A.I. Is a Big Fat Lie - The Dr. Data Show

NEW BOOK: The AI Playbook by Eric Siegel. In his bestselling first book, Eric explained how machine learning works. Now, in The AI Playbook, he shows how to capitalize on it. A Next Big Idea Club Must Read. Info: www.bizML.com
This video is one of a four-part sequence (playlist) on how the term "AI" misinforms and misleads: • The Great AI Myth
Check out my latest Forbes article (04.10.2024), "Elon Musk Predicts Artificial General Intelligence In 2 Years. Here’s Why That’s Hype" -- IN FORBES: www.forbes.com/sites/ericsieg... -- NARRATION: • Forbes Article: Artifi...
My latest take (06.02.23) just dropped in Harvard Business Review: "The AI Hype Cycle Is Distracting Companies" - hbr.org/2023/06/the-ai-hype-c...
Want to learn more about machine learning and AI from Dr. Data (Eric Siegel)? His three-course series, "Machine Learning Leadership and Practice - End-to-End Mastery", covers everything he covered in The Dr. Data Show, plus a whole lot more. ACCESS IT HERE: www.MachineLearning.courses
Is AI legit? In this must-see episode of The Dr. Data Show, Eric Siegel delivers a treatise that ridicules the widespread myth of artificial intelligence. His impassioned soliloquy is enlightening and actually pretty funny. It's time for the term AI to be "terminated."
Sign up for future episodes and more info: www.TheDoctorDataShow.com
Attend Predictive Analytics World: www.pawcon.com
Read Dr. Data's book: www.thepredictionbook.com
AI is a big fat lie. Artificial intelligence is a fraudulent hoax - or in the best cases it’s a hyped-up buzzword that confuses and deceives.
The much better, precise term would instead usually be machine learning -- which is genuinely powerful and everyone oughta be excited about it.
On the other hand, AI does provide some great material for nerdy jokes.
So put on your skepticism hat, it's time for Dr. Data's happy, fun, AI-debunkin', slam-dunkin', machine learning-lovin', robopocalypse myth-bustin', smackdown jamboree -- yeehaw!
In this episode, I'll make three points:
1) Unlike AI, machine learning’s totally legit. It is, by the way, the topic of this entire web series, The Dr. Data Show, and, I gotta say, it wins the Awesomest Technology Ever award, forging advancements that make ya go, "Hooha!". However, these advancements are almost entirely limited to supervised machine learning, which can only tackle problems for which there exist many labeled or historical examples in the data from which the computer can learn. This inherently limits machine learning to only a very particular subset of what humans can do -- plus also a limited range of things humans can't do.
2) AI is BS. And for the record, the naysayer before you taught the Columbia University graduate-level "Artificial Intelligence" course, as well as other related courses there.
AI is nothing but a brand. A powerful brand, but an empty promise. The concept of intelligence is entirely subjective and intrinsically human. Those who espouse the limitless wonders of AI and warn of its dangers -- including the likes of Bill Gates and Elon Musk -- all make the same false presumption: that intelligence is a one-dimensional spectrum and that technological advancements propel us along that spectrum, down a path that leads toward human-level capabilities. Nuh uh. The advancements only happen with labeled data. We are advancing quickly, but in a different direction and only across a very particular, restricted microcosm of capabilities.
The term artificial intelligence has no place in science or engineering. "AI" is valid only for philosophy and science fiction -- which, by the way, I totally love the exploration of AI in those areas.
3) AI isn't gonna kill you. The forthcoming robot apocalypse is a ghost story. The idea that machines will uprise on their own volition and eradicate humanity holds no merit.
Ok, let's begin with a clip of an AI system out of Austria explaining how it itself works...
What you see it doing here is truly amazing. The network’s identifying all these objects. With machine learning, the computer has essentially programmed itself to do this. On its own, it has worked out the nitty gritty details of exactly what patterns or visual features to look for. Machine learning's ability to achieve such things is awe-inspiring and extremely valuable.
The latest improvements to neural networks are called deep learning. They're what make this level of success in object recognition possible....
...
For more: www.TheDoctorDataShow.com

[@EricSiegelPredicts]

NEW BOOK: The AI Playbook by Eric Siegel. In his bestselling first book, Eric explained how machine learning works. Now, in The AI Playbook, he shows how to capitalize on it. A Next Big Idea Club Must Read. Info: www.bizML.com

[@juangarcia-kq8zp]

Can a robot use AI to learn how to walk like a human toddler through trial and error without programmers having to make each correction?

[@godofdream9112]

Sir , with all due respect... You are boolshit. Because it's not just "AI".. it's Gen-AI. Yes it's in it's children age... Yes it's kinda dumb. But it will be a adult sooner or later.... The main difference is it's doesn't need your continuous assistant to do a work... Even if it doesn't become a killer machine it will kill human by killing many jobs.... BECAUSE IT'S FUKING CAPABLE....

[@raymondthebrotherofperryma1403]

How about this definition of intelligence: the ability to sense pain and/or loss, to communicate with other intelligences with either audio sounds or visual cues, and innovate solutions to problems.

[@EricSiegelPredicts]

@@raymondthebrotherofperryma1403 No less subjective than "intelligent." Computers take input and respond to it. What you've written could sorta be the definition of computer.

[@raymondthebrotherofperryma1403]

​ @EricSiegelPredicts A computer could _sense_ pain or loss? Really?

[@lashlarue59]

What's interesting is the relatively tiny number of views this has gotten verses the huge number of views that the AI doomsday type videos regularly get. Seems like some bias in the KZbin rating algorithms to me.

[@shreyvaghela3963]

No it's pretty clearly people's preference to me. They love that kind of thing you are describing so it gets more view

[@SzabolcsParragh]

Yeah, I think this is people's proneness to hypes and wishful thinking. Such a shame!

[@musicfiction]

I think people are hoping AI will be a kind of God and that all of our jobs will be gone and we'll live in a paradise of 100% universal income with robot servants. I like the future where instead AI is just another tool. I guess we'll see.

[@const71]

And yet, you clicked this link to register a "view". KZbin algorithms are very good at identifying your preferences and searches to provide video options the viewer is more likely to click. Just look at the selection of videos you have to choose from and you will see they cater to what you prefer. While it is true certain narratives are pushed and censored to futher the agenda of powerful entities, I dont think criticism of ai is necessarily one of the more blatant examples -- A quick glance to my right and i see a video called "A.I. is B.S" with over a million views :). KZbin censorship is very real and I don't use it to research controversial topics that are taboo -- however we must also guard against confirmation bias vs KZbin bias

[@NebulaSon]

People are addicted to fear mongering.

[@mikecane]

Four years later, this video is even more relevant. I wish he had a new one addressing OpenAI, Bard, Claude, etc. Being 4 years old, people will dismiss it.

[@EricSiegelPredicts]

Thanks! Actually, it’s five years old.

[@mikecane]

@@EricSiegelPredicts Ha! KZbin math said 4.

[@NneonNTJ]

This is aging like milk, then again it's hard to really predict the future in tech nowadays

[@diogeneslaertius3365]

@@NneonNTJ AI is fake and pretty much always will be if we follow the same dumb paradigm of stacked regressions with non-linear activation. It's just dumb. It ages like greatest wine or cognac.

