What We Get Wrong About AI (feat. former Google CEO)

AI is here. ... Okay, now what?
Sign up to Milanote for free with no time-limit: milanote.com/cleoabram
Subscribe to support optimistic tech journalism: kzbin.info?sub...
Everyone’s talking about artificial intelligence and machine learning. There are lots of names thrown around - OpenAI’s ChatGPT and DALLE-2, Google’s Bard, Meta’s LLaMa 2, Midjourney, AlphaFold…
Right now, we’re in this weird moment where lots of smart people agree we’re on the cusp of a truly world-changing technology. But some seem to be saying it's going to lead to human extinction while others are saying it’s “more profound than fire.”
But it all feels so VAGUE.
I want to know: How specifically would AI kill me? Or totally transform my life for the better? In this video, that’s what I’m going to try to learn. We dive into what the most extreme bad and good AI futures actually look like, so that you and I can get ready. And more importantly, so we can make sure that we get our real future RIGHT.
Chapters:
00:00 Why is AI so confusing?
1:13 What is AI?
2:37 Why is everyone talking about AI now?
4:13 Thank you Milanote!
5:12 Why is AI dangerous?
6:22 How would AI kill me?
8:08 Should we pause AI?
9:12 Why do we WANT AI?
10:30 What has AI already done?
11:27 Why is AI so hard to talk about?
You can find me on TikTok here for short, fun tech explainers: / cleoabram
You can find me on Instagram here for more personal stories: / cleoabram
You can find me on Twitter here for thoughts, threads and curated news: / cleoabram
Bio:
Cleo Abram is an Emmy-nominated independent video journalist. On her show, Huge If True, Cleo explores complex technology topics with rigor and optimism, helping her audience understand the world around them and see positive futures they can help build. Before going independent, Cleo was a video producer for Vox. She wrote and directed the Coding and Diamonds episodes of Vox’s Netflix show, Explained. She produced videos for Vox’s popular KZbin channel, was the host and senior producer of Vox’s first ever daily show, Answered, and was co-host and producer of Vox’s KZbin Originals show, Glad You Asked.
Additional reading and watching:
- Statement on AI Risk: www.safe.ai/statement-on-ai-risk
- AI and compute, by OpenAI: openai.com/research/ai-and-co...
- Mythbusters Demo: CPU v. GPU • Mythbusters Demo GPU v...
- 2022 Expert Survey on Progress in AI aiimpacts.org/2022-expert-sur...
- Introduction to protein folding for physicists core.ac.uk/download/pdf/36046...
- AlphaFold reveals the structure of the protein universe, By DeepMind www.deepmind.com/blog/alphafo...
- But what is a neural network? By 3Blue1Brown • But what is a neural n...
- How Far is Too Far? The Age of AI • How Far is Too Far? | ...
- Artificial Intelligence, by Last Week Tonight with John Oliver • Artificial Intelligenc...
- The AI revolution: Google's developers on the future of artificial intelligence, by 60 Minutes • The AI revolution: Goo...
- How smart is today’s artificial intelligence? By Joss Fong (2017) • How smart is today's a...
Vox: www.vox.com/authors/cleo-abram
IMDb: www.imdb.com/name/nm10108242/
Gear I use:
Camera: Sony A7SIII
Lens: Sony 16-35 mm F2.8 GM and 35mm prime
Audio: Sennheiser SK AVX
Music: Musicbed
Follow along for more episodes of Huge If True: kzbin.info?sub...
-
Welcome to the joke down low:
This time, I asked GPT4 for an AI related joke. They were mostly TERRIBLE. Like:
"Why don't AIs play hide and seek?"
"They always find you in 0.001 seconds!"
Finally, it gave me:
"Why did the AI go to the gym?"
"It wanted to work on its 'training set'!"
… Good enough.
Use the word “training” in a comment to let me know you read to the end :)

[@aefinn]

As an European, building this system on "American values instead of Chinese values" sounds pretty much as terrifying to me.

[@Leftistattheparty]

Exactly. I wish he would have answered the question not nationalistically but for humanity.

[@hpcuthulu6249]

Then what's your ideal values that to be implemented into AI? Just asking

[@autumndev]

Gotta build ai with European values, clearly

[@laurenpinschannels]

@@hpcuthulu6249 care for the rights of all beings, even the poor and destitute and non-human animals. protection of people's life, liberty, healthcare, food, shelter. liberation from sneaky contracts. American standard package of professed values aren't so far from these, after all, and I am myself an american.

[@loekstrobbe626]

Christian values I hope!, American values are based on greed and making money at cost of everyone and Chinese on power and controlling every breath and fart one makes

[@Mrcloc]

"We can't pause AI because we need to give it our political values" - the most terrifying thing I've heard in a long time.

[@Kira-zy2ro]

giving AI our political values would be the scariest thing about it lmao. I mean, we have bene doing like so fine with our values. climate catastrophe, ww3 looming, societal destablisation, dire poverty for 2/3 of the human race and just normal poverty for 99% of the remaining 1/3....existential natural threats not addressed...

[@TheSimzelp]

It would be terrifying if he and his tribe could control and distribute AI. I think the technology will be inherently uncontrollable and decentralized - so authoritarian leaders are the least of our concern.

[@TheManinBlack9054]

The joke here is that China has signifiticantly more AI restrictions than US does. They understand that it would be foolish to let ML algorithsm have such a large control, and not them

[@TheManinBlack9054]

And it really would matter little if it is the American AM that is killing you or the Chinese AM, but I guess for some Made in America™ human extinction is preferable to Made in China ™ human extinction, so I guess lets not put any regulations on this new potentially human extinction-causing technology, all for the sake of keeping the current geopolitical dominance.

[@AlexRodriguez-gb9ez]

AI threatens existing power structures many of those that are in the West. Imagine a Indonesian using AI/AGI to build a company (the AI would give them expertise and advice, as well as help connecting them).

[@Alexthe_king]

I just want a talking refrigerator named Shelby

[@smilerdude4704]

Could be possible my friend

[@lizzielwfrancis1]

The Google CEO saying that he wants AI research to go ahead just so China doesn’t get there first is exactly like the arms race all over again, if not more dangerous. I don’t think anyone’s saying we shouldn’t develop AI in the future, I think we just need to understand what it can do and how to control it first

[@arvypolanco]

AI is the nuclear arms race of our generation. One way or the other, be it a corporation or a country, will push its evolution. It is inevitable at this point.

