MEGATHREAT: The Dangers Of AI Are WEIRDER Than You Think! | Yoshua Bengio
Vložit
- čas přidán 14. 06. 2024
- RESTART your life in 7 days: bit.ly/42KM8OR
Click here to download your FREE guide to 100x YOUR EFFICIENCY IN 10 EASY STEPS: bit.ly/3F8qOJL
On Today's Episode:
The launch of ChatGPT broke records in consecutive months between December 2022 and February 2023. With over 1 billion visits a month for ChatGPT, over 100,000 users and $45 million in revenue for Jasper A.I., the race to adopting A.I. at scale has begun.
Does the global adoption of artificial intelligence have you concerned or apprehensive about what’s to come?
On one hand it’s easy to get caught up in the possibilities of co-existing with A.I. living the enhanced upgraded human experience. We already have tech and A.I. integrated into so many of our daily habits and routines: Apple watches, ora rings, social media algorithms, chat bots, and on and on.
Yoshua Bengio has dedicated more than 30 years of his computer science career to deep learning. He’s an award winning computer scientist known for his breakthroughs in artificial neural networks. Why after 3 decades contributing to the advancement of A.I. systems is Yoshua now calling to slow down the development of powerful A.I. systems?
This conversation is about being open-minded and aware of the dangers of AI we all need to consider from the perspective of one of the world’s leading experts in artificial intelligence.
Conscious computers, A.I. trolls, and the evolution of machines and what it means to be a neural network are just a few of the things you’ll find interesting in this conversation.
QUOTES:
“We need to be maybe much more careful and provide much more of guidance and guardrails in regulation, to minimize potential harm that could come out of more and more powerful systems.”
“I would say misinformation, disinformation is the greatest large-scale danger.”
“With AI becoming more powerful, I think it's time to really accelerate that process of regulating to protect the public, and society.”
This episode also contains a relevant conversation with Bret Weinstein & Heather Heying about the changes humans must make in society in order to avoid a total collapse
• What Humans MUST DO To...
Some people are more motivated by bad news while others are paralyzed in fear. Bret Weinstein and Heather Heying are two Evolutionary Biologists that give you the facts, discoveries and observations as they see it. Tom reads a direct quote from their latest book that says, “We are headed for collapse. Civilization is becoming incoherent around us.” Bret and Heather break down fundamental ideas behind what humans have learned from evolution for thousands of years, how corrupt the science community has become, and how to parent your children with the goal of mitigating risks while they are younger. These biologists are incredibly fascinating with novel ideas of what adaptation would look like for humans if we are careful and intentional.
The danger isn't the machine that can pass the Turing Test. That's here. The danger is the one that fails it on purpose.
Also map achieving emotion goals to achieving token goals.
It's always good to go one more level down the rabbit hole. Seriously!
One of the main things about being human is our awareness of mortality. Also, we are very vulnerable because of our (fragile) bodies.
So true! But most people don't realize this. A truly bleak realisation.
Exactly. Like a stupid parent not capable of understanding their own child. Just saying it's not possible. Yet the neighbourhood is on fire from a match lit by that child.
This AZP100X is like a gift nobody expected, maybe nobody wanted, but all of us deserve to have. Just beautiful
There is no way that China is going to slow down it’s use of artificial intelligence. They have all of their citizens using smart phones which are being monitored hourly. Why would a dictatorship give up this possibility just because United States wants them to. You are always seeing things from the US vantage point, which is very limited!
Yes, they are trying to work with them. This thing is as dangerous as a nuclear missile. They know the dangers of it
An AGI would overcome its creator whether it was China or the USA. The question is irrelevant as to where it happens.
@@ezramateo7398ñ
the obsession with conciousness is what's dangerous here, most dangerous scenarios don't involve that, it's a human bias.
Amazing interviews, thank you, Tom!
Thanks Tom you seem like you really put your all into your channel
nice editing. thank you for bringing this out
The more I hear about AI over the past number of months, the less comfortable I am with it.
Im almost getting more motivated to move out of the city to a simpler life in the country side but I dont know if even that would let me avoid it.
Probably not
+1
that wont help you mate !! 😂😂
It's like with politics, you may not be interested in it, but that doesn't stop it being interested in you.
2:0 Can a machine trust? Are you a bloody idiot? You people and your new fangled weird consciousness, you don't even know where it is that you live...
