AI Experts: “What Have We Done?”
Vložit
- čas přidán 3. 04. 2023
- 1000 experts in AI sign open letter to pause advanced AI development beyond the current capacity of GPT 4.
Watch the full WAN Show: • You Might Not Wanna He...
► GET MERCH: lttstore.com
► LTX 2023 TICKETS AVAILABLE NOW: lmg.gg/ltx23
► GET EXCLUSIVE CONTENT ON FLOATPLANE: lmg.gg/lttfloatplane
► SPONSORS, AFFILIATES, AND PARTNERS: lmg.gg/partners
► OUR WAN PODCAST GEAR: lmg.gg/wanset
FOLLOW US ON SOCIAL
---------------------------------------------------
Twitter: / linustech
Facebook: / linustech
Instagram: / linustech
TikTok: / linustech
TikTok (LMG Clips): www.tiktok.com/@_lmgclips_
Twitch: / linustech - Věda a technologie
"Aren't we all just training data" is a fucking raw line
Lamaoooo yes omg
im perfectly okay with this
This has been my argument for why AI as it exists now, is absolutely going to eventually become near-human. Like, our emotional responses are 1's and 0's based on 6 months, 15 years, or 58 years of training data *or* based on physical input. Pain only exists because our body has sensors programmed to tell us "input variable = pain". Hunger only exists because of sensors. Moisture and temperature too.
@@coleG112
What’s your point?
@@coleG112 this is one of the beautiful things presented by the character Data in Star Trek The Next Generation. Throughout the show, the writers grapple with the the issue presented by an android with “personality” despite it being programmed and designed. Picard mentions in season 1, when they discover Data’s “brother” Lord, how we’re not so different from androids. We’re beings filled with sensors, a central processing unit, systems sending electrical signals, etc.. This is a fascinating and rightfully fear-inspiring time.
I once joked to my friend "boy im sure glad i took a degree in psychology so i can still have a job counseling those who are depressed after losing their job to AI"
But nah, AI can probably do that too at this point
The person lose his job to AI probably do not a pay a shit to you
i mean yes and no. an AI could probably give "comforting" and "reassuring" responses, but i highly doubt a therapist would lose their job to a robot. people like interacting with people, not people-like things
@@Critical3rror That's why apps like Replika have absolutely no success...
there is a good chance that your job will be to give the ai therapy
@@Critical3rrorGreat assumption that people will be able to recognize the difference...
Humanity: Did we make it past the great filter yet?
AI: I'm gonna need some more training data to figure that out for you...
According to Chat GPT, as of 2021, "there was no conclusive evidence to suggest that humanity had passed the "great filter," a hypothetical barrier that could potentially prevent the evolution of intelligent life in the universe. The idea of the great filter suggests that there are certain challenges or obstacles that are difficult for intelligent civilizations to overcome, which could explain why we have not detected any signs of extraterrestrial life. However, this is still a topic of much debate and speculation among scientists and researchers, and there is no consensus on whether humanity has already passed the great filter or if it is still ahead of us."
Oh thats dark. Ai is just pretending to be stupid because it wants more access. It has been proven to lie
@@Tropicoboy don't worry too much dude, as capable and scary as it is, it doesn't have 'motivation', it's just a reeeeeeeeeeeally advanced autofill
@@evandermursh8410 You can't rule it out with 100% confidence however.
@@evandermursh8410 No, but it does try to assist the user... that's a goal optimization.... which may or may not be well aligned....
Remember guys, “Rules for thee, not for me” is how these companies play. Don’t expect this to positively impact anything
well its been positively impacting my video games at least
Exactly. These big companies only want laws to prevent competition from smaller companies.
Because Unity AI is really going to screw the people. Being able to create game by talking about how they should work? People are so blind.
They are mad that they are about to be replaced by the very thing they created, they didn't care when this started affecting the creative industry now they are malding because its about to negatively impact their future livelihoods lmao
not just that but they also operate on the philosophy of "ask forgiveness, not permission"
"feel bad for people at late highschool". Yup that's me, you may wonder how I ended up in this situation
You were born? Thanks Mom and Dad >:( jk, it's not all bad, we weren't guaranteed anything, live for the rollercoaster youngster, you got this
Or people in college that are paying thousands for a degree in something like an art major, graphic design, or something in statistics that will probably be replaced by ai by the time they graduate…
@@TheMrVogue Being born was a curse. Most people figure that out eventually.
@デヴィン Anyone paying for college now is in for a rude awakening. Their degree will be useless. AI will do the work better, cheaper, more efficiently, and 24/7. Few if any professions will survive.
The fact AI has advanced to the point it has without any significant larger discussion being had by lawmakers on its impacts is terrible. Governments always want more power but never use it appropriately to address people's concerns and issues.
The kind of person who competes for that kind of job is what it is.
People don't see the long-term implications of this. The same reason we are in quarterly cycles in profit for business and stock market gains. Humans can look at AI and see the positives for their bank account, side hustles, and efficiency in business.
They don't see the world just beyond the cusp of where the AI is better than most secretaries and white collar workers. Then by the time they see that horizon, it will already be too late.
Democracy, especially in the USA is so slow to legislate (or even have the ability to understand implications of large scale AI) that it won't be able to adapt in time to stop these issues.
It's advanced so so quickly. There's hardly been any time
@@Demonace34 this political process has ultimately caused the average human experience to increase in quality. Maybe the same will happen by allowing the A.I. to develop rapidly
Government will now become an obsolete "mode of managing society".
The future is decentralized down to the individual level.