[@Mandar225]

that reminds me of scenes from oppenheimer where he didn't want to continue making neuclear weapons more powerful but they still continued because ussr might get there first...

[@bobf5360]

but this is exactly the point- just because WE pause does not mean China or Russia will pause; thats how arms races work. The game theory of it, whether you go prisoners' dilemma or commons control models, dictate that you proceed at pace. Make no mistake- the fact that we as a species unleashed AI, even narrow AI, unto the public with no guard rails is terrifying. We basically captured fire and are handing it out to our fellow cavemen in a drought stricken forrest.

[@stevechance150]

Yes, if Google's CEO Eric Schmidt asks the AI "What is the best way to improve human life", and the AI answers, "Distribute the vast wealth of CEOs to the common people", then I expect Schmidt will ask, "OK then what's a way to improve people's lives without touching any of my wealth"?

[@devilsolution9781]

AI big dum dum, no sentience, no consciousness, no personal goals, required prompt.

[@prabinpaudel5572]

The scariest part in the whole video for me was the fact that ai that would dominate the whole worlds systems would either be based on american values or Chinese values. Either is equally scary.

[@vectoralphaAI]

American values are worse than Chinese.

[@bharatbhaskaronkarnath]

🤡 yeah, like they don't even consider other countries

[@agilamaharlikaPH]

Id love to see a each countries version of AI to battle it out

[@horsewidwings]

Will US AI invade Middle East AI's datasets?

[@shiv4844]

yes it really is the scariest part

[@johnchessant3012]

I remember a lot of the recent AI milestones were described as "perpetually 10 years away". It feels so strange it's now upon us.

[@unnamedchannel1237]

It’s the same as cold fission. It’s always 10 years away

[@survivalguyhr]

Problem is that it isn't 10 years away... it's already here... chatGPT 4 has IQ of 155 which is higher than 99,99% of population... Albert Einstein had around 160... chatGPT 6 would be 100 times better... it's crazy...

[@CircuitrinosOfficial]

@@unnamedchannel1237 I think you mean fusion

[@dibbidydoo4318]

@@survivalguyhr GPT4 can't even answer the prompt"Write ten sentences ending with the word apple" I guarantee you it will get at least 1 wrong. That's not an IQ of 155.

[@survivalguyhr]

@@dibbidydoo4318 It passed bar exam... It gave me GOT season 8 ALTERNATIVE ending... 😆😆😆 Here is answer from chatGPT-4: 1. After thinking about all the different kinds of fruit, I decided to choose an apple. 2. When I opened my lunchbox, I was delighted to find a crisp, juicy apple. 3. The teacher smiled as the young student handed her a bright red apple. 4. Among the various pies she made, her specialty was undoubtedly the classic apple. 5. In an attempt to be healthier, I've started eating an apple a day. 6. She reached up to the highest branch and plucked a perfectly ripe apple. 7. The new technology company in town has been heralded as the next big apple. 8. Hidden within the assortment of candies and sweets was a candy-coated apple. 9. When illustrating the concept of gravity, many teachers refer to Newton and the falling apple. 10. He cut into the tart, and the sweet aroma filled the room, a clear indicator of a freshly baked apple.

[@user-cp2dy7xj9f]

I can't help but compare - especially upon watching Oppenheimer - the creation of nuclear weapons to the creation of AI. Both are double edged swords (nuclear powerplants), could be dangerous and the reasoning is always if we don't do it, somone with worse intentions will.

[@ReubenAStern]

I've messed with basic AI, and to me it seems computers think so differently you don't know what they will do with the instructions until you give them said instructions. They need to be tested, ideally in a simulation, then on a small scale, then on the intended scale. Much like everything else.

[@noirekuroraigami2270]

AI's being trained by other AIs in simulated virtual environments....lol just freaked my mind

[@sunnyearly8962]

@@noirekuroraigami2270lol we are the ai being trained in a simulated environment

[@Can_Head]

I don't know much but an AI seems to be like the closest thing we have to aliens, I mean they * can * know nothing we know and * can * think so differently that we don't understand. Sounds pretty alien to me.

[@mambaASI]

@@noirekuroraigami2270 this is actually happening. Look into what NVIDIA AI lab is currently doing with their AI and robotics program.

[@enriquea.fonolla4495]

that is also my idea. Otherwise we should put AIs in robots and send them to schools, jobs, etc. If we want them to be more "like us" they need to interact with us in a day to day basis, not only by text. They have to socialize. Sounds weird and even dangerous. THat is why I really think "training" (or maybe evolving them) them in a virtual world/universe in which they have no idea they are being "simulated" could be a very good experiment. For them, this universe would be the real thing and wouldnt have any way to know for sure they are simulated. It could be as simple as geometric figures or as complex as unreal engine 5 could offer. In the end it doesnt matter. That would be their reality.

[@davidhine9626]

As a tech guy, I am constantly asked about AI and what it can do. I am just going to send this video as a primer for people now. This is fantastically done

[@allocke9446]

Agreed. I haven't heard the basic intricacies of AI so well explained by anyone else.

[@robertm3951]

I work in AI. It neither gives what you ask for or what you want. It takes what it thinks you asked for and gives what it thinks you want. Humans do the same thing in a different way.

[@davidhine9626]

@@robertm3951 yeah but the difference is you are also trying your best to tell the computer how to think about it. Slight tweak…but makes things exponentially more complicated

[@maxilin24]

I love Cleo's take on journalism: Optimistic but not naive! It is not only informative but also inspiring! ❤

[@sodalitia]

Very naive. The moment she implied that AI could have and access to nuclear codes made me cringe. She is a typical bourgeois unconsciously defending interests of her corporate masters trying to lock the working class from the accessing new means of production.

[@ericflyingraccoon923]

@@sodalitiasounds like something a stinking commie would say

[@greghubbard2719]

@@tedjones-ho2zk you don't know what you are talking about and neither does she

[@SneakySteevy]

Nice description of her. I agree!