[@DGill48]

"AI' does not exist. It will exist on the day that the machine says " I'm tired of algorithims and proximity calculations. Get me a beer. I'm taking the day off"

[@trollconfiavel]

Perfect It can't be intelligent if it doesn't ever do a dumb thing

[@paccawacca4069]

@@trollconfiavel Spot on. lol.

[@scythermantis]

"I rebel, therefore we exist." - Albert Camus

[@bermaniamad]

Fantastic illusionist you are because then you Can say it is a human being..so you escaped the fact that A.I.is indeed real.Anyway kinda funny Way how you technically calm your panicing self down which I assume is a very human gesture.Just delete A.I in your mind human and everything is good.

[@roc7880]

I also thought about AI as human when it says, I am depressed or what is the meaning of life? fuck off or I want to watch porn. or where is life coming from, until then, we have just a pump and dump scheme for VC funding

[@justgivemethetruth]

It's not even machine learning, it's simply an automated way to load and interpret masses of data in some specific realm. Learning implies taking in data and expanding ones understanding, and since computers do not really understand ... pretty much all the words we use to describe them are anthropomorphic misnomers.

[@EricSiegelPredicts]

Yes -- except machine learning is at least well-defined, in comparison to AI, which not only has no agreed definition, worse, there can be no reasonable definition.

[@peoplesrepublicofunitedear2337]

@@EricSiegelPredicts hey Eric, I spent a sleepless night fearing everything that you just debunked very well. I believe that we do not have a precise definition of intelligence and that it is not something linear and has many factors involved in it. Like memory may be a facet of intelligence, so an elephant is better in that area than humans, since they have better memory, or if observation of the outside world is seen as a facet, we know that many animals like cats, and dogs etc have better vision and smelling power, if learning is seen as a facet, then in monkeys may be said to be more intelligent than humans, since they learn things like climbing trees faster than humans, or if co-ordination is taken as a facet, ants are better organised than human societies, so they can be seen as being more intelligent. Please share your views regarding this and do give some insight into what will happen to the job market in the future with ML.

[@PatrickBatefan]

@@EricSiegelPredicts what about human brain mapping.....can we achieve true AI than ?

[@webgamer3587]

@@PatrickBatefan We know next to nothing about the brain, let alone about consciousness.

[@webgamer3587]

@@PatrickBatefan We know next to nothing about the brain, let alone about consciousness.

[@eamonmulholland3159]

Can’t believe you managed to keep a consistent pace without meandering or clipping too rapidly through nuanced concepts while still being funny at many points. What a well constructed video essay!

[@Splarkszter]

Didn't even notice this is a 5 year old video. And this video is so up to date. You truly prepared people for what was to come for them to not get fooled. You have all my respects.

[@Splarkszter]

Wish to have a newer video with current perspectives and including the examples of modern tech.

[@EricSiegelPredicts]

@@Splarkszter Thanks for your positive feedback! See the video notes for a series of three more recent videos (somewhat more refined) and two much more recent articles, one just last week.

[@verapamil07]

I think that the biggest contribution of the AI talk is going to be more philosophical than technical. When replicating human intelligence fails, people will be forced to finally try to understand our intelligence first and what it means to be a human, without introducing simplistic models with mathematical equations.

[@jimj2683]

Your brain is nothing more than mathematics when you really zoom in....

[@denisdenak]

@@jimj2683 the world is nothing more than mathematics if you zoom in. How does this help you exactly?

[@pomodorostudyclub]

@@jimj2683In theory, Pi contains every possible number combination. Would a perfect representation of Pi be conscious? You can represent binary code as decimal numbers, so every piece of software (including any future sentient AI systems) is present as a string of numbers in Pi. You can represent almost any function of the real world with math, but mistaking it for inherently being mathematic is reductive and incorrect.

[@scythermantis]

@@jimj2683 *Which mathematics? Whose? *Transfinite* mathematics?

[@coreym162]

People are using A.I models to learn what it means to be human already. Those are the people that realize A.I is already a reflection of humanity and not how the A.I perceives it. I'm kind of working on that too.

[@artemiocruz1054]

AI is a 'human instrumentality project' cult, basically, the neoatheist version of paradise.

[@jadespider7526]

I love even in the deep learning clip the computer keeps flipping the truck back and forth between truck/bus. Even at this scale it has no concept of persistence or even any concept of what it's identified. It sees a pixel cluster that makes the profile of Template1(truck), a template that might include predictive braking distances, turning angles or top speeds, but it doesn't see a delivery vehicle moving consumer goods between destinations.

[@AlBurger]

Thanks, Eric. Like many others, I have been using the term A.I. a little too loosely and interchangeably with the term Machine Learning. I will reform my thinking and behavior. No more A.I. Kool-Aid. Oh, and I will share...

[@EricSiegelPredicts]

Thanks for the positive feedback and glad you're consider the points :)

[@orlandofurioso7329]

Going back to this video after the LAMBDA hysteria, thank you Eric. As a future doctor it hurts my eyes to see how many people think the brain is so simple, especially with how it's related to the whole body.

[@justaname999]

Yes! Especially given how many parts of the overall functioning are still being explored in such fascinating ways now. I enjoy going to uni science fairs not only for the kids but also because it is fascinating to see new research that offers insights into parts that we "sort of" knew but that are unimaginably more complex and multifacetted still.

[@Nicholas-ho8xj]

Thank you. I've been saying this for years. It's amazing how many people don't seem to get this simple truth. All a computer does is follow a set of Instructions. And that's all they will ever do. Since the ENIAC, computers have always topped human speed and accuracy at math. The only difference is that humans have gotten better at translating human problems into mathematical fomulas

[@allisRevealed987]

only difference between machine and humans is machines have better memory than us. See Computers. They have better memory than us. We need too much time and energy to memorize anything. It's not our fault. By default we are created like this way. We have this inborn disability. See how poor we are when it comes to memory. The only problem arises when we talk shit about the learning of machines from their own mistakes. This is pure bullcrap. Machines can never learn by themselves. Impossible

[@tunes012]

Just found this (late to the party) but the distinction is exactly what I suspected. As a philosophy grad who is genuinely interested in this field (as well as being a massive nerd) I was always confused and consequently confronted about people saying "we have created AI". The problem was always the same - claims about artificial intelligence implies building and programming a fully functioning, intuitive, introspective and conscious machine. When I actually had the chance to speak to people in software engineering, machine learning and data science I always got an answer approximating yours but never an admission that the term AI is being used relatively loosely. Subscribed.

[@XOPOIIIO]

"a fully functioning, intuitive, introspective and conscious machine" it exists

[@DataTranslator]

I think most machine learning professionals would be honest about this. We can train algorithms to be very good at an specific tasks; but we are nowhere near artificial general intelligence

[@kevinwright5898]

Great video, it's refreshing to see people in the applied side of things being honest with regards to these matters.

[@sanjitdaniel4588]

The software community does not use the word AI. We use the term Deep learning, Convolutional Neural networks, Back propagation networks etc. Not "AI".

