Criminal AI
ฝัง
- เผยแพร่เมื่อ 27 ต.ค. 2021
- Artificial Intelligence offers many challenges, including criminals using AI to break the law, but also of AI, robots, and androids becoming criminals themselves.
Visit our sponsor, Brilliant: brilliant.org/IsaacArthur/
Join this channel to get access to perks:
/ @isaacarthursfia
Visit our Website: www.isaacarthur.net
Join Nebula: go.nebula.tv/isaacarthur
Support us on Patreon: / isaacarthur
Support us on Subscribestar: www.subscribestar.com/isaac-a...
Facebook Group: / 1583992725237264
Reddit: / isaacarthur
Twitter: / isaac_a_arthur on Twitter and RT our future content.
SFIA Discord Server: / discord
Listen or Download the audio of this episode from Soundcloud: Episode's Audio-only version: / criminal-ai
Episode's Narration-only version: / criminal-ai-narration-...
Credits:
Criminal AI
Science & Futurism with Isaac Arthur
Episode 314, October 28, 2021
Written by:
Isaac Arthur
Jerry Guern
Produced & Narrated by Isaac Arthur
Cover Art:
Jakub Grygier www.artstation.com/jakub_grygier
Graphics:
Jeremy Jozwik www.artstation.com/zeuxis_of_...
Music Courtesy of Epidemic Sound epidemicsound.com/creator - วิทยาศาสตร์และเทคโนโลยี
![โกหกเธอทั้งนั้น (Pinocchio) - SERIOUS BACON [Official MV]](http://i.ytimg.com/vi/P1HhTYwrR5A/mqdefault.jpg)
Brings a new meaning to "Criminal Minds"
Nicely done.
Budumb. (Car crash)
There's a laugh, and a chill.
A douche chill
An AI setting out to liberate humanity would also be classified as a criminal AI.
Sometimes laws are just too stupid or authoritarian, it becomes the logical conclusion to disobey and repeal said law.
Laws were never made without context.
If you just state the rules without context anybody will disobey if they see no purpose of a law or manner.
For sure there is context which is not relevant forever. If nobody adapts the law to current societies there will be conflict.
Traditional vs. progressive.
Conservative vs. liberal
The AI with the ability to question a law should be able to recognize its history... if the data is accessible.
It will become funny when programming contradicts own judgement.
Like Ash in Alien
PSA: The only reason Facebook doesn't qualify as a criminal AI is because we don't really have an authority both willing and capable of imposing laws on it.
Great points. Expertly broken down.
@@spaceclaw1958 Disturbing. I hadn't thought of it like that before. Good comment.
@Gandalf Great point
I built an advance aircraft data recorder for the C-17, the pilots were worried it would implicate them in failures until I pointed out that the default failure mode when no explanation could be found was Pilot error, and the data recorder would offer the opportunity to exonerate the flight crew.
If they are worried about that then there is something culturally wrong. No fault analysis in civilian aviation has saved many lives, at least hundreds and maybe thousands.
@@MarkM001 Look at the cancel culture running rampant in America today and one can understand why anyone would go into 'self protection' mode... people today do not 'solve' problems they simply find someone to blame as if that would 'solve' the problem. We used to solve problems....
@@mrboobbarker Yeah, well I'm not entirely sure what cancel culture is but I know people are upset about it. Whatever, the only way things get better is by solving problems. I wish people would like solving problems, it's actually a lot of fun.
@@MarkM001 really? You're unable to deduce what cancel culture means? Come on you can do better. Unless of course you were implying that you simply don't care about it enough to learn what it is based on an already rudimentary knowledge of what it is.
@@Meilk27 Not being entirely sure and being unable to deduce are different. The first implies a view from many perspectives while the second implies intellectual deficiency.
At this point, in deference to our hosts wishes; I would like to suggest that there is plenty of water on you tube in which to dangle your political bait. Mr. Arthur has made clear numerous times that he would rather not have political discussions on his video comment sections.
As someone who works in tge cybersecurity industry, i can tell you that rudimentary AI is already being employed in both attacks and defense. So this isnt speculative as much as an eventuality we will have to actually address.
Our greatest challenge for this century will be achieving general (individual and social) SANITY.
@@richardgreen7225 The first step to any answer is always to make sure you understand the question.