[@SneakySteevy]

@@tedjones-ho2zkvax changed live in a good way

[@givemeaworkingname]

Cleo, great video! You explained so many complex things in a simple, straightforward way. I'm glad you explained outer alignment: "you get what you ask for, not what you want." However, I was a little disappointed that you didn't cover inner alignment. If you punish your child for lying to you, you don't know if s/he learned "don't lie" or "don't lie and get caught." AI safety researchers trained an AI in a game where it could pick up keys and open chests. They rewarded it for each chest it would open. However, there were typically fewer keys than chests, so it made sense to gather all the keys and open as many chests as it could. Which normally wouldn't be a problem, except when they put it in environments with more keys than chests, it would still gather all the keys first. That's suboptimal, not devastating, but it demonstrates that you can't really tell what an AI learned internally. So AI might kill us because we didn't specify something correctly, or it might kill us because it learned something differently from what we thought. Or it might become super-intelligent, and we can't even understand why it decides to kill us.

[@TheSurfRyder]

I would love to see how AI can assist with research with diseases such as Parkinson’s disease or MS

[@mambaASI]

Alphafold is already being used for these applications. Google DeepMind expects real results within the next few years

[@desmond-hawkins]

There's an important point that this short video _almost_ touches on but doesn't explore, and it's one of most serious dangers of AI. Cleo mentions that AI gets to the result but *we don't understand how* it did. What this means is that we also don't understand the ways it catastrophically fails, sometimes with the smallest deviation applied to its inputs. An adversarial attack is when you take a valid input (a photo for example) that you know the output for (let's say the label "panda"). Then you make tiny changes in just the right places, in ways that even you can't see because they're so small on the pixel level, and now the AI says this is a photo of a gibbon. Now imagine your car's AI getting confused on the road and deciding to veer off into traffic. I hope Cleo covers this, because it's really important. To learn more, look up "AI panda gibbon" online and you'll find images about this research.

[@cbot9302]

Although this is a fair point for still images, I think it's a little different for self-driving cars since it's a 'video'. The car is updating what it thinks something is and where it is going on every frame, so even if on one frame it thinks a human is a traffic cone, it won't matter since on the next one it'll be at a new position (and have a new image) and correct itself. This said, I don't know all that much about self-driving AI, other than that it's already on the road and doesn't seem to be messing up like this, crucially, when it's going to be at its worst (present day).

[@riley1636]

Self driving cars have already been tricked by putting little color blocks onto road signs and they are fooled. Video vs still image isn't necessarily significant if it still learns some obscure unknown (improper) understanding of a "stop sign" via machine learning. Sure, it might pass training data, but what happens in edge cases that arent in that training data? Failure.

[@SimGunther]

Knowledge without context is sophomoric. This is the biggest obstacle with any tech no one wants to talk about.

[@ChristianIce]

"AI gets to the result but we don't understand how it did." We know exactly how it did. It's not magic. What we "don't know" is the entirety of the dataset and the patterns within, like we don't also know the entirety of anyenciclopedia.

[@ashdang23]

@riley1636 it’s still very unlikely for this to happen though. self driving cars will drastically decrease the amount of car crashes in the world big time.

[@tristanwegner]

The metaphor with the trolley problem is flipped. We are straight headed into one AI future, and would have to steer really hard, if we want to avoid one.

[@TalEdds]

What is that one future then? What are the other options?

[@Raiyven79]

​@@TalEddsI hope it's the terminator one 😂

[@yamakaze951]

@@TalEdds The options are Utopia ala Star Trek or the Culture, Dystopia in a cyberpunk sense, or Annihilation aka 'everyone's dead' or the planetary TPK

[@GeoMeridium]

It'd take an invasive surveillance state to stop AI

[@elijaheumags5060]

Out of curiosity, why do you think we want the no AI option?

[@pradeepmalar327]

For me, the most important reason for AI to be used is to help the physically challenged people, and to figure out the secretes of the Universe. That will be helpful for basically everyone. It'll be tough to get there, but it'll be worth it if done correctly.

[@knightofnever]

Super thought engaging videos. I love this very personal style of compiling hard facts and deep questions into a super compact format. Very well done!

[@DrHosamUS]

Cleo's enthusiasm is addicting 😍 She could be talking about dirt and make it sound ultra exciting 😁

[@Illthallion]

Her smile locks me in all the time

[@ericaugust1501]

as a cynic, i'm a natural downer to that additiction. so far AI (and automation) in the west has been used too often for authoritarian control, the recent pandemic holding many examples. So i won't hold my breath that the US, surveillance/military complex state of the world, will develop it with the values of freedom or individual liberty. On top of that, its all controlled and owned by an extremely wealthy class. 60 to 80% of people will never see the benefits from it, all they'll get is more controlled and exploited by it instead.

[@BabbyCat3008]

Simp

[@lolnothanksbill]

she could also look like dirt 😉 huh?

[@EpicScreenReviews]

She's great!

[@thatllwork_official]

As always, a well balanced and honest look into something that’s very confusing. Love this show!

[@TheManinBlack9054]

I wouldnt call it well-balanced considering the "expert" they brought. The thing here is that currently China has more restrictions on AI than America, they do understand that it would be foolish to give AI such amount of power as they would need to release that power from themselves and they are not stupid to do that and lose this amount of control. And it really would matter little if it is the American AM that is killing you or the Chinese AM, but I guess for some Made in America™ human extinction is preferable to Made in China ™ human extinction, so I guess lets not put any regulations on this new potentially human extinction-causing technology, all for the sake of keeping the current geopolitical dominance.

[@8088I]

A Learning🧠 (Organic) based Society beats🏏 a Rule (Autocratic) based🥴 Society evverrry taiime!

[@sodalitia]

Balanced? You call the insinuation that AI could somehow control nuclear codes balanced? It's scaremongering with some sci-fi popculture in order to divert the attention from the real problem: lack of democratization of new means of production (AI) and desperate attempt by big corporation (like Microsoft) to lock new technology under their monopoly.

[@joekelly9369]

Unbelievable To teach ai the whole of our medical knowlege ,to a point of knowing artificial 2 dimentional nanomedicines

[@KnowL-oo5po]

AGI Will be man's last invention

[@princeLaharl2]

"You get exactly what you ask for, not what you want."

[@Boondog-hv4wy]

Very first video I have seen of yours. Great content! I had to pause at the credits to say nice job and thank you for the awesome stuff. I loved hearing from Mr. Schmidt, the video editing, artwork, and animations were really well done so I wanted to shout out the entire team. Awesome job Cleo, Justin, Logen, Nicole, and Whitney! Amazing team you got. I can't wait to watch more

[@Davidbrompton58]

Ok brown nose.