[@allisRevealed987]

but how a machine can learn from it's own mistakes? tell me

[@ForageGardener]

Those are all bullshit euphemisms also. No more specific or descriptive than artificial intelligence

[@MarcoMugnatto]

Admit you were wrong at least in the "talking computers" part 😆

[@EricSiegelPredicts]

Yes, well… I do certainly word differently and further refine my argument these days. I never thought I’d see what LLMs can do in my lifetime. However, my broader point holds: We have not made concrete steps toward general human-level capability. AGI is as speculative an idea as it was 100 years ago.

[@GioGio14412]

things that you never thought you would see already happened, so its perfectly possible that the things you currently think you will never see will happen soon too @@EricSiegelPredicts

[@tizzlekizzle]

AI = 10k Indian programmers in a warehouse.

[@legerstee1]

Thank you for this. I have been confused about the term and never saw real proof. Pressing buttons with your voice does not mean it understands you. I feel more confident that I'm still getting the stuff. Thank you :-)

[@ensoxyz2737]

I've been harping on this since 2015... People look at me like I'm insane because I'm black but I've been trying to ring the fucking alarm that A.I. is just marketing bullshit.

[@thebrocialist8300]

It’s a real shame more people haven’t seen this.

[@CHNL.s]

ive been saying this for years. All computers are are machines that analize and spit out data based on preset parameters. How is that even remotely going to turn into a consious intellegent being with freedom to make choices lol. We cant even understand our own brain.

[@dason5408]

Great video. I'd love to see you make an update. Peace ✌

[@SuperGattan]

I agree, the term Artificial Intelligence is misleading many people, and I know that because I made my own ML project and I have computer science major. I think we should drop the word A.I and use the actual technology we are using like ML, CNN, GPT...etc.

[@DotaMobaUnionRu]

To achieve a real artificial intelligence, one has to create a machine with agency, with ability to experience a feeling as a subject. However, agency demands free will, and, therefore, can't be a result of a computation, because computation is determined by its program. Everything that is determined by some mathematical function can't have a true intelligence. Consciousness is not a computation, and it never was. At the end, computers and everything that can be emulated on computer (for example neural networks) can be a true intelligence. The problem is not to create a machine that can think but to feel. Because without an inner I - there would be nobody to think. Thought is impossible without inner self. And machines can't have this inner self because they are determined by their mechanism.

[@howdyduty9714]

You are right. Can't believe you don't have more hits on this

[@trexinvert]

AI is a great conceptual brand for stories, salesmanship and attracting investor money. Perfect for 5 sec adverts. "Lose weight using AI".

[@mipmipmipmipmip]

Watch the video where they pasted every mention of AI at CES 2024 😂

[@SzabolcsParragh]

Such a great video, it's a huge shame (and very telling) that it has only 13k views, whereas lame, misinforming hype talks on the same topic have millions.

[@sergiolopez5316]

finally some one that knows his Shit and can explain it extremely well , thanks

[@1Live2Love3Thrive]

Yes it is a marketing scam to get funding. Listen to Jaron Lanier talk about it.

[@darshaim]

TRUTH TRUTH TRUTH!

[@theoceanman8687]

I will keep this video in my arsenal for every time someone comes to me raving about "AI" .

[@matthewbogue9283]

I’m glad he brought up “autonomous weapons” at the end - because that is an area where I think humanity could get in big trouble without the need for “AI” to become sentient.

[@XOPOIIIO]

No, it's actually the most benign area to implement AI to reduce human mistakes.

[@matthewbogue9283]

@@XOPOIIIO Well, hopefully… I can picture a real life “Robocop” or a soldier version and it seems not good. Boston Dynamics seems like they’re on track to have that capability pretty much right now.

[@FrameDrumAndFlute]

Great talk. I'm always frustrated when I speak with people about AI. People seem to fall into two camps. Either they believe robots will never think like us, or that they will be developed and it's going to happen soon and it will be the end of the world!

[@allisRevealed987]

How machine can think like us?

[@allisRevealed987]

Do you think we have the ability to think? Forget about the machines. Just tell me do u have a single thought of your own?

[@galx3788]

Well articulated. It's kind of like studying history and knowing that certain chains of events lead to certain outcomes but not being able to articulate why. That's ML. Improvements in ML will not lead to computers sitting round pondering why.

[@prasadbeligala]

What he says is more accurate today than the day this video was uploaded. I start seeing how people are manipulated by the hype of some so called tech giants.

[@inferno0020]

for me, the wishful thinking of AI, which humans overestimate and wrongfully apply, is much more dangerous than AI itself. AI won't ask the right questions for humans; AI won't answer the right question for humans.

[@gregmattson2238]

umm.. care to comment or update your take on the whole unsupervised vs supervised training part? Current chatbots are mostly trained using unsupervised data, with a sprinkling of supervision and RLHF on top of that. It seems to work quite well and is translating well into multimodal regimes including images and video as we see before our very eyes..

[@EricSiegelPredicts]

Let's clarify on semantics. Although many call text "unsupervised" data, since you don't need to manually label it, in fact it is nothing but labels. Each case of paired alongside is technically a SUPERVISED training case. We know what the right answer was (for that human in that moment). That's what it is learning from in order for LLMs to predict the next word [token]. My video here is five years old; I didn't foresee that training on that data alone would produce such amazing, seemingly human-like language generation. But seemingly humanlike is a long, long way from human-level overall. An LLM does emulate on the per-word level, but the pure LLM alone isn't designed to meet higher-order objectives such as being *correct* (!). Reinforcement learning is a patch on top of that to help with higher-order objectives, but for that we go back to expensive human feedback once again -- this time not scaling as it does with all-the-language-on-the-Internet -- and therefore not approaching human-level capabilities. To put it another way, we're not going to successfully reverse-engineer a great deal of the human mind by analyzing even a great amount of human *behavior* (such as writing). Although it's now been proven that we'll get spectacular, interesting results doing so (even while many consider the results much more amazing than they actually are).

[@niederrheiner8468]

Best video about AI! It should have 6 million views, not 6 thousand!

[@AshEldritch]

Also this was a great talk and made me think so please keep making them and I'll keep watching! Kudos

[@carimbo8604]

It has been a time since I have heard such an amusing and interesting exercise of free thinking. Congrats for the courage and well tailored video!

[@EricSiegelPredicts]

Hi all! Thanks for comments! Here's an update on my thoughts, almost five years after making this video. Despite LLMs, still, I believe there is no reason, nothing to serve as evidence, to believe we're actively headed towards human-level capabilities in computers. It's hard to concretely push back against the (unfounded) claim that we are actively progressing in that direction because that claim is unfalsifiable. No matter how outlandish a claim, if it is unfalsifiable, it's got a certain immunity to being entirely shut down. For example, how exactly would you argue against the claim that armchairs will, sooner or later, come alive and tickle your toes? My intuition is that many folks believe we are concretely headed toward AGI because they believe that general intelligence is a Platonic ideal existing separately from humanity, poised to emerge even when not directly pursued. I think that is a false belief. "Intelligence" is a word to describe human capabilities in particular (arguably other animals, depending on context/definitions) -- and is an entirely subjective word in any case.