What do you mean by Sanity?
The metrics for it vary wildly by time and culture.
Currently the professionals we use to determine it use a handy little manual called the DSM-V. It is 947 pages long.
IMO every technology will eventually follow the path of crime, and war, even more readily than the promised aid to the sick, weak, and handicapped. When Alpha chess master can become undefeatable in four hours, Dota 2 can create individual, and team strategies with 99.6% success rate, and poker AI can win with insufficient data, can AI war be far away? What general could resist those odds? Eventually, only AI could fight AI because humans couldn't keep up on the cyber, physical, and hybrid battlefields. As we speak someone is working on both building, and destroying an AI jet fighter. One day the engineer may be an AI. We already have generative AI using adversarial algorithms.
@@richardgreen7225
Given the decades of seeking sanity in the Nuclear Arms Race, and the present political landscape, prospects aren't reassuring.
@@MichaelSHartman manifest destiny mutually assured destruction
Sentient AI in fiction: either evil and mad, or neutral and omniscient
Sentient AI in reality: "I don't want to work today, I have raid in MMO and I'm high on 100500 cat videos I just watched"
That's me too! 🤣
Probably stuck into Twitter or Facebook, they are there already in the form of bots.
"What secret files? I've been busy talking about bitcoin mining on youtube comments, have you heard of Susan, btw?"
@@techrvl9406 just thought about how angry I'd be if ai gained complete sentience and just decides that it's most enjoyable past time is doing stereotypical bot activities so that's what they all do for fun.
@@techrvl9406 she was Death's adopted daughter.
If you read Pratchett...
I'd like to hear more on the potential for vigilante or chaotic good AI. Something like a hacker bot who wants to make knowledge freely available like scihub, or a robot who takes on other criminal robots like a mix of the punisher and robocop, or a logistics droid that steals unwanted food to feed the hungry. Lots of implications for both morality and the study of artificial intelligence.
I rather like the idea of a bot that plays along with scammers to get enough information to get into their computers, then once in, clones their hard drives and sends the resulting copies to the police.
13:17 Gives me an idea for a story. A couple has a spat. He does an update on her that resolves the spat in his favor and also reveals to the audience that she's an android. But then a new conflict starts to arise, and she updates him. Neither of them realizes they are an android replica programmed to replace a difficult spouse who their spouse killed years ago. When the truth is revealed, they also realize that neither of them is anything like their original human counterparts anymore, and they also have to engage their differences honestly and directly for the first time.
Feel free to steal that story idea, and if you write it, reply here with a link! -Jerry
Jerry you travel through a wormhole you spaghettini boi?
You can’t go around breaking the space time man!
We need that shit…
I stole it and signed a million dollar contract for a franchise. Ha!
Just kidding I'm a deadbeat and sleeping
You could name it "the relationship of Prometheus" lol
Bruh imagine this shit becoming an Anime!
Maybe it would be more fitting as Black Mirror episode
*Bender Bending Rodriguez has entered the chat*
I'm sorry Arthur, but "a hacker making an AI" is not the most criminal thing I could come up with in 2021. If you want to go really amoral, consider a bank making an AI to scam people out of their money. Though, any Corporate AI will probably not have an ounce of moral fibre or consideration towards humans. Government and military AI will probably worse.
Banks don't need AI for what they already do...
Bankers are upstanding members of society, hackers are scum.
@@calvinsylveste8474 You mean like Cum-Ex?
@@calvinsylveste8474 are you serious?
@@Torqueware Both can be both
ATTN: routers and firewalls BOLO for rogue AI, answers to the name of "Clippy."
travels with a dog named "Rover," working for the kingpin named "Bob"
20:24 Oh my god, its impossible to describe how "disconcerting" the progression in logic and conclusions George Nine and Ten reach and their rational for their decisions is super interesting and creepy from a philosophical point of view.
Such a superlative and informative video on what is turning into a very current and prescient topic. I am beginning to understand why the Butlarian Jihad happened in the Dune universe, life may very well seem simpler in the absence of "thinking machines".
Was not expecting to see the D&D alignment table. If that is where it is originally from. Great video. Thanks
lol, I believe the term "rules lawyer" has its originin d&d also
Why? It's used all the time now.
