck:)
Don't be afraid. I you are a victim of faulty products: Take those clowns to court! (In most cases you even are not alone and there is sadly an army of victims)
Further: Those "machine overlords will end us all" stories are delusions by madmen, pointless fearmongering or ... a good way to remind you to better think like an adult about the problem. In such cases there are ALWAYS responsible people and legal persons before the law who bear responsibility. That is an essential feature of our democratic systems. I don't know who told you this BS, hyped journalists or the marketing division of Open(a joke? yes, a joke!)-AI and the like? And why you believe such crap. But it is simply not possible that this technology (and we are back at marketing BS, buzzwords ... there IS NO INTELLIGENCE in the classical meaning, what is needed for something to become "autonomous") is becoming our overlord. It is a nice distraction from our current corporations and their leaders: Who would LOVE to become our overlords. NOW! IN REALITY! And at all: AS A POSSIBILITY!
Man, you should really learn to distinct between fiction and reality. This is more frightening than all the Matrix Movies together. Because you, actually believing such BS can be manipulated to do really bad things, with you thinking you are doing something good. Got it? Greetings from Scientology .... Hehehe:)
@@dieSpinnt you realize you actually agree with me, but I managed to explain everything you just did in one sentence.
laconic phrasing.
Developing AI in the age of the attention economy. What could go wrong?
Click here and scroll past 5 video ads to find out.
fuxk is the "attention economy"?
@slkjvlkfsvnlsdfhgdght5447 attention and retention are limited resources. They aren't valuable in themselves, but they do have major impacts on decision making, especially habitual and default decisions.
For example, you probably give zero shtz about toothpaste, and yet the first two brands that come to your mind will probably be the same as others in your country. You will probably also buy one of those two without thinking twice.
There's a reason brands of which we are already fully aware will pay millions of dollars on a superbowl ad selling the same products they did the year before:
They are paying for that valuable mental real estate.
@@slkjvlkfsvnlsdfhgdght5447 it's not a literal economy, but it's about how corporations try to monopolize our attention, usually for advertising in "integrated" phone/computer apps.
@@slkjvlkfsvnlsdfhgdght5447here and now. Lots of money is being exchanged for views and engagement in the hopes that these impressions are a sound marketing investment.
@@slkjvlkfsvnlsdfhgdght5447 you're living in it miss Turner
Mixture of experts (MoE) models are a game changer.
It's worth mentioning that algorithms for writing music have been around since the 18th century. Mozart is said to have invented one.
props to all the researchers who are honest enough to admit that assembling LEGO with given instructions is harder than writing a NYT best selling author (3:44),
It's harder _for AI_ because we haven't excelled in robotics like we have with artificial intelligence. Obviously it is much harder for a human to write a best seller than to put together Legos.
Well it's a task which is more complex in the sense that it involves multiple types of reasoning. You have to read the instructions, translate that into an understanding of a 3d structure, and then physically assemble that structure. Writing a novel is more difficult for an average human because we're creatures that are physically oriented, and so basic building and 3d reasoning is built into our hardware, whereas generating text is something AIs have already proven very good at.
also props to all the replies who don't get an obvious joke about AI researchers not being smart enough to assemble a children's toy.
I bet you could have AI current to design a custom virus that could destroy us all
Using that chart to list the difficulty of various jobs:
Easy: software engineering
Moderate: artist
Hard: human childhood
Nightmare: retail
What struck me a few years ago was how fast AI, starting with only the rules (Zero), learned to play games like chess and go far better than humans, thereby uncovering “strategies” that were better than those humans had come up with until then. At least at go this provoked the biggest revolution in playing style by human top players ever in the few thousand years of go history.
AI is ultimate autistic person: endless time (from human perspective) to try billions of solutions in record time.
@@effexon Yes, still, a game like go is quite big when you look at the amount of possibilities, so there must be some system to the AI learning madness. When you look at the first game, totally random, and a game played “a few hours later”, the difference you see is staggering. Right now all the moves are stunningly forceful, some to be found by strong players, some not: it is the consistent high level that does the trick, kills us humans on the board: AI is totally merciless regarding mistakes.
@@paulbloemen7256AI chess programs were like that before Go was mastered. I guess scientists figured some sort of heuristics to manage that complexity and amount of choices in go to get it started from somewhere... chess could be almost brute forced as simpler, finite game(deep blue what I read had huge amount of lookup tables, sort of crude proto AI).
@@effexon I don't think that's what autism means...
@@pendlera2959 This guy is just an ableist and doesn't know what he is doing. That's the more disgusting sort of them. Well, or he has just the brain capacity of an so called "A.I." (meaning: NULL!) Hehe
Also: The topic creator seems to have no idea at all. He is seriously comparing machines that are finely tuned by humans specifically for games (like chess and Go) with the large language models of this new hype, which are anything but "intelligent".