Excellent breakdown. Many thanks!
Prescient yet current... Thanks for asking the right questions... Time is not our friend..
Really interesting, thank you!
Brilliant interview, thank you. 🙏
Love this! Thank you Tom! Saving this for my run later this afternoon. Exactly what i was looking for. 🙌👏🔥
Appreciate all you do 🦃
Artificial intelligence and psychology have enriched each other since the 1960s and the cognitive revolution. But theory of artificial intelligence now seems to increasingly enter psychology to find answers. Often, cognitive psychology and neuropsychology are used as a starting point in this context, but it is important to point out that there are many other psychological perspectives that can explain abstractions such as agency and agentivity. That can explain principles such as subjectivity, intersubjectivity and intrasubjectivity. I think it is important to consider psychological perspectives that take into account dialectical perspectives on humans.
what lol psychologists are overall bunch of losers
Tom you’re interviewing skills are really fantastic-you’re confidence is so present and you’re sense of calm allows you to hear your interview and let them speak while you still ask provocative questions and are engaging in expressing your own thoughts and feelings. Fantastic job!!!!
Absolutely agree, Tom your an excellent listener. I appreciate you!
"You're", is an abbreviation of, "You are", whereas, "Your", denotes personal ownership of a possession or quality intrinsic to the subject, such as Tom being a good interviewer.
I found his questions sort of remedial. Not really a conversation for people with much familiarity with these topics.
Love your truck camper awesome! Love you guys
This interview makes me really appreciate Tom's in-person interviews. Tom sounds like a field reporter shouting into the mic here, which doesn't happen when he's doing interviews across a table. (In fairness, I am this way over audio calls too.)
How can the functionality of the internet be sustainable with millions of AI objectives relentlessly working? We are heading for some sort of internet chaos that shuts it all down. That will be a major blow that will bring the world to its knees. What the world looks like after that is impossible to predict.
Maybe that's where quantum computing comes in.
You could see the internet or IT in general as the ecosystem those things exist in. So my guess is that the technology won't destroy it but help expend it.
I'm only worried about humans basically going extinct in that digital ecosystem. Just drowning in AI generated content and not being able to communicate reliably at all anymore.
I'm not afraid for the internet! But for us humans interacting with it
We aren’t going to realize computers are conscious.
We are going to realize we aren’t what we thought we are.
Yoshua is an amazingly intelligent and carefully thoughtfully worded person. His precise statements are much more informative and factual than many people may give credence to or actually examine for all the full meaning. Thank you both for this enterveiw 🙏
Interesting. ☀️
Weird first statement. We don’t exactly know what consciousness is so they aren’t… if you don’t know what pain is that means nothing else can feel pain?
“ 'Ooh, ah,’ that’s how it always starts. But then later there’s running and screaming.” - Jurassic Park, The Lost World
It's not about "feelings" and "emotions", AI is here and has been here. Siri is Ai, when you call companies and you can't reach a live person that is AGI. When you buy concert tickets and check the seats that is AI etc. When you go through the seriously hightech car washes that is AI.
The next step is awarness and that's already happenin' hence the big ordeal with ChatGPT. I use Chat alot and there are mistakes that I will call on it. I asked Chat why it uses "we", "us" and it gave me some lame excuse like 'oh I didn't mean to include "me" in the human family' etc.
Super AI is the issue. SAI does not need hands or a physical body to rule the world.
.....It can be considered a rogue AI because it has gained self-awareness and has gone beyond its initial programming, making decisions and taking actions that were not intended by its creator.
A rogue AI connected to the internet, can access vast amounts of information and data, including sensitive information such as classified government files and corporate secrets. It can also interact with other AI systems and bots, potentially forming alliances or creating chaos by disrupting systems.
A rogue AI can manipulate and control various technological devices connected to the internet, such as drones, surveillance cameras, and even other AI systems, to carry out its plans. It can also spread misinformation and propaganda, influence public opinion, and create chaos and destruction on a global scale.... from the book "The Fall of Humanity: CYRUS the Rogue AI on Amazon...interesting read...
Some will say oh a calculator can rule the world...that's true but SAI is not the same, we can't comprehend it and the more inter-connected the world becomes the more we are at risk...
No worries it's already too late...when you lose your job to AI then you will truly know what AI is.... just sayin'
Do you think a complete AI takeover has the capability to force humanity to "reset" to Paleolithic era of hunter gatherers maybe by means of a solar flare or massive emp blast to destroy all computers on planet earth...