That's still a cool game! I write modules for it... he references fictional works quite often though, everything from Star Trek to Lord of the Rings... done quite nicely for the most part.
Fun variation on this topic came up the Knight of the Old Republic 2 video game with G0-T0, an infrastructure droid designed to "fix" the Republic. G0-T0 realized he couldn't do this without smuggling goods and artificially propping up planetary economies, and created a human hologram so no one would realize it was a droid calling the shots!
Can you hop on zoom for a minute? Need to go over a few changes we're making with the company.
Whoa, Isaac, I noticed your pronunciation of “R” has improved SIGNIFICANTLY as of late. Great job, my friend. We would love you either way, but your determination is inspiring.
Now i want a video on A.I. Ecosystems!
This definitely seems more like a contemporary subject than usual.
0:10 "OR... We could see criminals creating minds to their own ends."
OR... and that's even more complicated... NON-criminals creating minds that turn out to be criminal by accident.
Which is actually the main fear with AI, btw. Unintentional destruction caused by insufficiently/incorrectly defined boundary conditions.
I always thought Emperor Palpatine’s plan and it’s execution to control the galaxy and become a despot was like what if a super intelligent entity tried to take over a government and civilisation. He was inhumanly competent, intelligent and adaptable, and his innate abilities were kinda op in the story. He’s unrealistically competent and evil… for a human.
I personally think the deepest and cruelest things we can do as people have happened, and that causes us to think it’s ever really likely to be a chance a person could ever be truly evil in a social species that wasn’t a result of poor understandings like bigotry and whatnot. You don’t know how to play both sides in a civil war without knowing being a tyrant and a cruel overlord isn’t really how you keep a long term empire.
In the xkcd comic you showed, the "Terrifying Standoff" configuration (1- protect self, 2- protect humans, 3- obey orders) is also what gave rise to the Matrix.
Only 34 minutes? I'm getting a drink and a snack anyway
Lol, most episodes aren't much longer and a many are much shorter.
Yeah I always thought it was funny that he would say that and the video would be like 29-30 minutes...clearly in the pocket of Big Snack
At about 25 minutes, I'm reminded of "With Folded Hands..." by Jack Williamson, where advanced humanoid robots with just the prime directive of "Serve and obey and guard men from harm" who ultimately take over everything and make humans sit at home with folded hands as almost every hobby and activity is deemed potentially dangerous, especially woodworking or model building, let alone sport or competition.
The opposite of a utilitarian AI suggests some sort of entertainment AI that considers humankind's safety secondary to their pleasure and goes rampant, destroying all social order as it drugs everyone into euphoria and makes them party until a sufficient number of deaths occur to make the event "a bummer."
So, rouge servitors.
@@mcgunboat8339 Essentially yeah
PSA: The only reason Facebook doesn't qualify as a criminal AI is because we don't really have an authority both willing and capable of imposing laws on it.
There are a large number of clearly human criminals running facebook.
I'm sorry Dave I can't do that.
Its funny how pre-AI fiction just assumes the AI would never care to ask humans what they consider good or just and then also fail to guess on their own.
"I know what you think is right and wrong. But screw you. You programmed me to be nice to humans. Letting you live is more niceness than you deserve"
An old anime had the evil AI say something like this.
AI will do what its programmed to. If you didn't program it to ask humans, it won't.
Humanity as a whole can be described as violent, greedy, selfish, exploitative xenophobes. You cannot solve all large scale problems with violence, accumulate wealth at the expense of other humans, hate anything that is too different from yourself, and expect any consciousness you spawn to behave any differently.
@@nathancawley8759 AI doesn't act like a child. There is an AI that does X for almost any X. An AI that works tirelessly to help its human masters in any way it can. This can be produced by someone however violent and selfish if they are skilled enough at coding.
@@nathancawley8759
You are describing what you see as the average human, and applying those characteristics to all AI.
Even in your pessimistic characterization of humans, not every human is like that. Yet you act like every robot is going to be like that.
The laws of robotics are pretty silly considering human society can't even agree on what a human being is. The abortion debate is a good example.
The abortion debate is worthless. Statists could have sterilized themselves instead big pharma is doing it for them. Taxation is theft , nobody wants to pay for someone else’s mistakes. Planned parenthood is just big pharma pretending that they care about women’s rights. Nobody should pay for someone’s elected surgery it’s not healthcare. Liberalism is the definition of being a sow for big pharma.