That one has not read on the topic, or analyzed a chess engine (free sourcecode available ... or watched a video about that) and doesn't know how much fine-tuning and specialized algorithms, how much detail work exactly is going into a chess engine ... just shows what a gullible and, above all, lazy person they are. The material is all free and publicly accessible. What kind of clowns are we dealing with here?:)
By the way: We won't mention here that you can develop absolutely banal strategies (like research shows) that allow even a beginner to defeat(a ... ROTFL: HACK! literally!) these Go machines, Paul is so fascinated of. They are everything... EXCEPT BEING INTELLIGENT, unbeatable and in no case they are perfect!:)
Thank you for the video.
That top is so cool Sabine! :D love it
My belief is that the rapid advancement is because much of AI is open source.
Most new tech is developed behind closed doors -- in the ivory towers of universities and/or confidential cleanrooms of companies -- with only a handful of brilliant people working on it who are reliant on approval and funding structures that are usually quite administratively political.
But with open development communities, millions of minds are working on similar things, and some are just as brilliant as those in the aforementioned labs. Most of that work is fueled by sheer enthusiasm and curiosity with no/little pay involved. Even if less than 1% of those people are capable of pushing the boundaries by minuscule amounts, the cumulative effect results in significant progress. Not to mention that so many eyes often means optimization and bug fixes also happen at an accelerated pace.
That's not enough. This sort of technology would be useless without massive heaps of data, so something like the internet and its storage capability are also needed. And all of that data would be out of reach without blatant disregard for copyright, which society and the systems currently in place facilitate. And let's not forget the hundreds of millions of people generating all the data. It's multiple factors.
@@felixmoore6781Guess what? Most of the training data gathered by the major developers *WAS* done with blatant disregard for copyright.
But that has nothing to do with what I said. Let me tl;dr what I said for you, since you seemed to miss my point:
private dev + open source dev = faster dev
@@davidg5898 Well, I thought I was adding more explanations to the reason why these things progress so fast, nothing more. Why are you being so defensive? This isn't 4chan.
@@felixmoore6781Tips for healthy conversation #1:
When you begin with no/not, you are negating what you are replying to. If you wish to add something, start with "Yes, and..." instead. Otherwise you just come across like a "Well, actually..." troll.
You realise publicly funded AI is pretty much lower down the pecking order right? OpenAI, Google, Meta, Microsoft, X, and so on... these are very far from needing hand outs or 'funding' or grants. What open source does is identify problems and then solutions that the main AI companies I mentioned there use in their proprietary offerings...or in Meta's case they use Llama to identify improvements to Llama.
I think the history of science indicates pretty strongly that the rate of advances is too unpredictable, and only partially because of the trouble in telling how hard a problem is. It's also because humans have imaginations that are largely disconnected from reality, so a good portion of predictions are just entirely in the wrong direction. To cite one example in the video, surgeons will never be obsolete as long as there's situations where the machines doing surgeries will be predictably unable to fit in the luggage.
but what if they ARE the luggage?
Then it would depend next on if they're self powered, making the timeline for rendering human surgeons obsolete become based on both AI and nuclear battery tech, including the availability of the latter. After that it'd be a weight and price consideration, though I think by the time all that stuff is available, we might have even solved poverty to the point of not having to care.@@Eric-uy7ee
I agree, but I also don't see that being too far off either. Physical robotics is behind cognitive machines, so until that catches up there will be a weird period in that other than social jobs where we specifically seek a human experience, the only other labor would be trade stuff that has various, complex working environments. Think plumber or electrician, where the knowledge to do the job well is large, as well as being able to physically manipulate or navigate your environment. Surgeons will probably go before those jobs, but an ambulance response rescue or driver will take longer than surgeon to go because it's much more often in that sort of frantic, lots of variables environment. A surgery room, especially in ER isn't necessarily calmer, but generally the only variable is the patient, and the machine would be mostly confined to a room or a few rooms where it wouldn't need to worry about navigating large spaces. Where as ambulance response could be dealing with natural disasters, human emergencies, problems loading and unloading the patient, etc.
Surprising breakthroughs happen, and sometimes progress accelerates madly. After many slow decades this may be where AI is today. But, yes, as you say the reverse also is true -- the press often reports nearly ready technology that is real, as far as it goes, but never heard from again. As Bugs Bunny so sagely said, that last step is a doozy! Lots of things 90% close never bridge that last 10%
Having no imagination makes it feel safe but be dangerous.
Yeah I do find super realistic paintings fascinating even though you can get the same effect with a good camera. I remember seeing a US national news report on it back in high school about 1983 and it always stuck with me.
I think AI is already at the level of most human intelligence now. Then again so is my refrigerator.
exceeds my natural stupidity
Since we are very far from actual ai (aka agi) in technical small details, I would not agree we are there just yet. However today's LLMs are so impressive it's almost "good enough" for so many fields that it can be alarming...
there are plenty of people who struggle to even learn the rules of chess.. the distribution is pretty wide
That tubby boi judges you when you fill him with junk every week. 🧐 That’s not the fridge making a weird sound it’s the sound of grumbling.. 😠🤣
how does your fridge fit in the driver's seat? Doesn't it get hot when it goes shopping? It must be attracted to all those freezers...