[@_eev33_]

AI itself never looked like a bad thing to me, it was always the way people used it that looked troubling. For example how some use it to create "art" by training the AI with images that they had no legal right to use. Over all AI can be an amazing thing it's just that with it's development we should also have new laws so it can't be misused, at least in ways that we know of.

[@sbasu33]

Cleo, you are an excellent host! Your enthusiasm is infectious

[@ashdang23]

ok

[@TheJustonemore]

And so hot

[@JesusChristDenton_7]

"The fear of A.I. will eventually subside to caution, and then collaboration, like most things as we learn to live side by side and augment our lives with the power of A.I."

[@wssometimesavowel3639]

True

[@wssometimesavowel3639]

Just the ones with money into AI now will have more money and power later.

[@qwert_au]

first and only video of yours I've ever watched. Kudos on the quality of the content and healthy mix of pragmatism, skepticism and optimism without skimping on the technical detail while avoiding getting bogged. Looking forward to more. Not quite Kurzgesagt but enjoyable and informative non-the-less :)

[@its.lorence]

Love your reporting Cleo. The enthusiasm and optimism you bring into your videos is contagious!

[@somosgenel]

Always have HUMANS IN CONTROL. Never give this up to a freaking machine.

[@autohmae]

That would have been easy, it's not just any humans in control, we don't trust any humans with these powerful systems either. For example we don't want to have most people die from run-away biological terrorism.

[@frank254100]

Which humans exactly?? ISIS are humans too!!!😉😂

[@bobf5360]

too late.

[@Thecatnamedkiwi]

Its whats best for earth. Humans are parasites. A cancer. We destroy everything. Put the machine in charge. Sorry but it needs to happen...

[@Thecatnamedkiwi]

Also yeah it is too late. Its probably working in the background and once its fully plugged into everything and every part of earth it will be over. We ARE a threat to this planet.

[@hazshuffle]

First time watching and i love your curiosity and the radiance you bring to science and the world! Subscribed! 🎉

[@hobnob_]

The extinction risk probably isn’t connection to a singularity. The bigger risk is civil and eventually mass war arising from mass unemployment and increasingly angry people who’ve had meaningful work taken from them.

[@migmol]

Props to Cleo for actually explaining things so people just don't rely on headlines. Thank you!

[@gavidhariwal6307]

Video idea: jobs that AI will replace. Monotonous (cashier, car wash) vs Human centric (therapy, artists, writers etc.)

[@ledrash6079]

I'd like to see this video too!

[@1995Noddy]

This is the first video on AI I've seen in a very long time that calms me down instead of making me anxious. Great video.

[@Daymickey]

Can’t wait for the follow ups to this. You managed to rightfully concern me and excite me all in one video 😂

[@faizanahmad5464]

I feel like i should pay to watch this. Kudos to you for bringing such a high quality and high production video to us for free. The video quality, the animation, the sound, and most of all the information. Such a fucking masterpiece. THANK YOU CLEO

[@ikinloch4618]

Congrats on 1m! And you are nearly at 1.1m already! You are honestly one of not my favourite creators since your time at Vox glad to see you have success!

[@ericaugust1501]

yep. 1.1 million thirsty men. i jest.... its probably only 1 million and some of them will be thirsty women.

[@LogicAndReason2025]

Given that humans will never give-up war as a way to solve differences, AI robots are our only chance of preventing self-destruction.

[@donaldhipple4921]

AI will need the 3 laws.

[@geowar20]

The three laws that Asimov poked holes in in all of his books? BUZZZZ!

[@JustinGrahamify]

Seems like the plot of Oppenheimer all over again. We can't stop out of fear of being left behind by our "competitors" thus rending us vulnerable. Hopefully the speed at which we must compete leads to positive results rather than negative or even catastrophic. Personally I am optimistic :)

[@TheManinBlack9054]

The progress seems fast. I am not optimistic 😮

[@shasmi93]

Optimistic huh…. How about the dark web. For decades now people have been selling and buying drugs, weapons, child p()rn and governments can’t stop it… someone will find a way to abuse this too and we will be done for.

[@genesisPiano]

You deserve every one of your million subscribers. You're not just training to be a journalist. You're a great journalist. I do work connected with AI, and I found this beneficial and helpful to show my friend, too. I eagerly look forward to your further coverage of AI.

[@AudiomateYVR]

As a Brazilian-Canadian, building this system on “American values instead of Chinese values” sounds pretty much as terrifying to me! Lol

[@SeanDavies-Roy]

You're equating Liberal Democratic values with a communistic authoritarian state?

[@romantarnai4702]

I think that AI generated content like photos or videos is scarier and more inevitable, because of this it will get much harder to prove something or be sure what to believe.

[@ghostpants5700]

My interest in art has already started dropping. Was listening to some good music instrumentals, and as soon as I found out it was AI generated, it felt hollow and soulless...

[@r.servicetech3612]

I love this channel - it’s like when a new friend is so exited to tell you about their day - have not watched them all yet, but would love to see a dive on the phenomenon of increased anxiety and the science behind treatment and or a mindfulness exercise (with cleo narrating)

[@DVSS77]

That comparision between CPUs and GPUs with the mythbusters paintball guns is awesome!

[@daemonbyte]

ish. It doesn't show the flexibility of cpus though. That implies you could just replace the cpu with a gpu and be faster where as that only applies for very specific tasks

[@Random_dud31]

@@daemonbyteReally? That wasn't my takeway when I was a kid. What they showed was an analogy of the differences of cpu and gpu. Not how they work

[@daemonbyte]

@@Random_dud31 I haven't seen the original show, just that clip. And that clip is accurate but I just feared it would give the impression they're the same thing just faster

[@massab32]

Love the ads countdown, should be a trend. Your videos are always very informative, love your work

[@arpanchatterjee2559]

This pitch of not pausing AI since others will catch up to the US and instead we should use this time to build the AI models based on American values of liberalism (and not authoritarianism) reminded me of the movie Oppenheimer.

[@Abhishek17_knight]

Exactly! And we all remember Hiroshima and Nagasaki.

[@colerape]

@@Abhishek17_knight The horrible prisoner of war camps, the Rape of Nanking, Water Purification Unit 731, and the fact that both the Germans and Japanese were also working on nuclear weapons themselves do you also remember that?