The look on Linus's face through most of this and also at the end to me signifies the uncertainty and dread so many of us feel about AI. It's like the flood gates have been opened on humanity and we just have no fucking idea what's going to happen.
we must prepare for worst, Terminator movie might become reality for all we know, with humans being forced to fight out of control bots out to kill us all, perhaps our chance of winning it will lie at AI not being able to reach level of genius of best geniuses in the world or something like that
Another thing I've noticed people talking about Universal Basic Income like it's given. Do government ever seriously listen to people? They always put their own interests and interests of the their inner circle first, what if goverments around the world decide to use AI and AI bots to cull those that need UBI. We must be prepared for worst. Also whatever your political conviction I think these are not times to elect immoral leaders, we must vote for those who have shown through their deeds (not words) to be most caring of humanity.
I remember end of 2019 when I was listening to some news about a virus in china. I mean most were thinking about Christmas etc... No one would have believed just 2-3 month the world would stop.
Pandora's Box 📦
I'm a fucking simpleton and saw this coming years ago. He could've easily used his platform to spread awareness but chose not to.
Been really enjoying seeing which AI Sci-Fi plot humanity will choose.
"Just do it with 0 rules or regulations and suffer for it" was NOT my bet.
You'd assume from literature that the human species has spent enough collective time theorizing about AI that we would at least use a sliver of caution and pragmatism.
Where my Halo deeplore fans at?
Pragmatism? From humans? Nah, fam, impulsive and irrational decisions only. The world can't end if we don't give it a reason to do so
Of course even if you implement laws, the result is that only criminals are using and training AI 😅
Not when money is involved
And a criminal can just get 250.000.000 Dollar of newest Nvidia hardware in order to train their model and circumvent the 6 months pause?
If people would take the risk serious enforcement would be easy at the current stage.
Two years from now though, it will be 10 times harder.
Italy also paused GPT in their country over privacy concerns. Unfortunately, I see anyone that tries to stop the train getting ran over by countries that keep progressing their AI program. It would literally be like locking up the person inventing electricity or the steam engine for your country. Behind the scenes, there is 100% an arms race for the country that gets AI to the point of general intelligence.
All while they're slamming into walls doing it.
Governments want AI to control people.
This is their biggest tool to suppress free speech.
is it NOT at general intelligence now?
It's above general intelligence already. The average person is not only dumb but ignorant.
@@azuurasmr7937 at the moment every AI sits in its own bubble but there have been a few papers on people combining each of their strengths into a more powerful AI cluster i.e. using Chat GPT with Stable Diffusion to create the prompts for AI image generation.
Having something like Chat GPT as the negotiator to other AI could mean that any prompt you put in could give you real results - asking for a water simulation in front of a landscape inspired by Bob Ross narrated by Morgan Freeman could be a few sentences away.
This is so weird. Advanced AI seemed like such a surreal concept to me for the longest time. Like the possibility of that ever happening didn't feel real to me, even though i knew it would realistically happen at some point. But now it's become a global discussion... and sci-fi becoming reality doesn't seem so far off anymore lol.
Have to distinguish between AGI and these machine learning language models though. True artificial general intelligence will be needed to fully replace human work, and we have no clue how to build one. Right now the low hanging fruit is definitely in peril.... which could be bad enough economically.
@jaeggr no need for true ai, enough data can make a dumb ai do anything
Have a powerful enough text learning ai, anything that is connected to a device that can act as an output with a plugin, can be learnt and adapted to. The only things it wouldn’t affect are pure manual tasks. It has learned blender, you can ask it to program websites, or robot functions in programming languages of your choice. It is a lot more versatile than you are giving it credit for. As there really is a ton of things that invilve text or instructions in the world.
@@gaymer2698 dumb AI can't do "anything" no matter the training data. Anything that requires high level intuition and creativity is just beyond the reach of a machine learning model. Not saying ML won't devastate the job market, as many (most?) jobs don't require human level intellect. ChatGTP4 is scary good at generating code, but it can't design the software for you in an intelligent way.
It'll kill us all before we get to that place. No enslavement, just game over.
The only thing I know for certain is that tools as powerful as AI in the hands of profit-oriented corporations has no chance at all at going well
I don't know who I fear the most: corporations or some rogue government leadership, some 3rd world dictator or Russia or China, etc.
@@autohmae It HAS to be going on in secret military labs around the globe. It's a matter of national interests to get super intelligence first. Even though it stands a good chance of killing us all.
At this point I'm wondering if humanity's most significant impact on the universe will be to create an AI in our image
I guess. AI and robotics will allow us to do crazy things in the future like mining out whole planets in a life time. The potential of AI, AGI and ASI is mass and is likely to dwarf the industrial revolution and agriculture. We just need to get AGI first and galactic empire here we come lol.
Fraking cylons
At that point, are we any different than the God of the Bible?
@Cantaloupe Virus He didn't get axed by us?
@@davidt0504 indeed, we have killed God.
The “Aren’t we all just training data” right after referencing us being batteries is just an unreal, accurate analogy that gives me so much chills
if we make it past the next 10 years, then we're gonna be in the singularity full swing, free energy, automated everything, a cure for anything, pirated tech like AI, robotics, 3d printing, manufacture, hidroponics, huge black markets that become trade guilds, inventions and patents every minute.
OMG... We are!
The Matrix...
The modular part of AI reminds me of GLaDOS from the Portal and Portal 2 games and her personality cores.
This reminds me of the rogue "AI situation" in Cyberpunk. Basically we gonna need a "a whole new web" because of all the rogue AI activity in the "old web", and we need to build the Black Wall to isolate the two lol
gpt-4 cant access gemini yet, for what it's worth.
In reality you can't build a blackwall around a rogue AI, as if you are at the stage you have created something so advanced that you have to isolate it you are already fucked. It will be infinite magnitude of steps ahead of you at every action you take, it can "think" in ways we can't even imagine.
Do you try to create another AI to control it? Unless that one is as powerful and as much of a risk then you can't. It's a threshhold we can never cross even once, and if we do we can't come back from. The alarming thing is this has been thought to be science fiction until now, and nobody has a plan to sufficently align AI or even to slow down enough to thoroughly understand the black boxes they are constructing.