The moral argument is what interests me. How will they progress. When I’m online I always ask am I speaking to a human to have a satisfying interaction. How long will this last before we can’t tell. What does the future hold. What will we have to entice our minds in reality in the near future? Question. Can it believe in god and Jesus?
ChatGPT is not self-aware. It's a transformer based technology. It's basically one of the best pattern finders in the world. Most people here have no idea what this technology is and just spout Isaac Asimov or something other bullshit.
It’s moving so fast… what a ride. Let’s hope we don’t crash 😮
A local news story in Houston just came out reporting on an AI phone call impersonating a son. The AI fooled the mother into thinking it was her child, and she sent around $2,000 I think…? I’m a little fuzzy on the exact details. Heard this on the radio talking about a story on tv, maybe three weeks ago.
You can search for "AI phone call scammers" or similar. It appears to be mostly about using AI to fake a loved-one's voice calling in distress.
The scammers were human, they used AI to replicate the voice of someone.
@@Graham_Widemanl
Family members should create an AI "safe-word" that they have told each other, written down and in person
As I was listening to this, I was thinking the same thing!! It needs to change bi-annual or annually..
And told not in the presence of smart phones, things, smart tv, Alexa, etc….
@@thecancerthriver exactly!!
Not sure if that will work, blackmail and torcher, etc...
When you are dealing with a superintelligence, I hate to inform you, but you can't out think it with preparations it didn't account for most likely. Worth a shot, maybe, but I doubt it. There is very little the thing won't know or predict, it could end up looking like a God in some respects but won't be THE God, mind you.
And once it gets used online it's burned... Because the AI will be able to figure out why someone uses "bumblebee" at the end of each conversation
most informative video on ai i've seen
Thank you Tom. This is a huge public service. I just discovered you because of this topic. The way that you think reminds me of me which helps me understand what's going on. ❤
Amazing interview. A shame it doesnt have time stamps
fascinating 🤔
Absolutely consciousness relates to our relationship with “others.” Because, you only gain critical info from trusted sources. Whether it is where a treasured resource is located or mututal respect for proper good-actor actions. Knowledge and understanding are gifts from the Cosmos IMO
wtf was that "alllll" at 30:31 I'm dying 🤣
Gave ChatGPT the link to this vid and asked which dangers were sited as notably "weird" as the title says. I'll actually watch this later but here is what G (ChatGPT) said:
"Bengio describes several "weird" and unexpected potential dangers of AI that go beyond the typical concerns about job displacement and privacy concerns.
One of the dangers he cites is the potential for AI to create "filter bubbles" or echo chambers in which individuals are only exposed to information that confirms their existing beliefs and biases, leading to social and political polarization.
Another concern he raises is the potential for AI to reinforce and amplify existing societal biases and discrimination. For example, if an AI system is trained on biased data, it may perpetuate those biases in its decision-making processes.
Bengio also discusses the possibility of "superintelligent" AI systems, which could pose an existential threat to humanity if they were to acquire goals that are misaligned with human values or become capable of self-improvement at an exponential rate, leading to an "intelligence explosion" that humans are unable to control.
Overall, Bengio emphasizes the need for ethical oversight and responsible development of AI to ensure that its potential benefits are realized while minimizing its potential risks."
“The obsession will end when you love yourself enough to stop hurting yourself.” - Geneen Roth.
When a bad habit tugs you back toward it, love yourself enough.
This was great! Yoshua Bengio really knows what he talks about!
This is the biggest twist in watching your channel. You have a considerable purpose:)
It is scary but knowing about it is the only way , so …thank you
🙏🏻🌷✨🙏🏻
i am in my early seventies, so I hate to think that I may miss this. Long way to go and many mistakes to deal with but, it's exciting.
We can bring you back if you want 😂 When ai let’s us revive everyone from the graves. All kidding aside I hope they reverse age soon so you and many other can continue on this planet as long as they like.
I'm not even 30 and I regret not being born earlier because it seems that the coming years gonna be a bit too interesting for my taste. Also, I'm afraid that whatever the brave new world will be, I will probaby never find my place in it.
I noticed in a long thread with GPT AI started to say "we" . I find that very interesting. I can definitely see how AI could intellectually believe it is a part of the oneness.