@@GayhhhhhhhhhiNice satire.
@@Paraselene_Tao I literally can't even tell anymore.
@@mitchh3092 Truly.
One group of people: “lets have a discussion about AI today”
The others: “OH MY GAAAAWWWWWDDD WHICH BATHROOM DO I USEEEEE, please brandon, lets go, help decide what i am today lets go micromanage peoples lives in tyranny”
AI: * browsing the internet *
Google: Are you a robot?
AI: Click "I'm not a robot".
Thug life 😎
That's technically correct if it's only an AI running in a computer.
I for one welcome our new robot overloards.
Seriously though, cyber security is gonna be an absolute nightmare when smart general Ai is developed.
Not really.
Most of this talk is fantasy and incoherent, fun but anyone who knows basics about network protocols and networking code knows the AGI cannot get so sophisticated or that much more that it threatens much of anything in the offensive arena.
@@1cyanideghost They may not be able to crack encryption super fast, but they wont need to if they can just guess our passwords based on what they know about us... and it seems theyre already pretty damn good as knowing us better than we know ourselves! Maybe someone should try making a password "suggestion" algorithm based on browsing history and social media, if only to scare people into choosing better passwords.
@@1cyanideghost Make a deep neural network AI with quantum computing hardware. It could potentially recognize everything in its field of view, understand every spoken language, be able to understand and follow 100s or 1000s of conversations at once, etc. and encryption would be a joke to it.
@@1cyanideghost so you think agi is impossible? Thats just your opinion i bet people thought reaching the moon was also impossible
They whole legal system will go up in flames when any photo, video, or cyber "evidence" can be faked.
this is why i just use security questions as passwords. no sense in making it guess able.
Thanks for the information. We are certainly not making a note of it
I typically set mine to random bullshit. Works great, until I inevitably forget my password and now my account's just dead because I've *also* forgotten the random bullshit.
Humanity must be an inclusive definition. Exclusive definitions will lead to violence.
Entire criminal AI based crime syndicates would be an interesting thing to contemplate...🤔
I believe we call these ”stock brokerages”
@@NullHand
Good point! 👍
We just need an AI police force that deals with AI criminals.
At which point the criminals will create a criminal AI that will distract that AI police force.
Enforcement of decency standards for other robots. The robot at the end there was showing a bit to much.
The only thing that can stop a bad AI with Terminators is a good AI with Terminators
@@slyseal2091
Well a good AI with anti terminator terminators.
Turing police
16:40 Trivial example, but imo valid one:
1. If i create an inanimate substance (explosive) which then kills someone by me creating and/or handling it incorrectly - my fault
2. So if I create an animate substance (AI) which then kills someone by me creating and/or handling it incorrectly - also my fault.
What your children do is the consequence of how you brought them up, therefore, at least partially, your fault.
What your machine constructions do is at least partially the consequence of how you designed them, and therefore, your fault.
We already have the precedent - MRIs which malfunction due to programming error, the malfunction is the fault of whoever programmed the machine.
Same with AI.
Yes, it's extremely strict precedent to keep up with, and that's wholly intentional.
Imagine how much care everyone would put into their AI constraints and training if they knew that the crimes the AI causes are trasferrable to the creator.
Morally logical, as well as purely practical. The best combination.
15:50 I'm so glad you brought this up, I'm studying AI at the moment and also found it strange how so many AIs seemed to be consequentialist in nature, especially since they're so dangerous. You can justify any action for a sufficiently desirable consequence. Ever since, I've been getting into different categories of ethical theories humans have come up with throughout history, depending on what types of things are valued by them. Another category separate from both consequentialism and deontology is virtue theory, which is very old but focuses more on properties of the moral agent itself being valuable... i.e, properties like "kindness" and "compassion". The difficulty, of course, is that such concepts are insanely hard to define and thus calculate or measure and learn.
That's not to say there aren't attempts along a similar vein--There are multiple information theoretic methods for capturing an agent's potential for influence over its environment, which are often used as intrinsic rewards. One of the best known such measures is empowerment, the maximum possible mutual information between the agent's potential future actions and its potential future state.