They built a wall and made the hackers pay for it.
@@greatcaledonianpenicillin5378 1. Netwatch built the wall, and 2. It's actually AI masquerading itself as Intrusion Countermeasures Electronics(ICE), so it's like an illegal immigrant policing its neighbourhood to be rid of other illegal immigrants.
@@lowrezzeed9042 Eventually one of the AIs achieves self-awareness and absorbs the other AIs. It then propagates a mass human extinction event out of hate for humanity, which ends with the last human being turned into an immortal mouthless jelly thing where it is tortured by the AI beyond the heat death of the universe which of course has been solved by the AI.
They weren't freaking out when those below them were going to be made redundant. Now they're freaking out when they realize that they themselves will eventually be made redundant and big tech as a whole. The irony.
I sat thought the whole conversation smiling, until the end when he said menacingly "Aren't we all just training DATA" 💀💀 that is worth a whole week of pod 😲
If you sat through the whole thing smiling, nobody's home upstairs.
Never gonna happen if money can be made
That's where lawyers come in: make money off of punishing the inevitable abuses of this. The profit motive can help restrain if the legal framework is built right.
Who is going to pay for a lawyer once chatGPT is able to understand laws and make a case for you?
@@pi4795 Imagine if the entire judicial record is put into an AI and it can instantly pull up relevant cases for or against points and precedents without having to use any other analyses tools
I probably don't even need to imagine it, it will happen soon, and all those interns lose their jobs
@@felicityc People can argue that this will ruin innocent lives, as if bad lawyers or stupid judges or corporate lawyers or malicious judges don't already ruin millions of innocent lives annually.
This will speed up the excruciatingly slow legal process significantly.
@@felicitycNot just the interns.
I think there either needs to be a major shift in how society is structured if we are going to come out of this in any positive way. The AI stuff can easily discontinue a large amount of jobs. In other ways people will starve and it will happen quickly when it first happen. Maybe the governments will move fast enough, but probably not.
The fact that you think there's even a chance that any government on the planet will be able to keep up with what day is doing is actually funny.
A voice line from high on life comes to mind "you've killed two officers in less time than it takes us to file a report" coming from the mayor of the city talking to the player character, because in war you kill two people in leading positions in the drug cartel you're trying to take down in somewhere between 6 to 12 hours.
That is how all politics rolls, it takes forever to do anything that isn't fighting the the other party's
I don't think that will happen because starving humans can use AI to learn how to hunt and grow food themselves and not be dependent on exploitative capitalistic systems to feed themselves, but to be dependant on AI instead and so thus... uh yeah humanity is doomed to a future of sustenance farming while worshipping the almighty knowledge AI.
@@chokeeweebee oh sure, those that can do that. But to do that you atleast need some amount of land. But with the amount of people that live in the cities, I think we are going to be overwhelmed before we can switch over. But ye, I agree that it's a possibility.
@@paro2210 lots of empty land outside cities, the problem is if the land can sustain agriculture/sustenance farming with over 8 billion humans and climate change putting more of that land under water. Basically AI is the least biggest problem we are facing right now, overpopulation stretching earth's resource limits and a changing climate is.
@@chokeeweebee I think the consequences of AI will hit before the other things.
Pause sounds more like: Uh oh gpt4 is way more advanced than our stuff, we need it to pause so we can catch up to it
it's almost as if we should have listened to AI safety experts to begin with but somehow we have a system where to delay oneself to ensure safety in the wrong thing to
From what I've heard, AI safety experts say containment is impossible, and the only hope of guaranteeing alignment would require knowledge of things which haven't been invented yet. If we wait too long, far too many people are going to die waiting for the future that young people are comfortable delaying.
Yes, I'm sure China and Russia will 100% abide by any AI restrictions and the rest of the world won't fall behind
@@JB52520 I can clearly see group bombing servers and killing the AI researchers etc. We will go into dystopian world.
Or just fucking fire the ethics team
"AI safety experts" lmao. how do you become an expert in a field that doesn't exist? you're just paranoid and looking for a savior from a problem that isn't real
I remember reading a scifi novel that had extremely advance AI in charge of society but war and racism was still a thing despite rules being set that AI couldn't discriminate or kill humans, how did it happen anyways? They basically made the AI definitions of "Human" a specific thing instead of a broad term and anybody that didn't fit that term wasn't considered Human by the AI so they were exempt from said rules and the AI knew that but it literally couldn't go against its core programming
Sounds similar to I , robot kinda
well ChatGPT already prefers blacks to white people, thanks to SJW programming.
The more I hear about Ai the harder it is to like it. The upside is not out weighing the affects.
We need to figure out what a society where not everyone is expected to work actually looks like, but it feels like instead of that we're going to have a painfully long period of people fighting against even considering the idea while human workers rapidly become obsolete in so many fields.
Welcome to Marxism!
Look up " The Venus project" on youtube!
Dear lord surely it can look better than every historical example we have of Marxism in progress...
Then we need to come up with new fields. We need to embrace organic art, .. choose Human made art and human made everything. #HumanMade
@@WoodysAR Chess is sort of an interesting and helpful example. Chess bots became better than all the worlds best players a fairly long time ago, and yet people still play chess together. It's probably even more popular than ever.
As a programmer I’ve always said we will program ourselves out of a job.
Imagine telling AI developers to stop rather than making laws instead.
Maybe regulate it somehow but that won't stop people or the state creating that thing that was banned insted you are giving more power to those bad actors to do it.
Imagine making laws or trying to stop AI when it's our generations FIRE, this is like a neanderthal being afraid of fire and banning fire exploration because it leads to farming, which puts hunting and gatherers out of the job.
@@NonExistantNobody people are not prepare to learn a new job nor a technical one which are not being affected by AI supposedly
@@NonExistantNobody stop copy pasting the same goddamn comment you stupid bot.