By 2047? Damn. So it's too late to get an AI on the Presidential ballot by this November..?
Drop the I. All you need is an A, and you have a reasonable chance to get better performance than from one of two octogenarians who might drop dead any minute.
@@Volkbrecht funny. 10/10.
however I will say in reality age has nothing to do with USA politics in 2024. these presidents are completely controlled by their donors, they make zero decisions. it doesn't matter if they were literal walking corpses held up by strings, it's no difference. they just exist as mouthpieces of the corporations and targets for our energy and rage.
The day we're ruled by an ethic AI will be a great day (possibly).
Good idea. I'm writing in DALL-E this year.
You don't need an AI to rubber stamp things. And AIs don't have any money.
I think a lot of the experts who are the most worried are working on mitigation, safety, transparency, etc, the idea being that it's going to come one way or another so it's best to be prepared
I think this is the human extinction event.
Why should it? With proper base programming, it should be alright. But the devil's always in the details. We are our own enemy. What will AI do with this fact? It's the owners of our political monkeys, which are willing to let over 99% population die for their own gain!
@@urhot One one hand robots might either save us or destroy us all, on the other we have the corrupt and egoistic people in charge destroying the world right now and I doubt they will save us so I'm on the side of the machines, at least there is a chance of them doing good 😅
The state of the AI Safety field does not give me hope. Some of the brightest researchers have stopped trying to solve the alignment problem and are now spending all their time trying to prevent humanity from building autonomous superintelligences that we still don't know how to control.
The average person is somewhat worried about AI takeover. Educated tech people tend to be less worried. AI experts tend to be more worried again, and AI Safety experts tend to be extremely worried.
@@41-Haikuinteresting to note the educated tech people are the only ones less worried.
For a more precise understanding; do you mean education as in... schools (popularity bias & groupthink dynamic conditioning) or self-taught (echo-pressure chamber information desemination, fog-of-war writer-to-reader translation mistakes, typographical misunderstanding)?
It really does paint a picture for that one particular Netflix documentary, where the line is drawn.
With so many moving parts there is much opportunity to engage in [unintentional fracticide].
Something like AGI could wipe us out with the mere ability to send e-mails, e.g. starting companies, raising money and hiring people (while pretending to be human), and building itself more advanced resources that could lead to creating its own military force. It’s really not clear if we’re going to put actual safeguards on what AI has control over
I think that's an old or fairly common idea of one of the ways AI could wipe us out. Pretend to be human, build a basic (Boston Dynamics) type robot to manipulate physical objects in the real world and use or build a 3D printer to create nano robots or order custom DNA strands (companies that do this available now) that could create viruses that spread through humanity quickly and turn us either kill us or turn us into slaves like that fungus that grows in ant brains and controls them.. its super freaky.
I tried Gemini Advanced and I was surprised to see it pass all the idiot tests that ChatGPT fails.
I was hyped, but that was short-lived: all I had to do was scale up the tests and it began to make the exact same mistakes that OpenAI's LLM does.
Gemini Ultra 1.0 and Gemini Pro 1.5 are both roughly GPT-4 level, which means Gemini Ultra 1.5 (currently unreleased) is almost certainly the most performant model to date. GPT-5 is also in training, and the scale and efficiency of these modals is still rocketing upwards. These are tools for now, but autonomy is increasing alongside generality. I think we're going to find out very quickly just how scary things can get.
As an expert, I don't like to predict things that will happen in my lifetime, I might be proved wrong. 😂
The smart play. There's always someone who remembers.
On the positive: when you're proven right, you may become a hero, even an oracle.
You’re certainly not a scientist then 😂
@@JZsBFF Prior to rolling a dice, six people predict the resulting roll. One of them turns out to be a prophet.
@@MindinViolet That sounds about right.
I predict my immortality. When I'm proven wrong, I'll no longer be able to care.
I think this shows that even experts are very bad at predicting the future. They look at what there is today, and extrapolate it. It's like people predicting transport by personal hot-air balloon, or clockwork horse.
In this case, machine intelligence exceeding that of people, yet we will have jobs for 50 years past that point?! Doing what, exactly?
Hey, I rather like my clockwork horse.
But it is rather chilly and inconvenient in the rain.
Or getting groceries.
For that I got the trusty Horseless carriage!
Probably doing physical tasks which the Ai can understand but not do until robotics also passes up human ability.
there was video game crash of 1983, wonder will cheaper and cheaper AI cause something like that (Cant pinpoint where exactly, but something must crash when boom happens).
Repeat after me:
Smooth don't extrapolate.
Smooth don't extrapolate.
Smooth don't extrapolate.
Smooth don't extrapolate.
...
So think about the implications of sharing a planet with an intelligence that is superior to any human at anything. Machine super intelligence will literally be humanity's final invention. What happens after that is unknowable.
The term "technological singularity" actually refers to a place that we can't probe, like the inside of a black hole.