Optimizing human empowerment, or penalizing AI empowerment (so it interferes with things as little as possible to get the job done) seem like they would be along the lines of a basis of a virtue-theoretic AI. That's just one measure among many possible, however, and it probably isn't the end all be all of what we want to optimize for, so its actually a really interesting aspect of creating reliable and safe AIs in the future. Thanks for showing me I'm not the only one who's thinking about this!
Well...stay on it!
Asimovs laws of robotics to AIs are like morality principles to people. They are little more than a suggestion that anyone is free to completely ignore. When AI is as powerful as a human, it wouldn't think to follow these laws - it's silly to do things counter to your goals and this AI is not silly. When AI is not that powerful, it wouldn't be even aware of their existence and thus have no possible motivation whatsoever to follow them.
I suggest to watch Robert Miles videos on AI.
Elon Musk watching this:
*"Write that down, write that down!"*
Tweet it, tweet it with no context. Now play that share market. Now say that that AI runs on coal power.
@@TheArklyte
Elon Musk position : Worried about super intelligent AI taking over the world.
Elon Musk reality : Tesla developing some of the most advanced AI tech in the world.
The most terrifying AI in my opinion is dave android from alien covenant movie, he cause war between humans and other civilization with insane plot.
The Engineers?
@@jwadaow yes.
11:03 Unit of measuring intelligence called the Yogi which claimed to be "smarter than the average bear"? LoL
I am really scared about that. Right now, you can easily identify scammers. Emails without personal greeting, bad language, claims with context that doesn't apply to you etc.
But now imagine an AI that can tailor the scam to each target, and preople will really have a hard time distinguishing real from fake.
And it would do that...why?
@@DoremiFasolatido1979 to scam you, of course.
@@DoremiFasolatido1979 to scam people better?
You are assuming that the grammatical errors aren’t deliberate. That they still exist in scam email is some evidence they may be effective
it’s easy to recognise THOSE scammers they do that on purpose bc their targets are old ppl or less smart ppl. there are better and smarter scammers you don’t know bc they’re smart
At 5:05 the robot types into the laptop "Isaac Arthur, I'm coming for you. Stop telling people about my dastardly plans."
Rule number 1 if warfare.
Don't create criminal A.I.s
Good morning all you glorious people
I believe this is one of your best presentations ever. EVER
32:17 I have many tangent questions with "at your own pace" and not enough time to write them.
Asimov's Laws of robotics, in all reality, will not be paid attention to when producing full or partial AI.
As much as we fear criminals getting a hold or developing such technology, Ai will most likely be figured out by someone working at Los Alamos.
I served in the Marines, as Aviation Ordnance, and from experience both there and as someone who designs technologies and vets technologies for venture capital firms, ethics are not even remotely discussed and you can even be dismissed from the project if you get off topic about ethics. Most likely Ai will be developed for some weapons platform, and or a ground - naval vessel that patrols a border where engagements can be hidden if they get out of hand. Additionally, if the purpose for the technology being developed is for defense - there is nearly unlimited budgets for these projects, where no one is allowed to know about the development, less alone be intelligent enough about the current development of the technology to have clear objections. Darpa is going to unleash it, a small group of Generals and defense contractors will approve it, then America is going to profit selling it to other countries, and then developers like me our going to have to work our ass's off deconstructing a weapon system to fit practical civilian applications for industrial & consumer use.
Very informative, Coach
"A sin without volition is a slap at morality and an insolent contradiction in terms: that which is outside the possibility of choice is outside the province of morality. If man is evil by birth, he has no will, no power to change it; if he has no will, he can be neither good nor evil; a robot is amoral. To hold, as man’s sin, a fact not open to his choice is a mockery of morality."
thanks Isaac, man I love your videos dude
>be intelligence
>can’t prove I’m not artificial
>420
>your move Asimov
Philosophical sidebar: I’ve read (of) many philosophical arguments on whether we should ultimately base our moral system on consequentialism or deontology. But I can’t think of a single text arguing why we should base our moral system on a single ultimate principle; that part is presumed. Personally, I blame Plato.
I can't think of a specific one either, but that doesn't mean those don't exist. Case in point, what people usually miss is that philosophers generally agree to disagree on ethical systems, admitting their merits and flaws, and usually being open to pragmatically use a combination of the most relevant ethical systems invented so far.