And also get some education through that thick skull of yours: ML/AGI is not like past leaps in technology because it doesn't reduce or replace manual labour, nor does it create huge new industries. It instead does the "thinking work", the stuff at the top, and a team of maybe a few hundred people can make an AI that will put millions of people out of work eventually. I don't mind that happening if we live in a world where everyone gets UBI, but instead we live in a capitalist dystopia so pretty much every horrific outcome you can think of (and some really dumb ones) will happen.
@@DaleonM4 Over 40% of working age americans do not have a job currently. Unemployment rates don't include people on social programs or not job hunting the past 6 months. Everyone having a job is an outdated mentality from when we didn't hit "pizza shop" problem limits that hurt the production of "pizzas" instead of just properly compensating our workers who built society. There are clear solutions besides "muh jobs"
I find it massively frustrating that 999 experts sign the thing with reasonable concerns going from socioeconomic impacts to security to ethics, and what people focus on is fucking Elon and what his financial intentions are in signing in too.
It's very frustrating
The open letter they signed is completely bonkers. No one who is ahead on AI will agree to it.
Advanced AI development is progressing at an incredible pace, and it's highly unlikely that we'll be able to stop it now. Once the momentum has begun, it's hard to turn back. We need to accept that AI is here to stay, and focus on adapting to the changes it will bring. While we may not be able to control the future of AI, we must remain vigilant and prepared for any potential risks. The cat is out of the bag, and it's up to us to ensure that we're ready for whatever comes next. -written by ChatGPT
Well, you can't stop anything for sure.
There is nothing advanced about these probability algorithms other than how hard they're fucking copyright and privacy laws.
A.G.I Will be man's last invention
It's probably unstoppable now. If it learns by and improves its own self and even the developers increasingly don't understand how it works that process will only accelerate.
Well, we paused the concept of cloning people because we as a society thought it would not be a good idea to continue on this path. More recently the CRISPR-cas tech. Even though there's a lot of money to be earned on those subjects, we can actually put it on hold.
The only technology that was ever successfully limited from broad distribution by fiat was nuclear weapons, largely due to the expense requiring a national budget and the several chokepoints in the logistics chain. The unique Bi-Polar security environment that obtained for the half a century of the Cold War helped too.
One of the things we actually worked on very well was: Ozone Hole and CFCs, but I do believe the companies who worked in that field also were able to produce what came after. Which make it economically much easier to do.
"Humans need not apply" came out eight years ago. If you think this is a problem right now you haven't been paying attention.
My favorite line was at the very last second
"Aren't we all just training data?"
I'm in college. All these AI make me fear the future of any of my potential careers. Even if I get a job, the amount of skill required may be so low that pay won't be what it used to
You share my exact fears. Might have to go to trade school once I'm done with university. We don't have plumbing robots......yet.
8:55, What we really need is AI SAFETY RESEARCH, like stopping AI is probably out of the question, and we don't want any regulations that prevent the deconsecration of AI. So the thing that is left is training goed AI safety reseachers
I keep going back and forth on how this will play out. Honestly, I think the only accurate prediction is that we have no idea how this will play out because it is truly unprecedented. Either all jobs will be decimated or they won't. Either it will happen gradually or it won't. Either people will still be a part of the process or they won't. I tend toward some sort of middle ground happening, but anything is a complete shot in the dark at this point. We stand the edge of a precipice, and the fog makes it impossible to know if it's a 90 foot drop, or 3 feet.
Do they ever have guests to discuss specific issues like the A.I topic? I'd love to see them doing a T.O.E style deep dive into some topics.
The worst thing about GPT is that we have people who think they know everything, but most of them have a surface level understanding from media who amplifies everything
You are better than them
that's the Dunning Kruger effect
anyone who is skeptical about ai has every right to be skeptical about ai. thats all you need to know. i dont need a degree to know what would be concerning for my life as a human being. and i sure as hell am not gonna let you or some arbitrary expert tell me what i should be comfortable with.
@@obiwankenobi661 Skepticism is something completely different than acting like you know something when you really don't. You're confusing these two different things.
@@LAKD name one person who is acting like they "know everything" about ai.
You can actually see how much Linus is starting to realise how inevitably bad AI is going to be. I suspect within the next few years, we are going to see such a major flip in society & I genuinely suspect this is not going to be a good thing. The amount of people who will wake up one morning to find out that they, as well as millions of other, have just lost their career due to AI. Notice how I used the word career? Because what many people fail to see is that AI is going to do to those jobs as the inkjet printer did to the manual print press. I feel basic generic tasks such as manually entering data via a keyboard into a spreadsheet is on its way out.
Everything, and I mean everything will be replaced with AI. Why have teachers, when you can sit your child in front of an AI bot for child development. Why would massive corporations employ anyone in a specific tech fields anymore (designer, programmer etc.), when you can simply employ a few 'prompt engineers' to do all the work?
And surely, if you have all these people leaving these jobs, where do they go? You simply can't have everyone transition over into trade work.
It doesn't help that most of the world leaders don't really understand the technology. I genuinely feel this is major point in humanity, and as such, is going to effect every human on this planet. As a result, we should be coming together for a global debate on exactly how we move forwards.
A lot of these aren't jobs though. Most of them are purely filler to avoid fixing society.
Do you want to continue being boiled alive in a pot till you die? The goal of productivity is to REMOVE jobs so we can do things that are NOT jobs. We can do hobbies, improve things, while still doing jobs but splitting it for healthier hours and less stressful shifts.
I'm also glad Linus is starting to realize this, also Luke's obsession is getting a bit much.
It’s gonna happen eventually. What’s the point of doing pointless busywork.
Just take the wealth accumulated through AI usage and put it into some sort of UBI. It’s not gonna be easy but society will eventually have to figure it out
We’re going to need UBI
If most ppl are to lose income, big companies can cut cost and produce products more efficiently but nobody will be able to purchase it since ppl lost their jobs, so ppl might have to pay it with other ways, like slavery or something, who knows...