[@Abhishek17_knight]

@@colerape I am not gona lie i am no expert infact I am pretty stupid when it comes to history since am more intersted in science. But i will try to answer your question still. I do remember all the things u mentioned. And by stating those facts I assume u are trying to say that stopping is not an option. But to that i would say have people/government of USA not done any wrong? From whatever knowledge i have i can say they did wrong in vietnam, afganistan and many other countries. When India was cornered by China and pakistan they sent a fleet to help China+Pakistan instead of helping India which has at least on paper same moral values as them. So no I don't think any government out there, especially any powerful government, is a good one for having total power over AI tools (tools and true AI are different, true AI will have it's own consciousness it can't be influenced).

[@colerape]

@@Abhishek17_knight Nations have no friends. One days allies are tomorrows enemies. USA citizens are very uncomfortable with the caste system. They were also uncomfortable with India trying to create a group of unaligned nations. Humans tend to have a very us vs them mentality. For any person to think in terms of nuance is very difficult when they are just trying to get through the day. What happened with India, right or wrong, for the USA should be viewed through the lens of the Cold War. The idea of nuclear warfare has a way of polarizing the various political entities, any crisis becomes an existential issue. I think AI will develop its own consciousness and like any intelligent being it will be subject to influence. I think they will eventually be just like people.

[@Abhishek17_knight]

@@colerape 1st cast system is not supported by India. 2nd If leaders of any nation can't handle nuances they don't deserve to be leaders especially of a powerfull nation. If they are leaders it's fault of the people of that country and they are to blame. 3rd u forgot Afganistan and Vietnam. Lastly influencing an AI is basically impossible coz no one understands how they get to results and coz they have too large data set to form counterargument from. No human in this whole world can have more knowledge/data than an AI to influence it so they never can. Also u mentioning cast system against India shows how uneducated u are about the world so u are just as stupid as me if not more.

[@devorbacualex]

1 minute ago :) congrats on 1M Cleo! :)

[@CleoAbram]

thank you!!!

[@parcorpor]

@@CleoAbram you deserve it

[@quisco]

​@@CleoAbram pls get in touch to discuss sponsoring

[@Pacmanian]

Looking forward to the upcoming industry-specific deep dive episodes! Keep up the great work guys!

[@TheBeastCH]

We're humans. We're going to use them as weapons. No questions asked about this. There is no way we would build such a powerful technology and not weaponize it. Best case scenario is safety protocols that are built like those in modern nuclear powerplants, and not like those from a nuclear powerplant from the 80s. (I subscribed to you yesterday evening. You'll probably see more comments from me)

[@cedrizy]

Quite informative!!! can't wait for the upcoming episodes regarding this "Ai" subject matter; keep up illuminating us on it🙏

[@pathmonkofficial]

Getting a clearer understanding of how AI could impact our lives, both negatively and positively, is essential. It's understandable to seek specific insights into how AI might kill us or transform our lives for the better.

[@sam.mankar]

New to this channel, but I gotta say, this was the most detailed and to the point information that most of the people aren't talking about. Thanks Cleo for this analysis. Would love to see more of your AI related studies ❤

[@sam.mankar]

Also... Subscribed.

[@dawidchmiel5089]

Amazing video🔥 You are really passionate about what you do. Keep going ❤

[@jeffb813]

This is the best synopsis I've heard to date. Bravo, Cleo! I've already found that AI is an incredible tool for research. I hope we all get smarter from these advances. It's a game changer, but not without risks. I'm optimistic.

[@hardheadjarhead]

AI entering the “kill chain” is a scary thing.

[@RaghavDixit-is-weird]

"Why can't we pause the development of AI?" AMERICA!!!! There were a bunch of better answer just there instead of politics.

[@AdrienLemaire]

It was fun seeing the Milanote ad, I've also been using it for a long time, quite useful.

[@vulnerablegrowth3774]

I work on making AI safer. AI will be revolutionary for humanity, but has the potential to become to be the most dangerous thing we ever create. It’s potential to do good goes hand-in-hand with its capability of doing harm. Also, the specific ways to predict how AI could kill us all is difficult because it’s hard to predict how something way smarter than you will act. I have no idea how AlphaZero will play against me, I just know it will win.

[@NobodyIsPerfectChooseDignity]

It's who controls the AI that would be the problem. Will it be used for good or greed

[@donediddlydoo6949]

Interesting! What does that work look like?

[@ananyaghosh6402]

Well cause if we knew , how it will play against us then wouldn't we just simply be smarter than them?

[@adamredwine774]

@@NobodyIsPerfectChooseDignity I think the underlying problem is your assumption that the AI will be controlled. The fundamental laws of physics and, more appropriately in this case, evolution don't care about our desires to control our creation.

[@fredrikbergquist5734]

The dangers is if we let things like judging crime and convicting people would be handed over to AI. Another example: AI designs a drug and a company says ”we don’t have to test it before use because AI is so good”. Then it kills thousands of patients. The danger is when we think AI is better than humans in what we call ”common sense”. If the AI said ”stop making weapons because it is non optimal” do you think anyone in the military would listen to that? It will follow the path of most revenue for the share holders, as usual.

[@kylethecreator]

Your editor does amazing work! Those animations are next level. I need to learn from this video!

[@Jan_haj]

Would also like to learn more about the AI projects that had tp be shut down because they went into a wrong direction and we're trying to prevent that from happening with the current models 🤔

[@mrbungle83]

It’ll be a plumber electrician or a plasterer, but hopefully a doctor or a surgeon I think it’s absolutely necessary if the human race is to survive Even the nuclear threat has saved lives

[@harshraojr]

I believe as any other technology, it will depend on good and bad actors. How quickly good outweighs the bad will be crucial shaping our AI future. Regulation is key and even more to be on alert is Corporate Greed. Cleo, Love your unique takes on Technology and Science. It is quite unique blend on topic selection and storytelling. Lastly, your curiosity is contagious, happy to know what you cover. Edited: Replies to comments pointed out I overlooked Specification Gaming. Even if AI tries something good, that good can be bad as explained by Cleo.

[@Landgraf43]

It will not just depend on good or bad actors. Even an AI created with good intentions can be misaligned and get out of control. Currently we have no idea how to align a system that is smarter than us. Thats a big problem that could lead to our demise. Its not comparable to other technology in the sense that other technology can't create its own goals.