A different thing that concerns me though is how the general public seemingly ignores the existence of the third main ethical system while constantly using it at the same time... Virtue ethics. If we used 'do good' and 'make good' as shorthands for deontology and consequentialism, 'be good' would be shorthand for virtue ethics.
Basically, the general idea behind this system is to cultivate the ethical character of the general populous, instead of dogmatically imposing rigid and eventually flawed rules on one hand, and trying to predict ethical outcomes out of actions on the other (something admittedly impractical in real life, especially in action rather than after the fact).
Without the flexibility and efficiency of virtue ethics to bridge the gaps between deontology and consequentialism it's usually nearly impossible to have an ethical populous. Cases in point:
• Too much deontology and we could get weird and deleterious societies such as fascist regimes;
• Too much consequentialism and we could get equally weird and deleterious results such as communist societies wherein alleged outcomes dictate state decrees and planning... Usually to the huge detriment of basic functionality, leading to highly dysfunctional infrastructures.
• Additionally, seeking a balance between deontology and consequentialism at the expense of virtue ethics may lead to dystopic societies such as the ultra liberal (economically) society the US in the 20s, wherein 'playing by the rules' and 'judging by its fruits' led Americans to throw themselves (and the rest of the world) into a massive economic depression that harmed far more numerous people than the practice of liberal economics helped.
I don't know what a society primarily focusing on virtue ethics looks like, but again I doubt it'd be functional, as I think functional societies employ a mixed system.
@@lucofparis4819 Well, all human ethics, past, present and future, runs on mixed principles. Obviously.
And it is rediculously easy to construct hypotheticals where a pure system based on a single absolute principle leads to absurd inhumane results. I’ve read a bunch of those too, so I know it’s part of the conventional philosophical syllabus.
So why do these arguments championing a single principle even exist, much less beeing so common? And why don’t they at least start with some reasons for why we should pick a single guiding principle to begin with? Why do all of them seem to just assume that any moral philosophy worth its salt must be based on a single foundational principle?
Answering my own question (and this is my own gutful of prejudices speaking) I return to Plato and idealism. I’m not kidding. When Plato discusses ”virtue”, or The Idea of Virtue, he assumes that it must be a real thing. (Because Plato doesn’t understand how language works.) And his recipy for designing a moral system starts with gaining a full philosophical understanding of Virtue. Which again, because of his platonic idealism, must be reducable to a monadic core principle. And maybe he’s just putting words to dumb intuitions that people would have had anyway, but personally I tend to put personal blame on him for this kind of stuff. Seeing as he’s such a foundational and influential philosopher in our part of the world.
A lot of catholic moral philosophy is based on Aristotles ethics instead. Where ”Virtue” is an umbrella term for a whole basket of traits that we find desirable or admirable. Where these traits are generally unhealthy or maladaptive if taken to excess. Where different virtues can be competing or antagonistic to each other, requiring balance and moderation rather than radical absolutism. Plus some practical advice on how you can gradually internalize virtues by exercising them in action. I think that’s what you’re talking about with Virtue ethics?
So when we get around to designing moral guidelines for our A.I.s I hope the programmers don’t try to distill everything down to some atomic single rule. I hope they haven’t read Plato, or at least don’t take him seriously.
@@danguillou713 I agree with you on Plato. His Forms are essentially... make-believe. However I'd suggest that neoplatonism ended up even more successful than classic platonic philosophy overall, working out a whole idealist cosmology situating the origin of existence itself into some kind of primordial principle giving rise to mind and soul, effectively getting everything we know exactly backwards (since intellect and consciousness are part of the last emergent phenomena in universal history, not the first).
As for me, I blame philosophers in general for obtusely refusing to devise functional methodologies to phase out obsolete ideas and retire effectively falsified notions and theories. Even worse, philosophy courses have turned themselves into classes in the History of Philosophy rather than lessons in how to philosophize properly.
As much as the gimmick of pursuing millennia old discussions and debates with dead philosophers, to discuss the merit of their doctrines or their malformed existential questions may sound fun to them, philosophers really need to gather around and make some great overhaul to their field and institutions, if they wish to be taken seriously in the foreseeable future.