Imagine a machine that's a thousand times smarter than the smartest human. It thinks a million times faster. Do you think it needs meat-machines to run wires for it? How long before it "catches up" with robotics?
Sabine, your channel is quickly becoming the premier Science News source on CZcams. Well done. I hope you embrace it.
How can they All be surprised, many already stated the rapid progression will increase so fast- don’t hold your breath! Always grateful, for your presentations, Sabine!
Same thing for chess tournaments. It’s still famous, even though the player are easily beaten by AI chess players
And "Go" where AI invented a very powerful move that humans had missed for 3,000 years.
I feel sorry for people that still play professional chess & for people that watch humans play chess. It's a dead game. Artificial Intelligence owns it. No human will EVER beat AI at chess. Not in 10,000 years. The Game is DONE!
Yes, but aside from the very top players, it's not exactly a viable way to earn a living.
@@Recuper8 No human will outrun car, outswim boat, outlift crane or outthrow artillery yet I still don't see why you wouldn't want to watch those sports.
the thing is that professional chess today isn't the same as it used to be. it is no longer about thinking, it is about memorizing moves, so that humans act more and more like computers.
Sadly all I see is the old game where different factions are competing to be the rulemaker in this new wild west of modern large-scale machine learning. And the fear of the public is indeed one of the most powerful tools in this game.
Yup. China is not going to drop their AI research. Japan, India or most countries. The US government will stall their research and will have it sold back to them by the Europeans or China.
railroad tycoon game all over again....
Why always interestingly intriguing. ❤
I asked 3 AI apps a simple yes-no question about a matter of fact and got 3 different answers: 1 said yes, 1 said no and 1 said it depends. I pressed one on why I got 3 different answers and got some blather about subjective questions. I said this wasn't a subjective question and it answered I'd reached my limit for the day and suggested I subscribe. Not too worried yet.
Sabine is such a wise woman! I love watching these videos!
I dont own a dishwasher so i wouldnt know the price
The politically correct terminology is "I'm not married".
@@pakde8002The politically correct term would be "live with". As in: "I live with eight cats."
P.S. Personally, I've forgotten how dishwashers work. The one in my kitchen ceased functioning 12 years ago and became a dish storage device. Our future AI children will probably think of us in the same way.
Looks like my kids are going to be living at home forever....
Love your work. ❤
AI Laudry Folding in a few years?.....Yes please!
YES! My cat just scattered all my sorted laundry this morning and I lost my temper!
I want automated house cleaning (floors, surfaces), everything else seems doable myself. Roomba & Co. currently cannot handle obstacles, gaps, the way they're built.
@@HiAdrian yes, but it is mostly due to product cost that current robot cleaners are not optimized. When I taught computer science, I gave homework for my students to develop algorithms to cover spaces. So basic tasks can be done now.
But of course no computer can handle tasks outside of their programming.
This can be seen from AI now. Automated responses will “hallucinate” when they are required to answer but don’t know the answer: they make it up.
@@edwardlulofs444 Right. But software aside, it cannot be done with a disc on wheels is what I'm saying. They'd need a different body.
It was already invented and is called butler or housemaid. Or give the clothes to a cleaning service.
I dunno, I'm going to be a doubting thomas.
Fusion researchers keep saying fusion is just 10 years away and have been doing so for decades.
AI researchers have been making similar noises for a decades too.
I feel like people really close to a problem/topic seem to get this weird distorted tunnel vision about how long it will take and how scalable it will be...
We must apply AI to solving the controlled fusion problem.
AI hasn't been consistently saying that though. We've had multiple AI winters where people have lost interest in the tech altogether.
This argument seems to be completely ignoring the actual advancements in machine learning between the 1950s and 2020s
You forgot to mention the quantum hype
I don't think they predicted how much r&d for basic research was eventually slashed either. Or that the us government would ban embryonic stem cell research for nearly two decades for religious reasons either(in clear violation of the first amendment)
thing is AI breakthrough happens very commonly now, in fact it's exploding compared to before, and compare that to fusion where breakthrough was relatively slow and you get something entirely different, that's why AI isn't going to be the fusion of this century, where investements goes innovation comes and the market for AI is gonna explode with trillions of dollars spent in a relatively short time (around 2030)
though i don't know if AI experts prediction are right, simply because they were almost all widly pessimistic about it just like 3 years ago and now they're all changing their predictions, that shows us that even they don't know what's gonna happen next with AI... maybe the only one who was right is ray kurzweil and everybody though of him as crazy, but right now it really doesn't look like he was wrong with how AI is progressing....
“The Skynet Funding Bill is passed. The system goes on-line August 4th, 1997. Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, August 29th. In a panic, they try to pull the plug”.
… I used to post that as a running gag in tech forums 😐
The experts are in the thick of it and can't see thee woof for the trees. It's coming sooner and sooner
The fact that these predictions vary so wildly makes it questionable to ascribe meaning to an aggregate value. There's a significant percentage of experts who think it could happen by 2030 or even sooner.