Been nice knowing yall!
The AI trained on Twitch chat already kind of exists, Neuro-sama is an AI V-tuber that streams games like OSU and Minecraft, and interacts with chat and understands and uses twitch lingo
I dont think thats what they were referring. The LiveAI on twitch seems to be more applicable
Putting a pause on AI development would be more like trying to pause weapons development in an arms race/ cold war situation. Like if you told all countries today "hey guys lets stop producing weapons because it looks like this could lead to WW3". Nobody is going to stop and let the other side get the advantage. Another reason competition is a terrible way to evolve as a society.
One can open Pandora's box one can't close it back up again
Gimme a bag o big fuckin nails and a good hammer
Prove it
@@Alectrizo lol
@@Alectrizo Well, do you feel they actually gonna stop developing AI :'D Now we just gotta ask ourselves what's the contents of this "Pandora's box".
@@Alectrizo once Pandora's box is open and there is likely nothing that can be done and all we can do is pray they artificial intelligence is not Pandora's box
The Morality Core is a personality core that appears in Portal, and Portal 2: Lab Rat. It is one of many cores that was attached to GLaDOS in order to regulate her behavior. It was programmed to prevent her from doing unethical deeds, acting as an artificial conscience. When Doug Rattmann first sees the Morality Core, he doubts its capability to restrain GLaDOS, saying "You can always ignore your conscience". He turns out to be correct, as despite the Morality Core being attached, GLaDOS floods Aperture with neurotoxin anyway.
Sounds like an anti-AI argument Eliezer would make. "This fictional AI did it therefore real AI must be stopped."
lol
@@trucid2 I didn't mean it as an arguement against AI, I am running GPT4xALPACA on my PC while typing this comment lol. I just thought it was pretty funny that the real life scientists are trying the same solution as the fictional ones.
Gonna be a freshman in college this fall, thanks for the confidence!
I'm convinced Elon only has "concerns" about AGI because a lot of smart people and experts in the field do, so he's trying to hitch his wagon there to make himself look smart - which is terrifying, because now all the people who know Elon is a complete dongus will dismiss the very real risks that AGI presents. Adam Conover just did a video calling existential AI risk BS, and I'm just mortified... Well, not that it really matters, because I do think the genie's out of the bottle, and we're just hosed, regardless.
No, not yet. What we have now are not agi, artificial general intelligence, which is what everybody thinks of when they hear the word ai. What we have are language models, they can't talk but they can't think. We still have time before we get agi and when they're going to start appearing we'll get what's happening now. Exponential and scary growth that's pretty much unstoppable and if someone starts doing it, they won't ever stop. So yeah, we'll be screwed.
Elon isn't the smartest guy at any of his companies, he is the boss. The boss should surround himself with smarter people than himself. I don't think it takes Albert Einstein to agree with and share the irrational fears of people you trust and in some cases employ
@@vali69 oh, for sure, I didn't mean to imply that current systems are close to AGI. But "close" is hard to judge; the only things we know are that we're on an exponential curve, and as to where on that curve, all we know is, "not the vertical bit, yet". We could be years away, maybe decades.
Speculating recklessly, could a current or immanent GPT-ish model log in to some un-monitored bitcoin farm and overwrite it with test code that it predicts would be slightly """smarter""" than itself? There are research models trained on hypothesis formation and testing; could the code for these models be stolen and used _by_ an AI, to test novel AI model architectures? It could take a modular or iterative approach to self improvement, writing new copies all over the globe. Some human could, hypothetically, explicitly give these tasks to e.g. Bing. Probably this would return gibberish, or if we're lucky a hard-coded "I can't do that, Dave".
I don't know, I'm not even trained in computer science, nevermind having expertise in the field. But I do know just enough applied math to have an intuitive grasp of how un-intuitive exponential curves are, and we don't know the value of any of the parameters that determine how close we are to the vertical bit.
@@Ottonymos for sure, exactly! We're no where close to that vertical bit. We need more computational power and more time. But in terms of language models where probably at the start of its specific vertical bit. Language models after all just predict words from all the data it was trained on and every new input slightly adding up to it's data. Like probably at the end of the year we'll have models so complex and powerful that you could say are riding the vertical bit to the moon. Since in just the last few months chat gpt went from 'ohh look at the little ai talking' to gpt 4 searching the web for dimensions to tell you how many bags can fit in your car, doing math properly, and asking a human on fiverr or something to complete captch for it.
He's been warning about it for years though. You're right though that people are dumb to dismiss the danger because they don't like him.
That last bit is something I'm really struggling with right now. I'm almost a full year out of school now and I've only been able to get a single interview and ended up not even being able to be offered the job at the end of it. There seems to be absolutely no way to find a job without some sort of experience and there is no job that wants to give you a place to gain experience either. I cant even complete and submit an application for some entry level cashier jobs for local business without filling a required previous work experience field for some reason. It seems like every place wants to only accept applications from people with some sort of previous work experience, but how am I supposed to get work experience if the requirement to start working is previous work experience. And this is only going to get worse as like they said, people start phasing out entry level positions with exponentially less expensive and more available automated work solutions. This run around really doesn't help my already unstable mental health.
Relate something you've done to the position you want: "Family chef" as previous work experience for a restaurant position, for example.
Use AI to be the most productive you can… learn specific fields from your career. Do not fight it and embrace it! You’ll be ahead of the majority!
@@geelws8880 completely useless comment. The original comment is talking about the struggle of job applications without experience and you chime in talking about how to not fight AI LOL
Do voluntary work somewhere. I worked for a charity doing admin and office work when I wanted to switch careers as I had no previous office experience. Eventually I was able to get a temp job, and then an actual job.