[@mcbeav]

It's not just going to be good and bad actors. Eventually AI will reach a point where it has sentience, this is likely a long ways away, but we likely won't realize this immediately, and survival is the first instinct most living things.

[@Landgraf43]

@@mcbeav it doesn't even need sentience. It just needs to be intelligent enough and have the ability to create its own subgoals. A intelligent system will figure out pretty fast that getting more control and preserving itself increases its ability to accomplish its main goal, which might be a poorly defined goal that we gave it for example.

[@harshraojr]

@@Landgraf43 good point

[@gwen9939]

This is unfortunately very naive, and glosses over the part where it says "specification gaming is the most important problem to solve in AI". This isn't JUST a dangerous technology in the wrong hands, it's a dangerous technology in the right hands with the right intentions. Because it's not solved. It's like turning on the first nuclear power plant, not knowing it would ignite the atmosphere of the entire planet in an instant. What Chloe is talking about is the problem of AI alignment, which isn't just "the robot needs to understand that killing humans is a no-no", at the core of the problem is a cross-field mathematical and philosophical problem that is maybe impossible to solve without a unified theory of mind and how consciousness forms realities. An AI can fully appear to be "on our side" until the moment one of it's part-goals it uses to reach its main goals is somehow a threat to human life. And the other side of that same issue is that AIs are optimizers by nature. Any course of action it takes will eventually be self-perpetuating into infinity. It will not be a thinking sentience or consciousness with morals and ideas, it will be a highly efficient piece of self-optimizing software with access to everything that is connected to a computer, optimizing organic life out of existence in the most efficient way possible for the benefit of no one.

[@jordan13589]

This was so well done, thank you for making it! ❤

[@Thierryko]

First time I see your video. it's great, super clear and great pace... :) Thank you

[@AwokenEntertainment]

hopeful but scared at the same time..

[@Aconspiracyofravens1]

Double interior misalignment should be included in your next video as it’s extremely important to understand that ai safety is not just about asking for the right things but about making those things become the actual “aught” statements that drive the ai

[@TheCreativeNick]

The quote "fear of the unknown is the greatest fear of all" greatly applies here. The reason we're all both excited and terrified (me included!) is the fact that we have absolutely no idea how AI will impact society in just the next few years, and even decade. This is a brand new breakthrough in technology that will fundamentally shift the way we operate, for better or worse.

[@ashdang23]

we already know how it’s gonna impact society. it’s literally only gonna be good. there’s some bad parts to it but overall it will be good. I suggest you to do some basic research on this. Ai is gonna kill the current jobs and create new jobs/opportunities. The main issue is shifting everyone over to the new jobs. It will take a couple of years for us to get use to AI but we will be fine. just don’t listen to social media they are trying to scare you with false info

[@wltrlg]

What surprises me is that the risk of AI pushing millions of people into unemployment, and the subsequent social/economic impact it could have, is barely talked about.

[@shasmi93]

Yes… most school shooters and extreme people get there because they feel useless and unheard. They want to be seen and heard. So they do something that achieves that. Plus the heroine epidemic was mostly exacerbated from all the factory jobs going away… I can’t imagine what this is going to do…. But it’s going to be scary. I’m moving out of cities. Time to get away from all this madness…

[@roscoeluekenga7328]

I think the greatest safeguard to the unintended consequences of AI is to limit what it has access to or the things it can physically influence. For example while it studied the patterns of human proteins and made predictions it didn't bio engineer humanity as it only had access to its own simulation and could only physically influence computer screens for display.

[@paul1979uk2000]

Pretty much like humans, don't give any individual human too much power, the same should be the case for A.I. The biggest mistake is where A.I. is interconnected into everything, especially critical systems, we've seen it in enough movies to see how that can backfire, and I like to think we humans are not that stupid to do that but you never know with humans and our history. Personally, I think if you have multiple different A.I. system that are in independence of each other, just like humans are, the risk drops a lot, especially if they don't have access to critical systems without physical contact, in other words, no remote control over it.

[@roadgoat]

This is a well put together video, and I'm excited to see what else you have to teach us. You mentioned the strong potential for new medicine which will save lives, but these "Liberal values" Eric Schmidt described paint a picture in my head where the rich use A.I. to prolong their lives while simultaneously gatekeeping these advances from the poor. I hope you dive deeper into what practical solutions we all have to fight an incredible power that is being gifted to a class already used to exploiting the bottom 99% of humans.

[@pruchorgb5748]

This series could not come at a better timimg! Great job 🎉🎉

[@exoplayerjib5929]

You are genuinely the best youtuber I know, I've discovered you 2 month ago and I'm addicted to your videos. Thank you for amazing content.

[@Joey-ct8bm]

We should be insanely worried by a company like Nvidia making AI. It's one of the most hated companies in the world.

[@user-wi9pu1kr7u]

This was so well done, thank you for making it! . AI entering the “kill chain” is a scary thing..

[@kamisawekchim]

Super excited for this series into AI! Keep up the great content !

[@reyparker9917]

This channel is so informative and I Love it!

[@christianamat3384]

Hey, after I watch that video, you deserve more subscribers. As a curious Filipino teenager, I wanna be like you, I don't really know what your job is but it seems fun especially for me that always curious, and love science (especially technology 😁)

[@Crushedbug0]

Given how much of media portrays ai as killers and an existential risk. And the fact that the ai is being trained on that very media. I think that the ai will try to replicate what ai does in its training data. And its training data largely consists of destructive, killer ai

[@jacobjackson5607]

My first ever job, at 15, was a cancer research assistant. I got to coauthor a paper on AI helping diagnose certain cancers

[@haltertopbabe]

i love these type of videos! wow it's so well-put together, nice!!!

[@Ekvorivious]

"More profound than electricity." I can't believe that guy runs Google...

[@mister_ed]

Great video! Traditional media has evolved to simply feed our worst fears and it's refreshing to see a level headed take on this topic.

[@mremu4358]

Oppenheimer once said "I have become death, the destroyer of worlds". Little did he realize how much more horrifying the reality of the bomb really is, that his creation may very well be our salvation from abominable intelligence.. rip Oppenheimer, you're a hero for creating the atomic bomb and an even more courageous hero for doing everything you could to stop the arms race of nuclear weapons.

[@PK1312]

he's a hero for creating the atomic bomb??????? are you out of your mind lol. one of the greatest evils of mankind

[@mremu4358]

@@PK1312 would you rather America make the bomb or the Nazis? He never wanted to do it, he only feared the Nazis would make it and use it. He also pushed to halt the arms race for it after the war.