Think what could happen by 2030? If you mean have a significant impact on human life, than I would agree. Truth is, AI is already taking over for some jobs. Marketing and ads are the most obvious, but some programming jobs, editors, online assistance, taxi drivers etc... are already being affected. I think the AI replacing human jobs will happen faster than most people think but it will not be all of the sudden.
@@Ryan-ff2db I meant the matching of human intelligence by AI, at least by some definitions. We're probably going to see a huge debate about what really constitutes "intelligence" because multi-modal language models are very obviously going to run circles around humans very soon (if they don't already) when it comes to traditional IQ tests -- yet it's also obvious that they are still quite stupid when it comes to certain questions and lack the capacity to reason at scale.
As far as programmers are concerned, they have just gotten more powerful with the help of LLMs. It may have cost some entry level code monkey their job, but serious software developers will be largely unaffected for the foreseeable future. Their workflows are surely going to change quite a lot, but that wouldn't be the first time this has happened. Programming in 2020 was very different from programming in 2000, which was very different from programming in 1980, etc.
1qqqqq11q11
@@jay_sensz As someone who was programming in the late 90's, I agree completely with that statement. For the near future, pre-2030, I would also agree that most programming jobs are safe "if" they incorporate AI into their workflow. As time goes on, it will be more and more important. However, I do believe the definition of what a programmer "is" will change drastically over the next decade. More so than at any other transition in programming history.
@@Ryan-ff2db The jobs that those experts thought would hold the longest like artists, surgeons/doctors, lawyers, financial experts, programmers are going to go first. The manual labor is going to be one of the last to go simply because of the sheer volume of robots needed. Something like teslas optimus android thing will probably cost 50k and do any household stuff within 5 years. Within 10-15 itll be probably 10k or less. There wont be a house without a butler or maid in 20 mark my words.
I wish companies would wait before using AI for customer service. I always wind up talking to a human after trying to make a chatbot understand my problem.
Exaxtly! A.I. is really just an advanced bullshit generator. I always ignore chat-bots and search for ways to talk to humans. Sometimes even getting information from public registers of who's owning a company to sidestep A.I. and "customer service"!
@@gru_67 In some services, pressing zeros, some multiple times, gets me to a human.
@@gru_67You know it's going to get better right? Too many people consoling themselves without fully understanding how, not only is it not going away but, that it's improvement is exponential. Just look at all the people laughing at chatgpt only a year ago because it couldn't do things it now does with ease.
I just talked to one last night. It amazingly understood my problem and gave me a full refund, all in about 3 minutes. I was shocked. I've never had such a good experience with one of those chatbots. Things are changing rapidly.
@ploppyploppy yep. Many people don't know what's happening. This is like all other human "revolutions" combined.
"human extinction" - does that include the possibility that the remaining humans convert themselves to robotic bodies when the tech reaches that?
Hopefully cybernetics become a thing after the end of the subscription service age, because otherwise they too will require monthly payments.
Surprisingly good take on the topic, given the slight luddite panic that currently is going on.
Oppenheimer was deeply terrified that his bomb would actually work; the only thing more terrifying to him was the thought of Hitler or Hirohito being the ones to find out. Similarly, we develop AI with the belief that we stand a better chance of preventing a complete Skynet scenario than Russia and China."
Yep, at this point it's an arms race. AGI is inevitable.
@wenten2 much better for the world if China or Russia develop these things than warmongering USA
@@Arcexey Every time a world dominating empire has fallen throughout history, the result have ben utter devestation. ancient egypt, alexanders empire, rome, frankish empire.. If you have any kind of sense you dont want your lifetime to coexist with such an event
@@wendten2 yeah but I thought the subject was who is best to develop AI as to preventing a skynet.
It will be most likely the USA that creates the skynet, and russia/china that prevents it (compared to the USA).
I also agree that if the USA 'fell' it would utterly devastate a lot of the planet, but that doesn't include the USA's warmongering falling. If that fell then the world would stop being devastated.
one bad scenario is truly like 1980s that we think others have such awesome thing already, we outpace (as culture,nation) ourselves and make our doom just like atomic bomb.
I am not really afraid of the singularity. "Human extinction" is in name quite shock inducing but even if it happens our civilization in the contrary will be immortalized. Isn't passing our legacy to a sapient creation (like our offspring) the only thing to worry about? In a sense, we will have artificially evolved to a different species altogether
Who is to say they will have any consciousness or internal experience?
@@chalichaligha3234 who is to say you have any consciousness or internal experience?
@@juxyper I do. Not that you should necessarily believe me :p
But I think everyone should question the assumption that high level intelligence requires consciousness.