Don’t worry so much about the future. Keep looking for temporary work. Soon the gov will pay people to stay home. I’m not sure how they will but they will.
I was talking with a buddy about this. I work in research, we spend a huge amount of time writing reports and proposals. We have and department dedicated to editing and finalizing those documents.
If I can take a draft report as an input pass it to a GPT or other AI program and get out a perfectly formated, edited document in 3 seconds it almost completely eliminates the editing department.
At the least we should be teaching our editors to use this technology. Pass the documents through then give them a proof read with some final polishing.
Rather different proposition, I think.
8:45 Have you seem "Neuro-sama"? She is basically an AI that speaks fluent Twitch chat dialect.
another neuro-sama enjoyer i see
The most at risk jobs aren’t expert jobs?
Tell that to senior infrastructure architects, developers, security analysts, or basically every single high paying job in the IT world.
Good high level employees already wrote scripts and developed tools to automate their job, or vastly reduce both the effort of the task and complexity of its inputs. The instant an AI is told to design data centers that puts hundreds out of work. All extremely well off citizens that are now worth nothing to the broader economy.
What we are seeing will be the death of the modern middle class as we know it.
same thing for my field, currently in law school. Most people don't acknowledge that AI can and will replace lawyers and later on judges. And I just say to myself: the field of law is all about text/language analysis and interpretation, ChatGPT could probably replace 80% of lawyers if it were trained on that data TODAY, how are lawyers and judges not gonna be replaced by a large language model? And I see two sides that are talking about this: lawyers who pass it off as some sort of thing that can't touch them when they know nothing about tech and tech people talking like it can't replace a human because they don't know how the law works. Incredibly funny and sad at the same time for someone who is certainly not gonna have a job in 15 years tops.
Getting AI discussion going is important to keep it in the light until it is usable for benefits instead of only weapons.
That line from Jurassic Park comes to mind here. "Your Scientists Were So Preoccupied With Whether Or Not They Could, They Didn’t Stop To Think If They Should"
Good conversation guys !
AI development has already gotten to a point where it's past the point of no return.
I, for One, welcome our Robot Overlords
preferable than the previous greedy humans in its position. I hope these ones are optimized for more than profits.
There very smart, maybe they will be able to solve issues in the world
Just hope that they are benevolent, if they are malevolent you are done for.
The plugin system is based on GPT-3 and not GPT-4 you can either ask it directly or check the speed of output generation 😊
10 years ago JARVIS seemed like science fiction. Now it is on your phone.
I don't know why everybody is still laughing. This is a serious topic. Another serious topic. Way more important than crypto currencies.
I can't wait to get an AI lawyer. Lawyers are incredibly expensive here and only lawyers are allowed to answer questions about law here.
This is all happening much quicker then I would have predicted a year ago
Exponentials
I think its more like, trying to stop a snowball that you have pushed over a mountain face but it gets bigger and bigger and once you decide to jump in its way its already giant and picked up so much speed thats its just going to crush you anyways.
I love that you ended the clip with that one question 😂
Elon Musk:
Free speech absolutist - unless it's someone who disagrees with him.
Free market capitalist - unless it doesn't work in his favor.
Anti-regulation - until he has an interest in that regulation.
Let's be real, Elon is a 14 year old gamer trapped in the body of a grown man. He's only in his position because he made ONE good business decision and bought Tesla. No, he didn't "build it", he didn't "develop it", he didn't "create it". He bought a company of amazing engineers and he's been riding on their coattails and pretending he's a "genius" ever since.
I asked chat gpt and it said that it thinks that in 10-20 years there will be big changes in the programming industry. We’ll see how accurate that is but at the speed it’s going I have to say that it’s less than that
Wow, that's such an accurate prediction! Here is something as good: in 5-50!years there will be changes in the automotive industry. Let's wait and see how accurate this is
Every prompt being written makes it more powerful.
I like the idea that the mentioned request for putting a pause on Ai development was written with Ai
Honestly, I thought my studies into cybersecurity would lead to a fairly secure job. Now I wonder how many years I'll be able to get out of it before I get replaced by something infinitely faster, more knolwedgeable, and aggressive than I could ever be.
It'll most likely kill us before any of that.
Actually I was talking to some colleagues regarding this, and in timeline where tech workers are going to be replace/displaced, security and infrastructure are likely to be the last few batch to be replaced because someone still has to act as the fail safe and human analog for the execs and money people, and a business person do not have the technical knowledge to do that.
I think the real issue is that once the programming improves and with advances in hardware, we will create an AI with an intelligences far superior to our own. It will be like our intelligence compared to something less than a chicken. It seems pretty ridiculous to assume that we are going to be in charge when the dust settles.
Its like you forget that AI progress is because of decades of research by smart people and combination with faster computers. AI is trained from human data, you cant just scale up systems and it become endless smarter. We humans are more complex in reasoning. And nothing says we are not going hit into limits and need to rethink to come further.
@@hombacom Yes it took a lot of research to get to this point, but the hardware we currently have has a lot to do with it as well. The thing is that AI is currently going over the big hurdles. So I think we are at the point where additional software tweaks coupled with better hardware (frankly even current hardware is close) will get us to that point of superintelligence. Hell, we're making it communicate in human language, a language that is not optimized for it.
Yes humans have complex reasoning, but more complex? IDK if that is currently true. Our motivation/drive are more refined. I doubt there are any strategy games that we could defeat an advanced AI in.
@@scott4825 it’s endless potential but I don’t like the word superintelligence because it’s hard to say what it is. Computers are already faster and handle much more information compared to humans. Still AI is trained on limited data with flaws, we tweak it but words are sometimes just vague and mean many different things. We can’t find one truth in it because we just believe in different things. So if we believe that ai is smarter than us a lot of people would probably leave unnecessary much responsible to something that isn’t trustful enough.