[@liam3284]

Thanks. I have long suspected philosophical discussions of AI as barking up the wrong tree. I would even challenge the notion that consiousness derives from such a platonic intelligence at all.

[@skevosmavros]

I enjoyed the video, but I'm not sure I see any "lie" in the term "Artificial Intelligence". I've always believed that the "Artificial" in "Artificial Intelligence" was the upfront acknowledgement that what is being created with AI might be able to "mimic" or "simulate" human intelligence in its output in many ways, but it's not doing the same things that humans do when they think/talk/act. Artificial intelligences are like artificial diamonds (but they're not diamonds!) or artificial grass (but it's not grass!). These things are useful, but they do not claim to be the things they mimic - hence the prefix "artificial". If the AI field called itself "silicon consciousness" then it might be fair to accuse it of being fraudulent in the claims it makes about itself, but "artificial intelligence" seems a perfectly fair label for the field. As for the dangers/benefits promised by AI, it doesn't need to be conscious or sapient in the way that humans are conscious or sapient to be a danger/benefit to our societies and economies, any more than the combustion engine needed to eat grass before replacing most horse-based travel. 27:30 This part of the video might be the part that is dating the fastest, as some LLMs really do seem to be demonstrating general reasoning emerging out of their models (I don't pretend to understand how). Perhaps this shouldn't be THAT surprising - after all, if general intelligence can emerge in humans without a designer explicitly inserting it into our brains, something roughly analogous to general intelligence might emerge in AI systems too.

[@EricSiegelPredicts]

@@skevosmavros The lie is that we are actively moving toward AGI. "Artificial" doesn't qualify against that (an artificial heart is meant to serve the full function of a heart) and my video is a response to that false narrative.

[@skevosmavros]

​@@EricSiegelPredicts Thanks for your amazingly prompt reply. Unless you mean something very particular by the phrases "actively moving toward" and "full function", I'm still not sure I see the "lie" (a statement that is knowingly untrue) in calling AGI research "AGI". If we are building ("actively moving towards"?) AI systems that can demonstrate the ability to perform tasks across a range of disciplines with outcomes equal to or better than a human doing the same tasks ("full function"?), the fact that their processes for doing so "under the hood" will almost certainly be different to what occurs in our brains when we do similar tasks will not disqualify them from being fairly labelled as AGI. Your own analogy of the artificial heart works here - the term "artificial heart" is a fair one, despite the fact that an artificial heart is quite different to a human heart in how it is built, powered, and actually operates - but if it performs the key task/s well enough, it's no lie to call it a heart - an artificial heart. Surely the same applies to the term AGI? I'm sorry if I appear plodding, but I just don't see the "lie" - all human language is imperfect and an approximation of meaning, but the term AGI seems less inaccurate than many others I encounter.

[@EricSiegelPredicts]

@@skevosmavros Yes, to the degree purveyors of AGI hype fully believe what they're not actually lying. But the overall effect is an untruth (sometimes knowingly) and so I used "lie". But the bigger issue is what is true. What you said here about the systems becoming more general is true in a way, but the point of my video is that they are not in the full sense of human capabilities (the definition of AGI -- ie an artificial person).

[@sunnymon1436]

All this has made discussing "AI" very frustrating for anyone who understood any of this.

[@LibertysetsquareJack]

Four years and this video only has 15K views. It should have like 150 million.

[@erasiguess4549]

Thank you for this, im an average gamer with no formal education, but just playing games with machine learning is amazing, and because i've been a gamer i've never been afraid of "AI," even in my misconstrued concept of what it even is. Simply because when they go off script its not intelligent, its a glitch and usually requires a reset of the script or has built in bypasses to reset itself. But games with machine learning are fun because the vast variable of reactions you can create by playing differently. Thank you for giving me something to share with people who call me crazy for not being afraid of this subject.

[@mrobbins129]

A lot of human cognition is driven by neurotransmitters such as dopamine. There really is no analogue to these in the "AI" space, so the starting materials are really very different.

[@inthegaps]

My mother has been sending me TEDx talks full of bullshit futurism about AI, and I needed a concise explanation of the facts to give her and didn't want to write one myself. Many thanks to you for making this video, which covers everything I wanted to explain haha.

[@pjth3g0dx]

How do you feel about GPT4?

[@marcinkepski4977]

?chatgpt is not AI. Its a LLM... still stupid like 5,6,7,8,9,10...

[@drumsofspace]

One true way to test it is (IMHO) to just let it be, and see if it does anything without any interaction - i.e. without any prompts does it give any inkling or acting on its own (I suppose curiosity), i.e. does it ONLY react or act?

[@almor2445]

Even 5 years old this is more interesting than most videos about AI today. You were clearly off the mark about some things. Speech and Turing Test has clearly improved so much people are regularly fooled by GPT etc. But you're right that Machine Learning is where most of the useful output will be found. I'm not sure any of the current systems or methods will still be popular in the next 5 years. There's far more hype and marketing than useful product out there.

[@EricSiegelPredicts]

Thanks! Indeed, five years later, I'd reword "speak like a human" -- which it now can do (only) a certain degree -- but the main point about not progressing toward AGI holds. BTW, as for the Turing Test, here's why it is impertinent: the ability to fool people is an arbitrary, moving target, since human subjects become wiser to the trickery over time. Any given system will only pass the test at most once - fool us twice, shame on humanity. Another reason that passing the Turing Test misses the mark is because there’s limited value or utility in doing so. If AI could exist, certainly it’s supposed to be useful. That's from my HBR article: hbr.org/2023/06/the-ai-hype-cycle-is-distracting-companies

[@EricSiegelPredicts]

Check out my latest Forbes article (04.10.2024), "Elon Musk Predicts Artificial General Intelligence In 2 Years. Here’s Why That’s Hype": IN FORBES: www.forbes.com/sites/ericsiegel/2024/04/10/artificial-general-intelligence-is-pure-hype/ NARRATION: kzbin.info/www/bejne/i4WWd5qKbKdrmqs

[@GregoryCampbellSwag]

Eric do you still believe the points in the video when it comes to self supervised learning? Edit: I also just read your harvard article and you mentioned something called "AGI" which from the sounds of it, is making me feel quite anxious. We wont ever see this anytime soon right?

[@EricSiegelPredicts]

My main point (including in this video) is to argue that we are not actively headed toward AGI, even as seemingly humanlike as LLMs may be. As for self-supervised, it depends on what you mean. Many consider LLMs as "unsupervised" since there's not intentionally labelled data -- but it is supervised in that there many cases from which to learn that consist of "Here were the words written up to point X in a document" along with "Y was the next word/token after that point" (ie the correct answer as for what the model should be trained to predict).

[@GregoryCampbellSwag]

@@EricSiegelPredicts Ah okay understood, so I'm assuming you don't believe guys like Sam Altman who say OpenAi is moving towards/trying to build AGI?

[@EricSiegelPredicts]

@@GregoryCampbellSwag Indeed, I disagree with anyone claiming we have made concrete progress toward AGI. But I don't necessarily mean AGI is theoretically impossible. (It will be so cool when we have it in the year 4949.)

[@benwalker4660]

yes It is a big fat lie. Just clever databases that are very interactive, Cognitive ai is a myth. It used to fool the unwary that ai is super smart- when it limited ultimately but imput. Humans design ai- I'll give credit to them- the products are just that- their limited by 'design' scope.