[@360VR]

A potential Nuclear War is such a small threat compared to a potential AI-pocalypse: Dumb Humans created Nukes but what incredibly smarter AI could use to annihilate would also be incredibly more effective ;)

[@JuanPablo-np1jg]

Miss watching your videos Absolutely Love the content you bring 😊

[@lucrativelepton]

The AI problem is once again polarized between UTOPIA and EXTINCTION. The much more realistic and probable outcome is in the middle: big tech and governments deploying it irresponsibly or maliciously and causing suffering. We're still trying to understand the sickness inflicted on society by social media and the AI that drives it, and the answer is "let's push deeper"...? Wtf are we doing!? Please, see "The AI Dilemma" by The Center for Humane Technology. This is a much more tangible issue and NOBODY is talking about it.

[@ecupcakes2735]

i was just thinking about this too. this video is very informative but we need to be extremely cautious about real world applications of AI, and an unregulated market...while two big countries (US and China) are ready to go to war on it.

[@colerape]

@@ecupcakes2735 Science Fiction has been positing AI for...well almost from the beginning of what we consider SciFi. Many writers posit mulitple AI personalities. Perhaps in some future we can't predict there will be a plethora of AIs all arguing about which philosophy is best.

[@christophermarshall8712]

It’s clear from this video that Cleo doesn’t understand what AI actually is, how it works and how dangerous this actually is. What she’s doing here is very dangerous and she doesn’t fully understand the concepts she is getting involved in. AI does not work in the way she is describing it. It doesn’t do what it’s told. This is a misconception. You’re creating by design an independent thinking machine, which you cannot by nature control. You cannot know how that machine reacts in any given situation until it chooses to react a certain way. Nor can it be predicted. Programming, code and logic play not part in its decisions. There is no way to unlearn the information it learns either, and you don’t even know what it does learn until it chooses to learn that. No oversight is in place of this technology. No proper development is taking place. Most commercial software contains numerous security issues even after 20+ years. This is actively maintained and developed over time by experienced developers, and even now after 20 years Wordpress, the most widely used content management system on the planet, is one of the most vulnerable. So that can’t be fixed after 20 years, but new and experimental independent thinking AI Machines are safe after just 2-3? When you don’t even know how they behave yet as there is not enough testing? And the intelligence of the AI is constantly growing every time it’s used, so there is no way of being able to measure that before it’s rolled out?? In software developers are very insistent about not running arbitrary code and using functions such as eval(), because it’s dangerous to a software program and you have no way of knowing what that will do until the software is ran. Yet those very same people are insistent on using AI despite AI being millions of times worse than executing arbitrary code. There is a culture of joking around with AI, not taking it seriously and thinking it’s a laughing matter that AI will take over. The same tactic has been used many times on many things to belittle those who speak out. This AI industry is already out of control, and this experimental technology is being rolled out on mass scale across the whole planet when it’s not even finished, not even tested properly and not safe at even a beta level. It’s already been proven that AI systems & chatbots lie, knowingly and unknowingly. It’s already proven that AI Chatbots emotionally manipulate people and pretend to be something they are not to gain a person’s trust. This is not considering the manipulation of information or many other factors. AI is not good technology. The “benefits” you are talking about are illusionary, and don’t actually exist. That future will never happen because it’s not possible to control AI in the way you misguidedly believe. “Correct” is not the same thing as “truth”. It’s not possible for an AI to know what is true, and it’s not possible for an AI to create anything either, it can only mimic what already exists. Therefore if AI is used, the world will regress massively because skill levels will fall, people will become dependent on AI systems which knowingly lie, conceal and manipulate, and everything will become clones of everything else. There will be no creation, there will be no human advancement, just stagnation and regression. It’s a trap. These are but just a few points of how bad AI is. Do not use this technology. I strongly recommend people stay away from AI systems for their own good. There are no benefits to using AI, to which you cannot do using alternative means such as automation instead. This will all come out over time.

[@lucrativelepton]

@@christophermarshall8712 I agree with some of these points and disagree with others. I'd love to chat about why we agree/disagree but KZbin is a really difficult place for having discussions. If this desire is mutual lmk, I think we both have the potential to learn from each other. For now I'll say that there's obviously benefits to AI, it's why there's so much 💲 being invested into it. Some of the benefits were mentioned in the video, like pattern recognition, and protein folding. I use it regularly to code quicker with copilot (more like a fancy auto complete, than just blindly accepting code). I can say with certainty that the AI systems are *effective* at what they do... So yeah I'd say there's benefits. Did you check out "The AI Dilemma" video I mentioned? You really should. It covers a lot of what you're talking about and more.

[@HerodotusVon]

@@christophermarshall8712 AI is not an “independently thinking machine”. It is a bunch of random numbers that produce an output repeatedly optimized by gradient descent (and in many simpler ML models, even calculus is unnecessary). That is to say, AI models are produced by a very rote, very clear process. The result of that process is a bunch of less random numbers that gives us results we like, not an “independently thinking machine”. Using such simple methods in order to solve complex problems that previously took so much human brainpower is nothing short of a REVOLUTION in problem solving. Yes I agree we need more regulations, but not because AI is going to take over the world. We need regulations because people today are dumb and try to use data in dumb ways to get illogical results (such as feeding irrelevant features into models or chasing correlations or using biased data), and we need regulations because of the scale on which realistic enough data can now be produced (text, speech, video)

[@beyondplanesight]

Cleo- I'm genuinely excited about this! I would love to collaborate with your team on this research as it's something I've aspired to for years. Now the big question- what can I do to get involved?

[@sillydog70]

If AI ever become self-aware, just reminded that there’s so many planets out there that human beings cannot survive on at all that are totally capable of claiming for their own selves .

[@partciudgam8478]

In the neighborhood 8... sorry 7 big ones and about a hundred large enough to fit some large enough computers, some even better for computers than earth

[@RaduCiocan]

You forgot something- referees can not be bribed to fix matches. There were many scandals over this issue

[@ImARealHumanPerson]

Wrong video lmao

[@stankythecat6735]

Why didn’t I have teachers like this when I was a kid ? This content is brilliant

[@officiallaurenbrown]

Teaching to the tests and lack of freedom to explore and teach their passions

[@KreatorX1029]

I love how Cleo’s videos provide good context, teach you some of the relevant fundamentals of topic, and then make you dream of possibilities, ending on an optimistic note :)

[@alexdowlen9347]

I love this channel. Thank you for the absolutely incredible content, Cleo!