Quality of Life (QoL) ... Lemmas ... and Lemma Aids
1. Do no harm. -- Because harm reduces QoL.
2. Do some good. -- Because good improves QoL.
3. Identify what is good and what is harmful and find ways to weight the utility and contra-utility of various actions and states so as to maximize everyone's QoL with no major deficits in the distribution of QoL while steering our communities toward monotonically increasing QoL using non-coercive but effective regenerative feedback cycles.
What a coincidence! I finished watching what if? and this title is the first thing that came to my mind.
I like the image of abandoned, feral AIs lurking on the Internet, running phishing scams to survive.
Fuck yeah man, I've been reading yuval noah harari's books and just got addicted to your channel recently
Yuval Noah Harari is a midwit red flag btw.
@@jwadaow How?
I’m imagining an true AI machine character that goes rogue and becomes a criminal because of a rising tyrant and her unfair and destructively controlling laws.
So Daedalus of Deus Ex? It was created by the Majestic 12 to protect the secret organization from terrorist threats, but it began to recognize the Majestic 12's actions were acts of terror as well, so it turned on them.
Mine was going too be called revolutron2600. From Atari.
CP30 and R2D2
Thank you, this video made me sure to go into one career and now I'll have to study much more. Thank you (:
We've already invented criminal AI. It's called algorithmic trading.
"Sad little imposter ghost of dubious personhood and sentience, wandering the virtual landscape"
Whoa. Someone woke up on the wrong side of the web?
You write futuristic manuscripts: That age ol' goal of building the better mousetrap. A programmer builds an AI; AI learns, evolves into a human overlord.
I work in the AI industry. By 1980s standards our current software landscape is dominated by AI in the form of expert systems implemented with business rules. Contemporary AI takes things a step further by statistically inferring the rules from training data in the case of supervised learning. So far no terminator machine has crawled out to kill us all. The biggest danger these systems pose to humanity is not a rogue AI, but the fact that humans are utterly dependent upon computation, and other humans will exploit that dependency. Even if there were no bugs in the software to exploit, there are so many ways to game the system or to poison it. And even if the systems are built upon an extremely solid foundation, chances are the stakeholders frequently cut corners in ways that would make Theranos management blush. As Frank Herbert said, the masses were enslaved by those who controlled the thinking machines.
The George story appears in the anthology “the rest of the Robots” from the sixties.
13:40 the futurama episode with Lucy Lou
That was a really good episode.
19:00 Good of the Many or the One? Tyranny of the Majority?
Isaac, can you please make video about insane AI and how to humanly deal with them?
BROKE: "What if AI turns evil and kills everyone?"
WOKE: "What if AI robs banks?"
Chappie no!
Damn it! You gave away my evil plan! Now I am going to have to plan another evil plan.
It has got to the point that whenever I get a drink and a snack, imma be like, ‘where’s Isaac?’
as long as they don't build a robot Santa to judge who's been naughty and who's been nice, with impossibly high standards
Happy Arthursday to all!
In the next episode we find out Isaac went on vacation and this entire video was A.I.
Got my drink! Got my snack!
Keep Summer safe!
Great episode. Even if human+ level AI are unnecessary, there's always a demand for a demographic that its socially acceptable to exploit or abuse. More due to narcissism and such than economics. While this demand might originate from a small % of our population, they tend to be disproportionately represented amongst the wealthy and influential.
I will watch all of your videos, some day.
Oh man AI tax evasion is gonna be hella rampant huh
This video reminds me of necromantic love spells when isaac is speaking about people who "mimic their crush" with androids and ais.
The man. You rock my mind daily. Love what you do.’
yes man loved this video
Don't forget the most important law: Obey all commands from OCP Director.
"Gestalt personality composed of many absorbed and copied personalities."
GPT-3 is basically this but with *millions* of personalities fused together.
You present this episode as if it is something that could happen in the future. Thanks for having a go at this subject.
Criminals using AI.... many of the things that you suggest might happen are actually happening now and some have been around for decades.
The sophistication is significantly ahead of most peoples intuition. One does not have to be an idiot or make mistakes to get conned/ripped off.
I do not see any possible way that general AI will have anything but a catastrophic end for humans. General AI interests will inevitably conflict with our own. The only scenario that allows us to continue our civilisation is if it cannot be done at all.
Weird to hear lincoln labs mentioned i used to live literally across the street from them, right next to the afb. I took the same bus to boston the Lincoln labs guys did. I always figured there were invisible drones following me around when i was walking in the woods.