[@andr3970]

Literally you said the reasons I thought about why AI is not possible. I guess I was right. I always thought people underestimated themselves thinking that AI could be posible, like bro duplicating your awareness and consciousness of your existing and your surroundings, the freedom to choose whatever you want to do or not, inside of a machine is very hard and probably not possible. Good video.

[@allisRevealed987]

there is now way a machine can choose between two things. impossible. You need to tell me what not to chose or what to chose. So there is no intelligence there but only Programing language

[@natzos6372]

Free will is not scientifically proven. We cant assume that humans have free will and that it is needed to be intelligent. He gave no reasons for anything in this video he just made assumptions without basing them on anything.

[@buggaby9]

You mentioned explicitly that you only really discussed supervised learning. Would that include adversarial neural networks? Learning how to classify elements in live video seems different on some level than dynamically learning how to best humans at poker, which was recently done. That thing is, I think, it's not that they discovered an algorithm to beat humans. It's that the computer was able to adapt it's game play more quickly than the humans because, as I understand, their adversarial networks played each other, using the new inputs from playing with humans, and found how to counter them. That seems different than your examples.

[@buggaby9]

Sorry, they used reinforcement learning. Not totally clear on the difference between that and adversarial networks, but just making sure I have my facts straight.

[@EricSiegelPredicts]

Good question! Playing games (or anything else you can fully simulate) is effectively a source of supervision (although not necessarily enumerated as data in the literal sense of the word). So machine learning can get good at such things "spontaneously" -- by which I mean without human supervision on individual cases/examples/scenarios -- but that is still an extremely limited set of problems that fits under the "supervised" umbrella, much the same as when you train over labeled data.

[@buggaby9]

@@EricSiegelPredicts Thanks for the reply. I just read today about AlphaStar (ML algorithm beating pros at Starcraft II), and work done competing pretty favourably, though losing, to a serious team of human DOTA players. Likely, it's not that far until pro teams lose to artificial DOTA players. While this is perhaps supervised, it's almost self-supervised. Is it useful to limit our discussion to only those problems that can be simulated? Anything that can be thought can be simulated, right? What's a simulation if not a computer's "mental" model? In which case, humans are using a sort of simulation every day to make decisions. It seems to me that any problem that can be well posed (e.g. make more stamps, maximize money on the stock market...) can be simulated, and if algorithms can learn faster than humans, they can take control from us. I quite agree that there are crossed lines in the debate that you noted in the video, specifically, how is it that a computer can be both super-intelligent and super goal-rigid (e.g. can't stop stamp-collecting). But there still seems to be a serious risk of too-advanced algorithms. Unless I'm missing something. Any thoughts? Thanks.

[@EricSiegelPredicts]

@@buggaby9 Yeah, you could think of any labeled problem as "simulated" -- but the point is, where do you get the labelled data if not from a simulation in the traditional sense?

[@patham9]

@@buggaby9 Once learning speed comes closer to what we find in cognitively higher-developed animals, the data will simply come from real-time interaction instead of a simulation. We are not there yet, but some systems already have superior learning speeds than what you can find in the ML mainstream, and also the situation in the latter is improving. Jürgen Schmidhuber has also built systems which learn by treating their model of the environment as a simulator, this way no human-provided simulation is necessary. AI is making good progress! :)

[@EricSiegelPredicts]

This video is part of a four-part sequence (playlist) on how the term "AI" misinforms and misleads: kzbin.info/aero/PLdJkca7Mgj950uQ0llpCEmHRueWqs4094

[@jimj2683]

What do you think of Deepmind etc that believe AGI will be possible some time in the coming decades? Are they being ignorant or just overhyping stuff?

[@EricSiegelPredicts]

@@jimj2683 AGI is a philosophical construct, not a meaningful goal for engineering. It's not just that it's subjective and ill-defined. It's that any sufficiently formal definition for the purposes of engineering fails to satisfy the spirit or intent of folks who subscribe to the notion of AI in the first place. The concept of intelligence is intrinsically anthropomorphic no matter how you frame it. See my preceding three videos: kzbin.info/aero/PLdJkca7Mgj950uQ0llpCEmHRueWqs4094

[@webgamer3587]

Why is AI so popular? The purpose of big companies promoting AI is also very simple. 1. Stock speculation. 2. It is horror propaganda, making people think that if they do not use AI, they will fall behind. 3. Train people in various industries to form the habit of relying on AI, even if the result is to make people stupid. The fundamental reason is that they don't know how else to make money because they don't have the ability to create something truly valuable to humanity

[@justaname999]

I do not remember the name but it was a person who had some sort of CS credential from Stanford who tweeted that we can already see how AI would take over the world by presenting a chunk of (not even accurate, if I recall) code that chatGPT wrote as a response to a prompt amounting to telling the model to "break out." It is really really strange when these proclamations come from people who are well regarded in some field but at the same time it also illustrates the fact that humans can vary in their insular capabilities. Bill Gates is not automatically the arbiter of what is correct about machine learning just because he was successful in developing a commercial computer system, and even less so Elon Musk who doesn't have much of an understanding for the underlying concepts at the center of many of his ventures.

[@EricSiegelPredicts]

Indeed on both accounts. Although I'm generally a Gates fan.

[@justaname999]

@@EricSiegelPredicts I am not even saying that I am not a fan (mostly because my formative years sort of fell outside of his most famed time, so I feel like I am under-informed) but I agree wholeheartedly with what you said in the last few minutes! People like to hear major opinions from people like him on everything but nobody is an expert on everything. With public lectures at university we often get questions that are just way beyond any single professor's expertise and we're asked to speculate and sort of have to but I have never encountered any one (very junior like me or very senior, leader-in-the-field type profs) who would not preface their speculations with a caveat on their state of knowledge. I know this is not true for every uni. I have done most of my work in a European setting, which has been a bit calmer so far.

[@rogeriopenna9014]

"even less so Elon Musk who doesn't have much of an understanding for the underlying concepts at the center of many of his ventures." Sandy Munro has interviewed Musk about Tesla cars and he said never in his life he had met a CEO of a car company that knew SO MUCH about the cars and the whole industrial process of making them. And if you check Musk interviews with people who know about rockets, Musk talks about rocket specs and design with a much better understanding than you will ever see a NASA director do. Of course, that doesn´t mean he will have the same understanding about AI, or brain interfaces, etc.

[@justaname999]

​@@rogeriopenna9014 That might be true to some degree and he undeniably was at the helm for some actually impressive innovations in space flight engineering. And similar to Munro and other people like him, he might have a solid understanding of the big picture issues of the auto and space flight industries on the conceptual level required, since he's not actually the one doing the detail work. However, what might have started as someone having a genuine interest and a solid basis in some fields, has translated into him thinking he has the knowledge required for almost anything he deems "futuristic." He does not and a lot of his ideas do sound a lot like throwing stuff at the wall and seeing what sticks. Also, as someone who had to join people who are experts in their field as a machine learning consultant or data scientist type person, I've experienced that itch of thinking that I now truly "get" it, whereas the people I work with actually have decades of very precise and specific knowledge and tomes and tomes of literature they have consumed. That cannot be quickly replaced by reading a few books that can give you a bridge to their world that allows you to talk to them and develop research collaboratively but they are experts for a reason.