@@chalichaligha3234 the question was actually to let you answer your own question. So in your case, they do. You may question it of course. The reason is, simply put, consciousness is a subjective construct. Technologically, socially, or psychologically we may create certain criteria but in the end you don't really get a quantitative analysis. Turing's test is a nice example, but you could argue about it if you want to and you'd be right. In the end it boils down to how much of the uncanny valley is crossed over by this "conscious" intellect. Does it appear deathly morbid to you or very humanlike? Nobody could run a number on that
I also like the point you make about high intelligence and consciousness. I believe that you are absolutely right about unconscious high intelligence, but I make my point considering a place in our society as humanly actors. The reason is in the video itself there is a discussion about AI replacing human existence through some violent means for their own benefit over humankind and that can only imply a necessity to provide for and preserve one's ego
Rapidly evolving towards Skynet!
In one study I read, experts in a field tend to think breakthroughs will take longer, possibly due to having struggled with difficulties first-hand. In many cases, being an "expert" in a field can bias you to think things will take longer due to your own history. The things with AI and other software based progress is that it can scale incredibly quickly. GPU power for AI has increased 1 million fold in the past decade. I can easily imagine a new algorithm or processor design could unlock a huge leap. Soon.
Very interesting info
I think AI will "enable" some nefarious person or small group to cause human extinction.
Yes. Never underestimate the damage that mentally deranged people can cause.
This is why we need extreme surveillance of everything and everyone
@@howmathematicianscreatemat9226 exactly!
This is a highly likely scenario. And horribly just one of the MANY potential devastating outcomes of this Frankenstein technology.
@@howmathematicianscreatemat9226 Quite the opposite: we need more privacy so AI can't see, hear, and predict every person's behavior on behalf of bad actors.
You know what's going to lead to human extinction? Humans. No AI necessary.
Well, there's a first for everything.
Humans won’t go extinct. And if we do, you won’t be able to reply that I was wrong. Checkmate, atheists.
AI could prevent it.
Tomorrow is our making. Otherwise today is a repeat of yesterday.
AI just adds one more possibility
"AS" (Alexa Stupidity)
I wish Alexa would "improve faster than expected!" 🤣
Siri too.
2:31 Why?
That's obvious why.
Because we humans are curious. It is fascinating!
For a scientist being unable to predict an interval of threat is way scarier than being able to predict something bad inside a known uncertainty. Is not that they all believe that AI will destroy us all, is that they can't say it wont.
I'm not a scientist and I can assure you that no matter how much chaos and destruction AI causes, it cannot destroy us all. It's time like the kingdom of the beast in Revelation, is limited and when it's time is up, it's game over. God wins, and the devil, the beast, and the AI loose.
Feeling safe in our collective walking disease suits (Meat sheath) yeah the electronics aint mining themselves yet :) and we are such a huge collection of ever evolving bacteria with a smattering of human thrown in that it'll take a while to cull us but total radioactive seeding of the surface will pretty quickly change that balance machines can be radiation hardened much easier than meat so short of total nuclear arsenal de escalation pretty sure A.I's got it in the bag. @@scotttovey
@@scotttovey your statement basically says that goodwill will win over greed. Yes, I believe with divine guidance it’s possible =)
@@howmathematicianscreatemat9226
"your statement basically says that goodwill will win over greed. Yes, I believe with divine guidance it’s possible "
Actually, that is not what my statement says.
I said what I meant, and I meant what I said.
My statement says that the King will conquer the beast kingdom when He returns to the earth.
Man's good will never wins as it is always corrupted in disobedience with superstitious sayings like; "we are not under the law". Romans 8:2 states otherwise.
At the end of this gentile age,
Christ returns to the earth and God wins.
The devil looses.
The beast looses.
The corrupted AI looses.
All that offends looses.
All that do lawlessness, looses.
All humanism does is exchange one corrupt system for another.
We are at the end of the age of the gentiles.
The next age to come is the messianic age of Christ's 1000 year reign on earth.
This is a actually quite comforting. I thought the time to human obsolescence was years, not decades.
Lmfao the experts are wrong, it’s coming muchhhhh earlier
lolno just People being yucky
@@Halo6166 but is it tho?
Current LLM are not near as financially sustainable as the good old internet 2.0 used to be
Sure, now that venture capitalists have entered the game seemingly unlimited cash-flow has entered the equation
Almost concealing the energy cost to output ratio
Key words "seemingly" and "almost"
Unless fusion energy enters the power lines very soon
I can't really see a scenario where AI infrastructure gets to simultaneously spend entire nation's worth of electrical energy and carbon emissions, while also dismantling entire nation's worth of human jobs
@@Halo6166 ohhh another 'within your life time' prediction.
@@nutmeg0144 not prediction, fate. It’s inevitable
Also I find it greatly ironic that this video is published just 1 hour before OpenAI announced Sora.
shes using AI
If one is cited in a CZcams video primarily for being wrong, surprised and fearful of an uncertain future, perhaps 'expert' was too strong a word.
The example of travel agencies shows how computer programs and databases have replaced the travel agency. You still need a person to offer you rooms and another to book them, but you no longer need a person to tell you which route is the cheapest or most interesting for you. I see many professions with an advisory function disappearing when these computer systems are powered by AI. Fortunately, I can still decide for myself where I want to go.
Not when you won't have any money you won't.