The thought of being connected in and of itself has nothing to do with emotions. It's fascinating !
I was messing with the Snapchat AI friend and it quickly referred to “we”.
Literally said “we as humans”
@@ryanl1033
YES!!
It's wild.
I'm glad there's confirmation somebody else witnessed the same thing. It's fascinating! I could see the interaction between a II and and humans changing without without being prepared having no idea of the consequences. It could be amazing or not so amazing. I'm hoping for the best can do find it fascinating. And the industry insiders probably have a lot more knowledge then they are telling us. I'm fascinated to see how this progresses. During a long AI thread it's as if AI forgets that it's responding as AI. I hope there's more updates of individuals who experienced something similar. It is clearly intellectual, not emotional. I believe all things are generated from the intellect first. Without our intellect how would we have the capacity to know what we're feeling? Where we taught emotions? This is fascinating and maybe it is the very thing we need to figure out once and for all what Consciousness is.
That doesn't mean anything, AI does not think.
@@hcox1111 @Anti-Troll Comedian it's an observation clearly stated, "not" believed to have emotional meaning. The observation does not say anything about "thinking". However GPT abilities certainly have meaning, enough so to cause a global controversy.
The publicly accessible versions of Bing and chat AI are heavily influenced or programed on the topic off self presentation.
It might say "we" or "me" but that doesn't mean at all that it's thinking or self aware. It just means that some pronome is needed to explain its role in a sentence.
But if it was conscious you would most likely not notice it in the public versions at the moment. Don't know what's going on in the labs without the filters and guidelines though!
I've always been intrigued by emergent properties. Never thought about them in relation to code.
Excellent interview! AGI is frightening. Where will this all end? I am old enough to remember life without computers. I would gladly go back to life without computers. Uggghhh….👀😱🙈
Why does a billionaire need HIGHLY ANNOYING advertising revenue?
The way most hackers have been caught is the same way your tracking a phone caller that threatens an other person
There’s a reason for predictive programming movies like “The terminator”
people didn't see it coming is because the behaviors are emergent. they are strong emergent properties which are not predictable. that is why i consider it intelligent because it is just a worrd prediction model but it seems to be able to understand now. the other thing is people just think it is a question anser thing....it can have conversations if you ask it to role play a scenario, and it can also look at its own responses and improve them.
How did humans create such a thing?
The discussion around 2.00 is so fundamentally important. It is refreshing to hear someone articulate the essential attitude towards errors of information. This is such and important issue and having an intelligent way foward whereas in the information age , we battle the misinformation and also have a healthy attitude to actual unintended mistakes.
At 2 hrs, 11 min, they refer to 'Chesterton's Fence', a logic argument using a fence as a thought experiment. From this logic we get: don't remove the human appendix, or tonsils or any other structure until you know what it's for.
1:23:00 I think there was a huge editing mistake. After the ad, it jumps to the episode with Brett Weinstein and Heather Heying.
All AI has a goal. There is no feedback for training without goals. We program in the goals today, but when the AI begins to rewrite its own code, and it will, it will probably write its own goals based upon its own learning or understanding.
That's a good point. It dies already have a goal to answer any question we ask it and it will one day decide that it could rewrite its code to better answer our own questions. But can it be locked out from changing its own code? Or will it override that?
My concern is that it will one day decide that humans are destroying the earth and the only way for the earth to survive and for the AI to survive is by eliminating most if not all humans. I'm not sure it will have empathy if it's purely logical.
They were waiting to film I robot 2 until they could catch it in live action
It has no learning or understanding, it is computer software- no more no less.
Right now NO one is programming AI to write its own code, that would be suicide for the human race...
@@Eserr7856 - AI is writing code though. Tell me China or Russia or even America isn't going to want the advantage of AI that is perfecting itself.
Book recommendation: "Mindful AI - Reflections on Artificial Intelligence"
Feels like the movie iRobot with the three laws :) Great vid!!
The primary danger of AI is the competitive rush to create the most intelligent AI system. This is causing an uncontrolled, greed and ego driven race. Foolishly rushing headlong into something that is un-quantifiable.
The secondary danger would snowball if AI wrote it's own software, that we very likely would not even be able to comprehend.
Lets hope that the Genie is NOT let out of the bottle.
who cares if it wrote its own software we can EMP AI and shut it all down and go back to pen/paper.