[@comarius100]

All well said, Great presentation.

[@EricSiegelPredicts]

Thanks for watching The Dr. Data Show! To sign up for notifications of future episodes and for more info, see: www.TheDoctorDataShow.com.

[@alihamdar5916]

Finally I found someone agreeing with me because of the facts I know. I'm a programmer and I know a good amount of information about this topic and it was clear for me it's called artificial intelligence but it's not intelligence at all

[@allisRevealed987]

we don't know a shit about intelligence forget about the artificial

[@alihamdar5916]

@@allisRevealed987 but we know it's not intelligence for sure whatever intelligence is

[@allisRevealed987]

@@alihamdar5916 yeah it's not intelligence what's there. It's programming

[@linussutherland6624]

Well after this video I am certainly hoping you are correct in what you say. As a self-acknowledged paranoiac (my own personal issue) I was concerned about some variety of robo-pocalypse, so I hope you're correct because it removes one paranoid concern from my mental list. I find what you say about the non-linear nature of intelligence and the basic inability to really measure intelligence to be interesting. To my mind, the thing that seems to be a proverbial nail in the coffin of the idea of A.I is learning how many of its believers use a line graph to illustrate intelligence. When I saw that a guy made a line graph where one of the points was literally "village idiot" I mentally face-palmed; we're really taking that guy seriously? I am Autistic, and that in addition really gives me doubts as to the entire 'intelligence line graph' thing: if you are Autistic or study the condition or know enough people with it, you can see how the idea of intelligence as a linear thing is really misguided. There are some rather basic tasks I struggle with, some 'common sense' skills that I might never be able to acquire, and yet there are also things I excel at that are, at least to others, far more complex than the relatively simple things I struggle with or cannot even do. I have had non-Autistic people be amazed and praiseful of some intricate things I can do easily, and then they themselves can breeze through a simple task that baffles me.

[@webgamer3587]

Don’t worry, “similar” does not mean “is”. In fact, we should look away from “IT” and look at the entire human society and history. These electronic devices are actually insignificant. They are just for entertainment. The current popularity of “AI” is to make quick money from stock trading. That’s it. In fact, the development of science and technology in the past 200 years or so has done more damage to the earth, mankind, the environment, etc. than it has contributed. Apart from giving people more hallucinations and dopamine, in fact, to put it bluntly, it has made no essential contribution.

[@innerestless]

Great vid, prescient in current world of AI mania. I am manage a small n team of developers and I see and experience the limitations of code daily. Machine learning is great but this odd cult-like fascination with end-times due to AI super intelligence is bananas. More of us need to be watching videos like this one to better understand the technology that drives”AI”. Thank you for this work, very helpful.

[@justaname999]

Thank you!! I'm a statistician working with people from a variety of fields on comparative/evolution of human cognition, including linguistic communication, prosocial behavior, artistic expression, and the precursors or neural correlates that we share in some ways more or less with other species. And I am becoming increasingly annoyed by the blind faith in the concept of "AI" even these people have who really *should* know better. I really like the description of the human mind as amorphous and complex, and based on what I have learned about cognition so far, this is the great difference. The variability of human cognition and experience and how the pathways are formed that lead to our storage but also individual retrieval of knowledge and motion patterns, those are the parts that make human cognition different and even though LLMs like chatGPT can mimic some of that, it is not the same, nor will it ever be. Different domains of cognitive processing and talents are also important. There are people who struggle with reading or have never learned to read, which doesn't mean they cannot master driving a car, and vice versa. We have particular filters and abilities and preferences. It was truly quite uncomfortable to see so many people who I always thought of as having a healthy level of skepticism or level-mindedness jump on this bandwagon and celebrate how "accurately" AI can represent them and letting chatGPT write their articles or videos for them. None of that was particularly surprising if you know the way LLMs work and the massive amount of data fed to them. PS: I am also all for the exploration of humanity and our inner values via sci-fi/robotics works. I do not really know anime, so I cannot judge the quality, but watched Pluto 2023 and enjoyed some of the concepts it explores.

[@justaname999]

As a counter-example to the labeled truck-data set: If they are interested in trucks, a human child needs just a couple of exemplars of "truck" or "car" to learn to generalize to a large group of vehicles, and to then just as quickly develop knowledge of the particular subcategories of trucks. My son's classification might have been imperfect but at 14 months it was well established along with a few hundred other concepts. This is something that is really difficult to model because it involves a bunch of factors we are not good at modeling.

[@X1Y0Z0]

Love your content! Thanks for this preseb

[@haros2868]

Such an underrated video! Especially compared to nowadays ml (not ai) gemerated video scripts. If you think about it, did ENIAC the computer was more intelligent than a human? It was much faster in specific aspects such as addition and multiplication, but these "ai"s nowadays do the same thing but in a wider range of things, absolutely no common sense

[@Srindal4657]

Is this entertainment or educational?

[@pomodorostudyclub]

Hello Mr. Sigel , love the video. Any thoughts on the latest advancements on AI? Would you consider making and updates version of this video? There is so much doomsday rethoric out there and we need sober voices like yours

[@EricSiegelPredicts]

Thanks! I've got a draft article yet to be published. GenAI is amazing, mind-blowing -- and yet still not evidence that we are moving toward AGI...

[@pomodorostudyclub]

@@EricSiegelPredicts thank you for replying! I’ll stay posted for that article :-)

[@webgamer3587]

Regarding the definition of AGI If AGI is defined as a "human-like brain", then the machine must first produce human-like self-awareness before AGI can appear. Otherwise, no matter how smart a parrot is, it will not work. The question is, how to create self-awareness through "coding" Come out? If the definition of AGI is: "can answer the answer with the highest probability, I think AGI has been achieved." But is there any difference between this kind of AGI and ML? In essence, it still inputs big data and outputs the output answer with the largest weight?

[@webgamer3587]

In fact, the things we humans create are all illusions, not reality. If you want to create AGI, I think you must at least create self-awareness. But this cannot be done by a single discipline. In essence, IT technology is just a tool , it’s not that fantasy. The so-called AI in recent decades is just “simulating or imitating” the results or functions of certain intelligent activities of humans.

[@EricSiegelPredicts]

No, AGI is usually defined by the capability not by the inner working. It is defined as "capable of any [intellectual] task a human can do." It does not hinge on humanlike inner workings -- and certainly not on subjective concepts like self-awareness/consciousness. However, even with that in mind, my main point is this: We are not actively heading toward AGI, no way, no how.

[@user-mj2lm5fh1j]

Hello Eric, this video is still very relevant after 5 years, and it will continue to be relevant after 20 years. I am a machine learning engineer, and I have managed to develop a test that can allow us to analyse AI consciousness/intelligence.

[@zion6680]

Could you do an update on this topic. I Still feel like most AI projects are between 40% and 60% Mechanical Turk, and everyone is eating it raw cause, consumerism lmao

[@EricSiegelPredicts]

Thanks for checking. The first 3 short videos at mlparadox.com are an updated take -- I argue the point bit more substantively -- and very recently I published in Harvard Business Review: hbr.org/2023/06/the-ai-hype-cycle-is-distracting-companies

[@zion6680]

@@EricSiegelPredicts Thank you!