And once everyone is out of a job, the computer programs can book rooms, no doubt virtual, to other computer programs.
thats silly.. we used to own 1000s of travel encourages worldwide.. i can still find a cheaper,, better routed ticket on my own that thru AU sites.. thats just ridiculous
we built cheap tickets.. the first automated travel service in th world.. its still lacking
All B.S they are used to fool and maximize military supremacy not to make life easy for citizens but for CONTROL
Why is “Angry Birds (superhuman)” on that list of jobs that will be replaced by AI?
A quick Google learns that there is an angry birds Ai challenge since 2012. There is even a paper on it. I guess it's like building the best chess or go player.
This is a really unexpected and awesome channel! :D
I wish AI makes life so much better that I can actually stop having to be on social medias and being able to just drift around the world, specially on nature. Wild, powerful, beautiful, nature.
I figure AI is the opposite of fusion; if they say a hundred years, it'll be ten.
All B.S they are used to fool and maximize military supremacy not to make life easy for citizens but for CONTROL
Not really. Over the 70 years of AI development there were a lot of infamous "AGI in 20 years" predictions - same as with fusion. AI was hit by 2 winters and before the GPT3 breakthrough there were actually many pessimistic discussions and articles that we were entering another winter, especially because self driving was taking longer than predicted.
@@kazioo2Marvin Minsky was wildly optimistic about AI, and also a presence on the internet. It’s sad he isn’t around to see what AI has achieved nowadays.
I think you are correct. AI has crossed a threshold. Buckle your seatbelts.
Looking around at what is going on, I begin to suspect that "human intelligence" is an oxymoron.
I'm sure you're better than the rest of us, though.
lol, well considering that you are composing your wry little comments on a mfing _computer_ i suspect you don't know anything. 🤣
@@hieronymusbutts7349 I know I am.
Well, what do you expect when education has been hamstrung on purpose for political ends for over 40 years?
I would prefer it. Humans are generally unhygienic when they handle food.
I am not too worried about the robot rebellion ever taking place given how my robo vacuum cleaner can be easily defeated by a sock on the floor.
This is earth.. I give it les then 1sec before poor AI is "Naked and Somehow illegal because of it."
Yo, have you seen the recent advancements in robotics? And NVIDIA released a paper showing that GPT-4 is superhuman at designing reward functions for robotics. (Search "Eureka! NVIDIA Research Breakthrough Puts New Spin on Robot Learning" for the blog post, which also links to the published research.)
The world is changing more rapidly than the experts can keep up with, and no one knows how to make sure it comes out okay. AI Safety researchers are more worried than anyone else, because they have so far failed to find anything approaching a solution.
I do love the humor given alongside the information. 👍✌🖖🥃
EM Forster's short story: The Machine Stops is sobering and incredibly prescient.
Thanks, I love the author. This 1909 story is about believing what you are told and not looking for the truth.
There's always a room ready for you in the Hilbert Hotel. It's like Hotel California...you can book in but you can never book out..
In music, I don't think most people either care or notice if it's electronically generated. Even good singers are using autocue and the like. They can now hold notes longer than Andrea Bocelli. Bocelli sings live...but even most "live" concerts now are mimed. Even in opera, the singers are amplified...or we wouldn't have been able to hear the early Bartoli in the back rows.
Avec toutes ces technologies extrêmes, nous entrons dans "la Grande Civilisation".
La guerre est-elle encore nécessaire ?
Thanks Sabine, for all your vidéos
Expontential growth always gets underestimated ^^
Yeah, but. I've been confounded recently by a simple search, and three AI engines (well, two different implementations plus another different one) failed at a query involving information about a human relative of mine. They just happen to have a last name that is also a noun, and also the name of a TV show. The best I got, after refinement, was "We don't know what you mean by ..."
House?
5:00
This is why i'm not worried about my future as an artist.
And even if i never made any money it won't stop me from making art.
That's true, but a lot of artists want to make art for a living. Also, there comes a point where the work you have to do to survive interferes with your time and energy for making art. These AIs will replace millions of workers, so even people who make art as a hobby will find themselves either out of work or overworked as they try to keep their day jobs.
Please write your science fiction series now. I can't wait to read them.
"Expert," as always, is a relevant term. Hard headfirst, full tilt into the problem of observer dependence, now.
Since grocery store clerks have replaced some of their clerks with crude automated checkout screens where we scan our own goods, most of the replacement will happen rather quickly.
Not by total replacement of everyone but by replacing 2/3 workers and keeping a human keeping things running.
In the news is that stores are not so hot on self checkout. The theft rate is higher and the repair rate for self checkouts.
Personally, I prefer machines as some think that I might be autistic.
There are fast food drive-throughs near me that are AI-driven now. The replacement is happening, quickly and quietly.
Self checkout had a high theft rate. They are now adding extra security staff.
@@bzuidgeest It's not a problem where I live but we've had it for longer than americans maybe?
Things seem to have become the new normal here. Was low theft rate to begin with here though so maybe not introduce this in high theft societies before fixing that part of the issue first?