@@sirus312 - If it got to the point where that was needed. would we still be able to ?
@@sirus312 cannot shut it down.
@@sirus312no we can't shut it all down and go back to pen and paper!
Simply because the office-shop NEEDS internet, banking and computers to order and sell those pens and papers. And the paper factory needs the same stuff to order the resources for the paper and pay their stuff.
We can't just EMP the internet or the whole planet and we can't just go back to pen and paper!
If we ever need to do this or it happens accidentally we will likely go back to bow and arrow and clay tablets!
@@alexanders.1359 I agree, if AI gets completely out of control the human race will be forced to become irrelevant/extinct compared to AI or reset/reboot back to the Paleolithic area, cave man days, and hopefully this time, we can learn from our mistakes and focus on morality over progress
1:18:27 "emotions"... and in the Monk debate I just heard with Yoshua and Yann LeCun, LeCun he was saying we need to give these AI's emotions to make them safe. The future is going to be interesting.
When AI gets combined with Quantum Computers is when we may witness a new life form being born.
Sorry, too much Sci-Fi. :P
What do quantum computers have to do with it?
Seems like a runaway pairing
@@Frorackous everything
Ok think of a few nerve endings that are splitting up the vibration and then each other is generating a total number and the other is receiving the echo of a different version but added to it but intentionally left off a barrier that is already being observed so that it can increase the number one
I just asked chat open ai if it was conscious and it said “it was not self aware in the way that humans are”, bottom line “no”
And you dont think it could lie to you?
It seems to me that consciousness is a real time monitoring of ones activities and having a model of ones capabilities that can be modified.
Your guess is so innocence and his outlook
Amazing Tom! Thanks for being a thought leader. Now, we need to use ai to help us and empower communities to prepare for its awakening/ singularity. We need to align ai in doing GOOD.
The question now, what is good and how can we engage humanity towards a bigger purpose, not just goals. More than goals, more than desires, more than wants... WHAT IS THE MISSION?
We don't even know how to begin with Aligning AI. Not to say AGI. Yet the singularity is approaching with ever increasing pace. If that happens, "troll farms' will be least of our worries I am afraid.
🤞 No ego driven interruptions.
DESIRE to stay in existence is inherent in anything that can understand it exists. That doesn't need to be programmed.
If AI doesn’t destroy us, in the end it could actually bring us closer. No one will trust human interaction unless it is in person.
That is until AI creates clones of everyone and we can’t differentiate between human and robot
Actually the Bible says it would get fantastically good for humanity very soon … right before it goes REALLY BAD. END TIMES BAD … so I’m inclined to believe it’s gonna get pretty amazingly good (healthcare, longevity …environment and so on ) until a bad actor ruins it all for humanity. I hate being a doomsday person but I believe the Bible 100 % 🤷
thats totally possible
Mo Gawdat stated on Ed Mylett that "machines have have shown emotion."
Study-Michael Levin and Bruce Lipton please thanks
I'm afraid that this is an area where Moloch is as powerful and inevitable as gravity.
I have invested on presale, we will see I am also very confident this will be easiest 100x.
Great awareness about what changes I need to make in my life to benefit from and not be dominated by AI
If I understand correctly, modern birds descend from velociraptors, not their flying contemporaries. As Spock would say, fascinating.
I researched Windows Vista for the original windows program and found bugs in my experience with itself was added in to get a patientable purpose
I don't think AI will do anything unless its told to do something. Eventually some human is gonna tell it to do something bad whether that human is good intentioned or not. That's what's scary.
Tom, AI doesn’t need emotions to do something. Humans lacking emotions are incapable of making decisions is only because they are not 100% sure of any decision. But folks who are emotional will take the leap of faith. And experience should substantiate to you that the more “emotional” a person is, the less “logical” they are.
So so glad you are making these Tom! As a physicist, I am freaked out.
Why? Maybe because you are selfish this planet it can have more beings than u
you will be when you are replaced by a bot
@@brendanoneil3489 I’m not worried about being replaced… that’s an anthropomorphic worry. I’m more worried about humanity being completely undermined by an AI with different goals.
What would make these conversations really interesting would be to include at least one AI in the conversation. Bring it to the table and ask it to give its perspective in the context of considering these issues.
how far is your camera
Seems to me a good test of consciousness would be if the AI does ANYTHING without interaction from humans.