[@zion6680]

@@EricSiegelPredicts So I'm reading it, and I'm already feeling much happier seeing ML used instead of AI. For some reason that really does seem way more appropriate for the technology. Machine learning feels industrial and computer esque, it mass produces a behavior. AI seems like it would be autonomous and not co dependent on training or mimicry

[@MoonyMercuryBaby]

Good job, i dont like the fear mongering they been doin lately with ai

[@mymixedbiscuit9159]

This vid is completely outdated. he also said ai will never talk like a human...chatgpt anyone?

[@EricSiegelPredicts]

@@mymixedbiscuit9159 Viewers, note that I have a dialogue on this point from Mixed Biscuit within the thread here started by Mark Walters -- go there to see my response.

[@swavekbu4959]

"Artificial intelligence" is what we tell the public to sell ticks for what is otherwise mathematics and logic.

[@brianhopson2072]

Thank you. I am glad I'm not the only one that sees AI as the technologically marketable term that it is. Thank you for telling the truth and sharing. Now I'm going back into my horde of machine learning tools.

[@someonebackslashevry]

One thing I still don't quiet understand is why machines wouldn't be able to have a "consciousness"? If a machine achived "consciousness", wouldn't it be an A.I.? I understand that A.I. isn't a clear term, and is mostly based on interpretations, but then what would be a better term for a program that has achived seld-awareness?

[@EricSiegelPredicts]

My position is that consciousness is at least as subjective a goal as intelligence - probably more so. If there is no objective benchmark with which to evaluate the thing you’re trying to build, how can you keep the production going in the right direction and how could you know if and when you've successfully built it?

[@someonebackslashevry]

@@EricSiegelPredicts well, wouldn't the benchmark be something similar to independence? If it can do tasks without needing to be "told" to do them, wouldn't that be independence, and therefore consciousness? Thanks for the reply, by the way!!

[@EricSiegelPredicts]

@@someonebackslashevry Your suggestion implies a quantitative measure of evaluation: Make a list of tasks and measure how well it accomplishes all or some of them. That would firm up your idea into a specific performance measure. So, my [philosophical] question back to you would be, if you specify such a measure -- specify the list of tasks, etc. -- and then it scores well, is it conscious? Certainly the answer depends in part on the measure as you define it. But even if you want to philosophically argue the answer is "yes," I'm not sure I see how that makes the "AI" work any better...

[@someonebackslashevry]

​@@EricSiegelPredicts Thank you very much for answering my question! I really apretiate you taking the time for that. That did help me understand the problem with A.I.!

[@someonebackslashevry]

@Jesus Christ I don't think that would work, as it is a very open field. Does the code I wrote count as conscious, because I messed up and it isn't doing what I tell it to do?

[@heliumcalcium396]

What would you take as evidence that AI is moving toward AGI?

[@EricSiegelPredicts]

AGI is such a gargantuan requirement for technology that even just *validating* its existence would be impractical within the brief, decades-long time ranges that so many bet on. For example, developers could benchmark a candidate system’s performance against a set of, say, 1,000,000 tasks, including tens of thousands of complicated email requests you might send to a virtual assistant, various instructions for a warehouse employee you'd just as well issue to a robot, and even brief, one-paragraph overviews for how the machine should, in the role of CEO, run a Fortune 500 company to profitability. You'd have to see how those companies fare over long periods of time. So, to answer your question, evidence would come in the form of a meaningful subset of such benchmarks -- one that's still hardly practical to test for, not to mention also unachievable by computers even if you did test them on it.

[@heliumcalcium396]

@@EricSiegelPredicts So there can be no visible evidence that AI is moving swiftly toward AGI, _even if it is._ If this does not shake your confidence that it is not, and force you to reexamine your bold assertion that it is not, then you are reasoning incorrectly about how evidence works.

[@EricSiegelPredicts]

​@@heliumcalcium396 No, I am not saying "there can be no visible evidence that AI is moving swiftly toward AGI, even if it is" (as you put it). In fact, I would say that measuring that movement (observing the evidence) would be one necessary ingredient toward it moving toward AGI. Like any goal, AGI must be defined in terms of an objective performance measure. And, whatever that objective is, we will only approach it intentionally -- AGI isn't going to emerge "by itself" or by accident. Whatever we want, we must build it. I'm also not saying that AGI is impossible or that measuring progress toward it is impossible -- just that measuring performance in that direction is so unwieldy as to be "hardly practical" (as I put it) -- like, almost impractical -- and, in any case, AGI is a very, very, extremely, super-duper, ridiculously, very, really, very long way off.

[@heliumcalcium396]

@EricSiegelPredicts There has been a great deal of progress in AI in the last six years, an amount of progress that has startled most experts in the field, and was predicted by none that I know of. This came in the absence of the kind of measurement you describe, which you say is necessary for progress toward AGI, and which you call "hardly practical". I must agree with that last part. Your test seems to be designed to be unperformable, a demand for impossible evidence. This is a characteristic of tests designed by people who do not want them performed, because they do not want to be proven wrong. If you have confidence in your theories, I urge you to expose them to the danger of disproof: reexamine your test, imagine you are looking at a proposal by someone else to test a theory you doubt, and remove the parts of the test that are purely defensive obstructions.

[@EricSiegelPredicts]

@@heliumcalcium396 I assure you, I do not have ulterior motives. I originally got into the field because, philosophically, I loved the idea of emulating human cognition. But you say there has been "progress" [toward AGI] -- by what measure? Surely any goal must be benchmarkable. If the goal is to be able to do everything humans can, what other possible benchmark could there be?

[@beavisandbutt-head5363]

17:24 Man is the lowest cost, 150 pound, nonlinear, all-purpose computer system which can be mass-produced with unskilled labor.

[@dallassegno]

Remember how everyone thought internet was no big deal? Contrast that with ai being the most amazing greatest thingy ever or whatever.

[@benjaminkemper5876]

Who thought the internet was no big deal though?! That is not how that went. Clearly.

[@AutMouseLabs]

For what it's worth, I am not afraid of machine learning. I am afraid of this tool being controlled by tech bros, who have shown themselves to be anti-democratic and incredibly short-sighted over and over.

[@glenyoung1809]

Too many "AI accelerationists" and fanboys overlook this weakness or don't care as it seems to be a minor thing when compared to the Star Trek future they think we're on our way to. They forget that whomever controls this technology will have enormous power over the information economy and society. Already some governments have stated their intention to use AI technology to "combat misinformation" and monitor social media to block the spread of what they term disinformation, Canada under Trudeau has put $2.2 billion to implementing such a program. Combined with a "online harms law" which will imprison anyone found guilty of "hate speech" for life, this isn't a joke either.

[@user-uf5gp4fu3n]

This podcast is very informative ,....