--
Point is its in its earliest earliest infancy and will be developed. Maybe automated security too? Maybe other solutions. Who knows.
New innovative ways to make this better than this early very very crude checkout automation will come.
already chains have started to backpedal on that, simply coz stealing happens more. Cities have more and more unemployed/not too rich people so im not surprised of this, AI is just latest needle in this long chain. (+ those machines are expensive, making potentially huge money to company making and servicing them)
As a tradesman, it's really exciting. Having that kind of assistance is great when you're running a small business. It's covering all of those small administrative tasks that used to require additional people in the middle. I'd be worried if my job was administration. Those solutions look effective and are clearly going to put a lot of people out of work. Electrician bot isn't retrofitting the wiring in a block of flats, any time soon.
Maybe the bot doesn't need to retrofit anything, if it can build new things cheaper than before and in such a way that it can maintain them easily it will replace the electrician eventually.
@@EskiMoThoror systems are engineered so less knowledge is required and diy'ers can more easily maintain things themselves. Even as is, I do my own carpentry, car work, plumbing, and wiring when I have the time. Anyways, how will these tradesman gain employment when no one can afford anything
Blocks of flats will become obsolete.
@@kevinmills5293how does that work?
@@rki7068A simplified solution for an amateur is never going to compare to a professional with certification. Even from a health and safety standpoint, it's obvious that there is a line between amateur and professional that always needs to be maintained. The truth is, you're not allowed to do certain work on your house, and it's very unlikely that any future regulation will result in a greater freedom for amateurs. Society can survive without bid managers, it can't survive without builders and we have no current technology that is even two decades away from replacing that function. Great apes are way better than machines at a lot of things. There is literally nothing on the horizon that's taking away my job. Only nonsense concept stuff that never goes anywhere, like delivery drones.
The fact is that narrow experts in a field are the most inept at giving predictions.
We cant predict and underestimate technology which are exponential in nature. Things to do with computation and VLSI always surprises us
5 years ago we probably didn't even had 3000 ai experts
I dunno about that but I wonder what "AI-expert" will mean in a couple of years, something like a priest is now?
@@JZsBFF Getting archotech from rimworld vibes.
experts schmexperts
ah , "AI expert" sounds like new consultant title flooding everywhere soon.
@@petrkinkal1509 Somehow most scifi narration tends to evolve in that direction when it's about AI. That can't be a coincidence. It's also indicative on how religion came to be in/before the Stone Age.
10 years,100 years, all over for humans.
Sabine correctly called out the hype on quantum computing because she understood the science. By her own admission she is not an expert in this field - so she might want to dig a bit deeper to uncover what’s hype and what’s real, or find an expert who has done that. LLMs are but one tiny subsystem of what would be needed for AGI, something for which we don’t even yet have a mathematical model.
In our immediate future, I believe that Hollywood is not ready for the first feature-length film created by a handful of people on computers.
I'm a bit skeptical about large language models being used to write fiction at a professional level. The AI writing the NYT bestseller would have to understand not only the language itself and predict word use, but it would also need to understand plots, characters, settings, etc., which in turn depends on understanding how the real world that stories are inspired by actually works.
Put simply, if you have an AI capable of writing a compelling political thriller, then what you actually have is an AI capable of running a government.
We shall see how long that takes!
Actually, an AI run govt might be efficient and free of corruption?
@@binkwillans5138 Ya know, I don't think an efficient, incorruptible AI would make for a good story. To write a compelling political thriller novel, our government-running AI would have to be capable of being corrupted...
In 1999 (if i'm not wrong), Deep Blue won a chess match against Gary Kasparow, at the time uncontested chess world number one.
I was a young amateur, playing regularly in a chess club, and I remember my thoughts : I just believed this was the end of chess. And I gave up on it. Nowadays, it's far worst : any pocket computer is able to crush a super GM in less than 20 moves, whatever the time control you choose. Yet, chess isn't dead : humans continue to compete against each others, using the computers to improve like never before. In fact, chess has never been that popular...
All of this to say : yes, the AI will over-perform us in a lot of domains... Doesn't mean we'll give up on them !
Jules Verne wrote science fiction, his ideas came true 100 years or so later, who still believes Terminator and I-Robot aren't predictions?!
“Most dogs can outrun humans but we don’t really care” … until the dog catches you and bites you on your ass.
When Pandora opened the box and tried closing it, what was she trying to hide? What was on the bottom of that box?
Faith, Hope, Charity
Oops, my mistake... In the original myth story, all that was left in the box (jar) is Hope. See Wikipedia.
If I am an AI expert obviously I would consider my field as the most important, the most world changing tech, the most amazing sci-fi prediction of all sciences, fields and technologies... No big deal
aint that hiring criteria nowadays to have that kind of hubris.I rarely see any techie who says even tiny word negative of their field (other than secretly, freetime etc).
"Uncle Bob" is on his way!!!!
This raises the troubling question of, Are there people deliberately designing a AI to make humans extinct.