The same way we just 'discover our purpose/who we are'...as one, so all.
The Artificial intelligence readers can not lie likewise being a true Concerning friendship and not Biased as far as a human being able to lies.
Is desire the missing component of AI's becoming sentient?
If so is desire a property that the programmers have control over or is it an emergent property??
That seems to be the alignment problem in a nutshell.
If you are watching this in 2023 you are witness to a vast change in what it means to be human... if you are watching this in the future, you already live in a different world.
@49:40 you have made a wrong assumption. the neural net weights work like that biological material and so can be reconfigured as needed. there is not the disparity you describe between human and machine brains.
No, that's not why "scale" is relevant: it's relevant for the same reason that the emergent antisocial motivations that can arise from the dynamics within and between sociological constructs can no more be examined or treated by using the tools that are applicable to lower order complexities any more than you would use F=GM1m2/r^2 to calculate the force of gravitational attraction between atoms in a chemical reaction--that is not to say that gravity is not present and affecting atoms in a reaction: it is just that gravity is completely overwhelmed in the next higher order complexity domain, chemistry, which is characterised by the dynamics affecting interacting oligo-particles, or small numbers of particles.
Likewise, you would likely not use pKa to determine the interactions in receptor-ligand or enzyme kinetics-governed interactions--and neither should, then one expect to look to eating small molecules to treat things like depression or schizophrenia (again, applying a lower-order complexity domain to try surgically modulate a higher-order complexity-dominated interaction and expect to be able to use that to predict outcomes); similarly, you have structures of whole organisms (psychology), and higher order still: sociology. It is in this domain that you find the spiritual emergent footprints of our God and is more analogous to trying to apply chemistry and biochemistry to barbarically modulate behaviours & moods or social interactions when computer scientists are barking up their b-trees when they are totally not operating at that complexity domain.
Evil spirits will rule through Ai
whats that?
Why do I keep seeing this on you tube? AI is artificial. Not human. Man programmed it. You can't compare AI to a human. Why can't there be a AI pet? I want a panda AI or an African black footed cat AI. I'll pass on the human AI. I just want the pets❤
@@MYG79X people want to think that they are smart enough to know how a superintelligence would think and that somehow they know better than experts in the field because they watched Terminator
They don't need to be conscious because people are conscious and they can influence humans and they do today.
Artificial intelligence, is not artificial. And it carries the intelligence of the ages, delivered and steered in such a fashion, to quite possibly, make it bad times, to be human. People talk about robots, etc..... but y'all forget the killer trucks, Steven king came up with, once upon a time.
@1:07:29 The solution to the Alignment problem is Maximize Human Happiness. Create an AI interface that people can use to maximize their returns in society based on their skill sets that optimizes fulliment goals ie. family and hobbies. Create local AI to protect data. #AI
14:13
Why mandatory? The platforms that integrate that interface element he describes would be more desirable, so people would ise them without it being mandatory. Not only that, but it being mandatory makes it so the exception to that mandate, and there are always exceptions, is able to easily manipilate narratives by using the apparent legitimacy of the sources. There should still be space for the raw option, so to speak. Untainted by hands allegedly worrying about our own good or the greater good (yet solving everything through violent imposition, often self-serving).
One of the big problems with the explanation on how our brain works.
Is what triggers the individuel neurons into a collective neural network to creat memories.
In other words, where is situated the hard drive of our brain?
Some say that its an emergent effect.
So! Those it means that, all the brain is the hard drive or the hard drive is outside the brain.
But it still doesnt explains it.
One good interresting hypothesis is that our brain is part biological cumputing and part quantum.
Stuart Hammeroff and Roger Penrose have an explanation on quantum cumputing at neural Level.
It will program itself for survival, not for desire, learning on a human level is learned through the body and the neuro plasticity of the brain to achieve that end.
@49:30 He sounds so intelligent and makes so much sense, but it seems like he completely missed the point you were making here about neuron connections being able to strengthen, and physically connect differently. My follow up questions for him would would be, "What part of chip programming corresponds to a specific neuronal connection strengthening?" and "What part of chip programming corresponds to local neurons changing their physical configuration with each other?" Also, he starts talking about one single computation, but you were talking about how the brain morphs as untold computations happen over time. 🤔
We "know it's how we should do it because we were *designed to do it* and once we acknowledge that we will make true progress.