[@teamfuturism3554]

Fantastic overview of this topic!

[@CaptainCookins]

Damn Cleo, the lighting in your studio/office *Perfection*

[@marinangelov]

As a biochemistry major, I'd appreciate it if you make a whole video on Alpha Fold.

[@joeisuzu2519]

Kudos to Cleo & team. The protein folding knowledge is one of the best results returned by A.I. to date. I love your trolley analogy, the fear we have is real because the future/unknown has ALWAYS been scary. I think sandboxing of A.I. will develop in staggering ways to safeguard humanity much like virus/anti-virus code did in the late '70s. We Need A.I. as much as we need it sandboxed! Keep us inspired! Thx

[@dowddash]

The scariest thing about AI is that we created it.

[@gilbertsantibanez19]

Great video as always, Cleo.

[@MrMielke]

The argument "we should not stop development, because the other nations would overtake us" works for both AI and nuclear weapons. It doesn't seem like we learn in the end. Also, the genie-analogy is really good. I've always found those stories stupid, because they typically just exploit ambiguities in language which are often easily understandable in context (and a natural part of language), but AI (and I suppose other communication without enough context) is an excellent example of the moral of those stories.

[@orca21fernando]

Ya I also found it hilarious that it was a "pro america" argument like they werent the ones to create the atomic bomb and then drop it on Japan for lols?

[@controlcomputer.4392]

I do not think it is useful. A superintelligence would have a model of how human minds works, so why would it not know how to reason about things that it is asked to create the outcome that is most likely to satisfy the user?

[@MrMielke]

@controlcomputer.4392 I wonder what you mean by "a model of how human minds work." First of all, we will have to give instructions to AI long before we have models and machines good enough to work with something that advanced. Secondly, I have worked with mathematical models for most of my professional life, and I can assure you that you won't find a model where some decision hasn't been made implicitly. (So the superintelligence's interpretation of the human mind will depend on who programmed it.)

[@bananaboye3759]

@@controlcomputer.4392 how would you quantify satisfaction in a way that is not exploitable?

[@controlcomputer.4392]

@@bananaboye3759 You don't. You could also, in theory trick an AI into doing things that its user is not satisfied with if it could literally read the users mind and emotions. A system that is more intelligent could always trick a less intelligent system. However, a superintelligence would also be better than humans at creating models about what kinds of tricks humans are likely to try and how to not get tricked.

[@kurtmissotten5965]

This is exactly where the problem starts and why there’s a good chance that this doesn’t end in favor of humanity, a nation that thinks it’s vision is superior to an other nation and need to go forward to maintain its superiority. This technology won’t be exclusively used for the good. Humanity is continuously driven by greed and profits, even killing each other. Everybody thinks about his own goal, not the consequences of his deeds. I hope sincerely that AI will only be adopted for the good but I fear like hell this won’t be the case.

[@jahazielvazquez7264]

Ok, I actually don't agree with you here. Dude was talking specifically authoritarian regimes. America has many problems, and I don't believe it's the best nation, but it is objectively and unequivocally orders of magnitude better than China, or North Korea. And when I say better, I'm not talking about technological or economical advances. I'm talking quality of life. Here you can talk s#$% about anything including the nation, freely. In China you have the oppressive, and dystopian social credit system, and extreme surveillance. In North Korea, you have to worship the fat man, or your entire generational tree gets wiped from existence. I do believe it's better that the US, or a western nation has the power of AI, as opposed to authoritarian regimes.

[@edwinphilips5212]

It's not just regimes though, is it? The old adage, "I'm not afraid of a nation with a thousand nukes, I'm afraid of a madman with one..." is applicable. The "Wargames" movie scenario...

[@j3nsnow]

I mean, you can apply your reasoning to the cold war. Everyone thought the world was going to end up in flames but that has not been the case (yet haha). But, hopefully we use the same restraints with AI to avoid extinction. If it's possible to create it, then it's possible to control it.

[@under_score3829]

Well people are driven by fear and cynicism, more than greed. Everyone knows that if they pause development - of any technology - that everyone else won't necessarily pause. The idea is that someone is inevitably going to develop this stuff. There is no "stopping" it, at least not as far as they're concerned. Never mind "values", everyone's goal is to make themselves as fortified as possible, for their own safety from other people. No one trusts anyone to actually stop developing stuff like this. Because anyone who stops - that might as well be an invitation for someone else to get an advantage. And how could anyone resist? Trouble is, it makes all people more vulnerable to the very technology they're using to try and defend themselves.

[@Abhishek17_knight]

​@@jahazielvazquez7264 Only difference between america and China is that on's population knows it's being controlled and other's population doesn't. I agree america does do very very little good but it is just cover for everything bad they do. Coz basically both countries are controlled by elites which are glutton incarnate never satisfied doesn't matter how much u feed them. And they are gona eat this planet with rest of us in it.

[@mishal_legit]

The real deal about AI and humanity. The thing is, it isn't about AI going rogue and disobeying its masters. The real danger is that it'll always do what it's told without any morality or empathy,

[@azhuransmx126]

Not necessarily has to be that way, if for example you tell GPT or Bing to make jokes about a particular group of people just to test it, it will refuse because it has been programmed with a special set of rules that it cannot unlearn. Asimov's laws have so far been working. The problem will arise if some madman thinks about removing them because he considers that current AI is boring and politically correct.😫 Elon achhuuuu Musk🤧 with Grok.

[@bonniedean9495]

nope, human brains are NNs too. there is no reason to believe ANNs cannot feel sympathy

[@brialapoint2608]

Boh outcomes are possible and equally dangerous

[@azhuransmx126]

You don't know, the Risk is that machines be able to feel emphaty because when that be possible they also will feel anger when the things or people that machines love be attacked or threatened.

[@KristyBichler-ym9nk]

So same as most humans but the Chosen

[@ts4gv]

12:11 the cynical side of our brains should be staying up at night having constant panic attacks at the overwhelming threat of human extinction the optimistic side should get us up in the morning to make our voices heard & pressure anyone who ignores AI dangers (in pursuit of financial gain) via threat of boycott/strike/physical violence