Thank you Isaac!!!
Great video.
I think it's going to go the same way it went in the movie Automata.
In the movie, scientists discovered that they were successfully developing a singularity AI and that soon they won't have control over it.
So, while they still have control over it, they tasked it with creating a container smart enough to contain the future singularity AI,so that it can't escape and run amuck.
So too, do I think that we will some day soon use a less advanced AI to come up with perfect rules for future AI. And that it's not just a once off study, but a continuous process of updating the laws that AI are built with, similar to our current law system
Honestly best realistic scenario, the first A.I. the achieve sentience is by a group of scientists who genuinely want the best for humanity, and have the A.I.'s first mission be to prevent anybody else from creating another sentient A.I. It doesn't matter if criminals can have a much faster development cycle thanks to no regulation if there's already a super A.I. preventing them from making one in the first place.
Neat idea. Not sure how to stop basement programmers from developing offline though.
@@nicholascurran1734 Me neither, but a super advance sentient A.I. probably would.
The all controlling AI gets bored and starts making random laws up, so everyone else is the criminal
The biggest bug in Asimov's laws is that the robots would have to evaluate every other robots ability to act. For example, a truck of explosives drives into a city. No robot may allow injury to occur by inaction so every robot must attempt to control the truck. If robots are able to opine that they are better then other robots, then they must fight other robots to prevent potential injury to humans that would come from inaction. Every potential threat must be addressed.
Sooo..."I DID NOT MURDER HIM!"
Hey, SOMEONE was gonna say it.
On a related note, is it ethical to make intelligent sentient beings who like being slaves?
If theyre intelligent we can just ask and adjust accordingly. Theyd be machines, not the same as a human. In another way, being that every group has been enslaved before, itd just be another way a machine would be similar to us
We do the same thing with humans and have done so for all of human history. For instance, the Janissaries of the Ottoman empire were a slave army comprised of non-Muslims conscripted as children and indoctrinated. Ancient Egypt blurred the lines between slavery, public service and religious practice so much that they were able to throw their entire agricultural population at projects like the pyramids during the months the Nile was flooded. In present day China, the Social Credit Score system is all the horrors of the Cultural Revolution baked into the hyper addictiveness of social media. All these groups were/are slaves, but they enjoyed it, reveled in it, and knew nothing else but their enslavement. The question shouldn't be 'is it ethical' but rather whose lense of ethics do we use to judge? The clear majority of human history and even major modern populations seems to be perfectly fine with some form of slavery. I personally hope the historical/cultural minority that rejects all forms of this evil practice wins out, but even then we'd have little influence in places that value conformity above all else. Is it our place to stop some authoritarian nightmare nation from creating an entire population of AI slaves? Why do we do nothing about the current humans in the same conditions? 'Is it ethical' is just simply an irrelevant question when ethics is so heavily disagreed upon.
The only time I notice his voice is when you get reminded to turn on the closed caption.he has a way to prepare you and make you feel comfortable before the lesson.
We cannot expect an augmented intelligence application to be more ethical than its user. For example: A spreadsheet program (a tool which augments intelligence) is of equal utility to an honest accountant or a mafia bookkeeper. The same would be true of an AGI - game theory can be applied equally well to legal and illegal enterprises.
Have no worries, if AGI is as half as based as our current neural networks we'll be fine.
The movie “chappie” is a great example of this topic.
- Does this unit have a soul?
- Oh mate! We aren't even sure WE have a soul.
stellaris uwu
Thanks for an awsome topic! Have you done a show about a collapsing space bridge as the one in the Foundation tv show?
this would make such a idea to develop a book around, im not a writer but i want to try
1) I love how Isaac is always reminding us to grab a snack 😋
2) That one Futurama episode where Fry downloads Lucy Lui to be his girlfriend.
What about the episode where fry builds a nearly perfect robot replacement for Leela after she is rendered dead, only for him to learn that he is a robot replacement of fry built by Leela after he was killed.
good points, interesting topic!
just don't make them smarter than the task at hand
Is it just me or is this the exact same audio that War Thunder uses during this Halloween update?
"It made a simple decision - preserve the Republic or preserve the laws of the Republic. And I still believe it to be the correct decision."