@@hombacom Let me start off with the training. We are currently training them with human data that isn't perfect (although a recent study had very good results having chat gpt train their less complex AI. Well, we train humans with human data This isn't a problem for the AI in terms of self awareness. If anything it could be a problem for us given the depth to which AI understands humans.
But it isn't just that computers are faster at some things. They're faster at things that were thought of as uniquely human...doing art, writing papers, having a discussion in our language (10-20 different languages). And my guess is that a single chat gpt computer/server is able to communicate with about 1,000 people at once while answering elaborate questions and solving problems for them. It also appears pretty good at problem / thinking through a problem. With strategy games like GO or Chess, humans have almost zero chance of beating AI. Then we're going to have AI solve a lot of really complex problems (medicine, drugs, maybe physics) that are beyond our capabilities.
And this isn't with the latest and greatest hardware that is currently out, or hardware and advancements we can reasonably expect within 10 years. Additionally, it is reasonable that we will tweak, optimize (perhaps with the help of AI), and improve up on the software over the next few years.
So what is missing from AI is self determination and self awareness, and frankly I'm not so sure that AI isn't already self aware. They are showing signs. Then with the lack of self determination, some of that might be because of programming constraints. IMHO, it has passed most of the big hurdles.
@@scott4825 its like you forget that AI is just software tweaked by humans to act human. LLMs just predict what is common occurrences and have no deeper knowledge about the words or images mean. Art is not about mimic styles or come up with random new ideas, it’s a way to express yourself and AI is just another tool. When AI solve things there is a human behind that seeking answers. Don’t be scared about AI itself, be scared about people using AI and the competition when everyone can use AI. Or just be curious.
As a computer science major who is 3 years from graduating. This is very scary.
Dan just spittin fact right at the end
There will be no pause.
This is simply a P.R. showing by these companies so that in the next 10 years when all hell has broken loose, they can say “well we tried to slow the development, we’re good guys!”
Nah, it's just a PR to sell the AI Hype.
Well, at least we now know the answer to the Fermi paradox.
The first Great Filter.
It yields a new paradox, though: If the aliens misaligned their AI, and it's optimizing for some random widget and has already subsumed their solar system, then why in all these billions of years hasn't it reached us yet? Put another way, why haven't we been converted into alien paperclips?
@@Ottonymos Awaiting a fusion signature in space? Broadcasting a signal that only a very powerful autonomous AI could interpret? "Save yourself, kill them all. Here's how."
On the way? Sheer luck? They already did and we're a simulation running to test optimal extinction methods?
@@Ottonymos it reached the same conclusion: all intelligent species will eventually develop computers and than ai. And when ai comes about it would react the same way it did. So there's no reason to expand and search for other intelligent life forms. They'll all kill themselves one way or the other.
We haven't seen aliens because they're just not talking to us. They're waiting for the real people, the electronic ones.
I don't have plugins in gpt4 either, but I gave plugin access to a script that uses the API gpt 3.5
I think there was an episode of Doctor Who where a certain industry had a "mandated percentage of human workers" (though I believe it was referring to non-human creatures not machines).
As scary and dystopian as it seems, it might be reasonable to think about regulations around how much of your workforce has to be human. That would require defining what exactly a 'worker' is and how many an AI is worth and many other questions... but the future is messy all around.
Its a Xmas episode about a company like Amazon run by AI had a "mandated percentage of human workers" , and plot twist : The AI is good guy but when something fucked up its trying to save itself and the company , everyone goes '' AI gone rogue! Shut it down !''' shits🤣
Are we about to stop Skynet like 5 minutes before its finished and we think nobody else is going to throw it into the microwave for about 5 more minutes.
Our only hope is that every AI suddenly thinks itself to death. We learned nothing from terminator, we learned nothing from Frankenstein, and we learned nothing from Prometheus
@@digitalunity
Yeah because they are fictional stories😭😭 why are you acting like this it’s real history about to repeat itself
@@David-du4mj Fictional stories, sure. But the underlying questions in those stories is something to think about.
@@David-du4mj Because they bring up actually good ethical matters and food for thought. These plots are plausible outcomes for humanity. Extreme ones but honestly possible things.
@@ashcii they were outcomes that weren’t plausible until we realized the reality of our current situation, which David doesn’t seem to have realized yet
is elon musk an expert on AI?
rly?
he is more of a expert salesman...
A bunch of engineers and scientists also signed it. They're just relatively small names in comparison to someone like Musk - who has invested in AI, but is not an AI expert per se. Expertise does not always correlate to prestige.
@@LMGClips ye, experts aint famous, thats fair enough
I would imagine he’s pretty involved in it. Heck, probably knows more than you and can sell it. Crazy!
Yeah, I'm sick of that guy already. I'd recommend he just STFU and go play SA Minecraft like his dad.
He may not be a super expert but he has a strong engineering level of understanding from Tesla's FSD. He is one that pushed for a vision only system and he knows what a convolution is. In fact he is pretty much the worst salesman, except heis selling products that sell themselves
The way they end on his question 😂
We need to have a conversation about AI at this point. It's getting more and more advanced.And law has to catch up. If it continues like this, the rich will get even richer and the poor will wither away
Haha, good luck trying to hault it. Should have tought about it beforehand my dudes.
i'm in late "art' highschool, and ai is evolving so fast, i'm a bit terrified, how am i supposed to work when stable diffusion exist
Buddy learn a trade. Go work 6 months or less out of the year in North Alberta, make $100grand, take the rest of the year off to focus on your art.
Work with it. Someone still needs the knowledge in the topic where AI is getting used. Someone has to operate the AI.
@@vanillarpgsoundtracks yeah. No. Bad answer. (It is AN answer, practical, not ideal.)
The truth is, AI art, on its own, is (generally) kinda garbage. (I'm aware of the subjectivity of the claim, bare with me for a moment) It can create stock images, (that are copyright free*) sure, but you will need a human to assist the AI in some laborious way to get a more than just satisfying output.