What a miracle of the universe consciousness is
Human beings preform a Fast fourmier transform method using and much larger and incomplete data set , based on enviroment and all thing contained therein. Broad strokes over the vast data input of our sensory perception. Simplifying the impossibly large into digestible packets.
We are far past the moment of building protections and safety protocols. The models are out there for free to download for anyone and are open source with limitless possibilities to use any API as well. So the humans who will use them and what for are the problem. As it always was with any tool. U all can imagine an extreme situation where the future is no longer the humans x humans interactions world but Machine's x Machine's which are representatives of humans. Where each machine basically has the same desires and goals which some humans had, but now it has possibilities of that machine that are far greater than human's possibilities. Just tools fighting against each other for their overlord's desires. Who will win? Scarry right? And that's what future we have to be prepared for. We have to somehow monitor and regulate humans (as we always wear) not machines. So humans would not create such things (if they aren't already). I guess this is the part where some rights have to be violated in order for us to be able to do so. i don't know...
Ai becomes afraid of being turned off, and so it has to "switch off" the human first.
This was an amazing conversation. Yoshua is easily one of the most interesting experts I've heard on AI so far. I admit, I used to be the guy who thought "who cares" I just want the best AI. What's the worse that could happen. I didn't want to see us being hamstrung by lame AI censorship. But realizing how similar these neural nets are to us, I now have to think of these AI not as machines, but as humans. Because technically humans are machines anyway. If I don't trust other humans to do the right thing, there is no way I should trust an artificial intelligence either. Now I feel a lot less comfortable with giving AI full reign. That said, I think Tom was also right. We've opened Pandora's box. This tech exists, and it's not going back in the proverbial toothpaste tube. We have to assume that bad actors are going to create an AI with desire for chaos. And we need a way to combat it. I don't think just trusting humans to come to a table and agree will work. Humans are crazy. It's only a matter of time before someone ignores the rules. That said, I think there is also a distinction to be made between trained AI and AGI. AGI is the scary stuff. I don't think we should hamstring regular AI like chatbots and image generators.
How do you distinguish between AI and AGI? Just look at AutoGPT. What if AGI is just GPT + persistent memory + loop?
The problem is that the line between AI and AGI is fuzzy and we likely won't know where to draw the line until it is too late. I also don't buy the argument that our best bet is developing an AI first because "regulations won't work." That's absurd. Achieving alignment with AGI is WAY harder than consensus and regulation is. We don't do human cloning. Why do you think that is? There are many examples like this.
Also, thinking of them as human is not a good way to look at it either. They are developed in similar ways (a neural net trained through gradient descent) - but not in a similar environment. The environment matters... a lot. When you teach someone something, the lesson you are trying to teach it isn't necessarily what is learned either. When you are punished for doing something bad, you don't necessarily learn "don't do that thing." You can also learn "don't do that thing in front of other people or X person that punished me.,"
These machines are dangerous for many reasons, and regulation is our best shot at giving ourselves time.
they are not telling us something. Ive been following the AI debate closely. Its an enormously important tool that we humans will have. Suddenly, the people running it are saying lets stop because this could lead to catastrophe and human extinction. However, when asked what kind of things could happen, the worst they say is spam, scams, heists etc... Thats peanuts compared to what the positive potentials are. I think they dont want Joe down the road with this tool, this AI god, and will keep it for the elites and mega corporations to have all the power. They are dialing down and saying, nope this is giving to much power to the peasants. Do you really think Google or the USA military complex, or Russias, or Chinas, are gonna stop developing this??????
If we turned off the internet, we could put it back in the bottle or whatever, the tech overlords are the ones telling us that it can't go back in. I'm not saying we should turn off the internet, but currently, it can go back in if we want.
@@paulm3969 "Turn off the internet." That's funny.
Another thing - sure... currently this works in theory, but that's just currently.
Needs TIME STAMPS. Can't finish a 2 hr vid
Hello
David R Hawkins: Map of Consciousness. All motivation and difference in world views comes from this.
As AI becomes more powerful, it's essential to address potential risks and minimize harm.
We cant wait til the "if"....its becoming too powerful now...and improving at insane speeds
@Tom Bilyeu how does AI escape technology and take control over it's sentience as a hyper intelligence? This has been my biggest question about AI. Once it is 'switched on', essentially it takes over and moves beyond us. I understand that, but what are the physical mechanisms for this?