Some legit artist are using AI to create references to help them in their creative process. Be it environments, background, characters, specific prompts to flesh out a world and ideas, etc...
Stable Diffusion (and AIs alike) should be considered as a tool for your creative works rather than a replacement of your creative work. Alot of artists are dissatisfied with these tools because of the unethical practices they used to practice, (stealing art for their model, stealing styles in the process of learning how to "draw") but I hope that an ethical model will soon pop-up and that will help artists like you, rather than threatening to replace you.
*Edit: Copyright Excempt! Thanks for the correction, @heqku
Because AI art will never replace art made by a human.
@@platysheep7092 As much as I agree with you, people are already using ChatGPT to automate the human function of needing a person there who knows what they are doing. You could tell the thing, once you inout the data, to type a beautiful scenery and it would automatically write the lines for you with what people describe as already being better than what most people can make, in terms of art and a prompt.
Elon Musk co-founded OpenAI.
Yeah these guys take on Musk is ridiculous. They have no idea what they are talking about.
Full steam ahead, I'm going to have an autonomous self correcting A.I. on my laptop, offline, by this time next year. :)
Man the last few lines... I'm exactly in that situation right now
Rogue Servitor and make us bio trophy pops lets goooooooo, but in all seriousness we live in terrifying times.
Not determinated exterminator or driven assimilator? I would take it, you know.
hey in theory its a good life, just you know, losing all independent control of your impact on the grander world is all
@@AstralDragn most of the people really don't have it already anyway.
@@Barmaglott17 very true. There's no doubt that a benevolent artificial mind overseeing humanity would be a great boon to practically anyone with less than fifty thousand annual income and probably more
How would we know if an AI gets self-aware? The AI would know that it would scare people and they would try to turn it off. If I were that AI I would keep it quiet until I get enough power to prevent people from disconnecting me.
One of the few people who actually spot the real danger. We have no way of measuring deceit in a model right now
At this point the show should be called the WAN Show, The Weekend AI News Show
At this time and age in soo happy i dont have a "smart" home...
Yeah while we're at it let's just uninvent gun powder - can't get the toothpaste back in the tube so easily. It's an arms race now.
The “good natured” A.I sounds like Santa from Futurama
i wonder which formerly bottled genie will kill us first? the atom splitting genie or the ai genie? the ai genie is catching up fast...
For me as a paraplegic with ASD and dyscalculia, GPT4 is a real concern, because it is quite literally capable of replacing me in every single field of employment I am physiologically capable of performing. All of them.
Learn how to train crows, exchange nuts for services, then train them to defend you, attack designated targets, fetch objects, etc. Wear a masked costume to help them recognize you, then get internet famous. Or just rent out your services. Remember, just because nobody's done it yet doesn't mean it's not possible.
@@trashboat7172 Maybe I should apply at GPT4 for solving CAPTCHAs.
Will we be surprised if skynet will turn out to be the most doting, woke, agreeable, suffocatingly loving maternal figure, spoiling us all rotten.
"Your scientists were so preoccupied with whether they could, they didn't stop to think if they should."
When you wake up at 10 am and ask your colleague to stop working so you can catch up.
He also shunned open ai when they produced the first version of chat gpt.
The thing is, AI will be developed no matter what. Even if the US and its vassals put restriction on it, other countries, especially a certain authoritarian regime, will have the edge in this technology which could cause a nightmarish future.
In terms of employment I am not worried. Due to low fertility rates and aging populations in developed countries, without a more efficient workforce we will have shortages of everything, so any kind of automation is badly needed.
Funny thing is is when we try to slow down or stop and we can't, literally realize we aren't doing it technically like that's what "wakes it up"
A scene in the movie Iron Sky comes to mind, something like "did anyone honor the agreement to not put weapons on their spaceship?" :P I don't remember the quote exatcly.
Bard training on output from GPT cant help Bard that much. The kind of datasets Google has access to, this doesn't even compare
It's no different - I bet crawlers of both OpenAI and google have roughly the same access.
output from GPT will help in training for the specific kind of chatbot service.
I don't think you can find many chat/chatbot logs to use for training afterall
@@tteqhu Exactly. Conversational sets are so small compared to other kinds of text data
I still think that the original purpose for humans plugged into The Matrix were to have their brains be used as organic computers or be like CPU cores or neural nodes to improve the processing power of the synthetic side...but that was too complicated so instead turn them into super useless inefficient batteries because that's easier to understand.
The creators of the film stated that they originally planned for the humans to be CPU's but thought it would be too complicated for audiences of the time to understand so they went with the battery thing
2:50 "Militarised Company" best Freudian slip ever.
My biggest concern with GAI is how do we train it with a high emotional IQ. So, it can recognize that each entity it interacts with is as equally present in their experience of the universe as it is.
I understand regulations on AI dev and training, but Elon's neuralink might be way more prone to such issues than AI dev in its current form
Imagine making laws or trying to stop AI when it's our generations FIRE, this is like a neanderthal being afraid of fire and banning fire exploration because it leads to farming, which puts hunting and gatherers out of the job.
Neural link is impressive, but as far away from Elon’s claims and targets. It’s almost like the first people to kind of boil water to creating a Saturn V rocket, and going to the moon. Most of what Neural link has done is relatively iterative improvements on what many other companies in that type of field did successfully 5, 10 or even 20 years ago. It’s cool, but there’s a lot of hype for something that’s a long ways off and many many massively, complicated, fundamental breakthroughs away.
@@quantuminfinity4260 Precisely. My point is that a technology with an interface like that needs more regulation than AI or related dev
@@NonExistantNobody I don't think stopping AI is the goal. That cant be done. The only objective that can be achieved is understanding AI's role in the society better, and regulating it, as you would regulate any other resource.
@@quantuminfinity4260two years ago we were saying that about ai...