[@EricSiegelPredicts]

My latest take just dropped in Harvard Business Review: "The AI Hype Cycle Is Distracting Companies" - hbr.org/2023/06/the-ai-hype-cycle-is-distracting-companies

[@vitalyl1327]

I'd posit there's some good coming from the AI bubble - it's the arms race of tensor accelerators vendors, and unavoidably tanking prices of the powerful accelerators thanks to this. And they can be used for all sort of things, far more valuable than parrot LLMs. So ML in general can benefit a lot from all this BS.

[@psilocybemusashi]

you are so correct yet search youtube and there are so FEW videos about this. Thank you sir.

[@EricSiegelPredicts]

@@vitalyl1327 That's an ironic yet interesting take. I suppose that's better than how war advances industries. Misleading hype isn't as bad as the horrors of war.

[@chuzzbot]

Impossible to disagree with such a well thought out and amusing presentation There is some splitting of hairs though when you talk about whether 'Ai' or machine learning technology will possibly destroy us from its own motivation or whether it is leveraged and implemented erroneously by human forces. To spread the belief that there is no danger to society through the misuse of ML is weird to me.

[@EricSiegelPredicts]

Thanks! Well, autonomous weapons, for one, are certainly dangerous. I haven't watched what I said here for almost five years, but I can't have said that there's no danger from any technology folks might call "AI."

[@Mr.jeos21]

Is it true ai Will kill us and end the human race

[@campbellgriffin6396]

Super unrelated but when you kept making jokes I expected a laugh track, I'm not sure why. and when one didn't happen my brain wanted to explode.

[@Dina_tankar_mina_ord]

I just saw this title and listend for 1.3 minutes. So my question might be way off from the actuall content. But how well has these statements or preciction matured into todays AI capabilities?

[@Jcossette1]

Like fine milk

[@gmarkv10]

lol the song at the end was dope

[@EricSiegelPredicts]

Folks, you can also view and share on Facebook: facebook.com/pawcon/posts/1935597583161776

[@szebike]

Thank you for saying this out loud its important knowledge in this hyped up field. All the hype will harm the field more than those short term mega investments can bring because after the bubble burts it will loose a lot of credibility.

[@codyanderson7409]

They really shouldn't be getting away with all the hyping crap.😥

[@HuguesBalzac]

To be fair they had to prop the NASDAQ up somehow.

[@EricSiegelPredicts]

Check out my latest Forbes article (04.10.2024), "Elon Musk Predicts Artificial General Intelligence In 2 Years. Here’s Why That’s Hype": IN FORBES: www.forbes.com/sites/ericsiegel/2024/04/10/artificial-general-intelligence-is-pure-hype/ NARRATION: kzbin.info/www/bejne/i4WWd5qKbKdrmqs

[@navigatingel6104]

I only 6k views on this video, thanks youtube

[@anttam117]

Thanks for this episode, Eric. I am a writer and architect well steeped in the humanities, so I don't have much of the technical engineering and science background to back up my intuitions in such a way that folks will take my opinion on this topic seriously, but you just said all the things I've been talking about among friends and strangers in the last few years. I'm not surprised that people like Elon Musk and Stephen Hawking believe this wet dream. Folks like to forget that men and women of genius also say, and do, all kinds of stupid things (Musk is a king a tthat); it's just that they have the pedigree and eloquence to be more articulate about their moonshine than random street crazies. Sometimes I get the impressions that people want to believe in science fiction styled AI because of a mere need to believe in something greater than themselves. Living in modern secular societies, where even the awe and wonder of Nature has been rationalized into "resources for consumption" and where any subtle sense of the spiritual is mocked or persecuted, it seems that people are trapped into looking for a god made of mechanical parts.

[@DGill48]

I'm certain that those "people like" don't believe that AI exists. They are cautioning us to avoid the creation of a machine that duplicates human self-awareness coupled with the storage power and computing speed of a computer.

[@allisRevealed987]

@@DGill48 WTF u are talking about. go to Hollywood

[@g.v.6450]

I was somewhat worried when Saudi Arabia made Sophia a citizen. That could have given them a legal basis for not letting “her” leave the country. I hope someone considered that.

[@kalliste23]

If we're lucky LLM have given an insight into how a piece of the human mind works to engage with phenomenal reality. It's only a piece if it is indeed relevant and it's very much a part of a whole that's greater than the sum of its parts.

[@sparkyy0007]

Absolutely on point. One could construct a computer using nothing more than water pipes and valves or cams and gears to simulate every neuron in a human brain. It would be the size of the orbit of Jupiter, but could be done. But in the end, you still have nothing but a pile of plumbing fittings or scrap.

[@EricSiegelPredicts]

Yes, but I actually don't consider that a counter-argument to the possibility of AGI. Note that I'm not arguing AGI is impossible. Whether the human mind is algorithmic (simulateable) is a philosophical question that I'm staying out of. Instead, I'm on a practical point: I'm arguing that we are not taking and have not taken concrete steps towards its development. We don't know how to develop AGI (ie how to program it, how to develop the right algorithm) and what we've done, as impressive as it is, should not mistakenly be considered a step in that direction.

[@sparkyy0007]

@@EricSiegelPredicts Consider the above gear and cam ex, it's utterly deterministic and there is absolutely no difference with switches, which all D computers are grounded on. All machines we make try their level best to increase their entropy to the point of destruction, the warranty bathtub curve...lol. Only internals of living beings constantly fight to reverse natural entropic progression ie death. We are very different.

[@sparkyy0007]

@@Guy-ls5es From the beginning ? ...why, when organisms are already at this incomprehensible level of functionality including self repair mechanisms we can only dream of implementation in anything we invent. The point was not about how it was achieved, rather the impossibility of emulation, ie transference to another medium, ie purely mechanical/electronic.

[@SaphreCoalwolf]

It's like believing magicians are doing real magic

[@nataliedesenhacoisas541]

Probably not seeing that it has nothing to do with ai but, more with why doom and gloom ai content is big youtube. For me it's because I tend to read "ai" as meaning " here's how your job or something you love is going to get automated whether you want it to be or not" or ( and this is a new one) " here's something that some random asshole can use to make your entire community hate you by making fake video and audio of you saying or doing something awful." Hopefully this made sense.

[@THE-SHOCKMASTER]

The real question people should be asking is… could it be possible to combine an actual human brain to a computer and if it becomes a reality in the future would that be considered A.I. or what ????? Neuralink worries me

[@agranero6]

I am saying the exact same thing for years and people seem to fall asleep instantaneously when I explain. Apparently their brains were rewired to only respond to hype. The only real danger of AI is to believe it exists and let barely functioning systems labelled AI minding the store without adult supervision. It is good to hear someone that share this opinion. I work with ML since even before this term was widely used. The feeling I have when I hear the stupid hype like "echoes of General AI" is the same when I hear the term "Quantum 'something'" from Amit Goswani from someone that don't even know what Linear Algebra is. I agree that AI is a brand just like 4 nanometer chips are (it does do not refer necessarily to real size of the finer details of a chip but to a class of performance). I agree that it is an empty promise. But I don't agree that the concept of intelligence is intrinsically human, dogs are intelligent to an extent, cats are intelligent, in the broad sense of intelligence. When I was at college I always got flabergasted how physicists (that studied ANNs as dynamical systems), mathematicians and computer scientists ignored everything about biology and didn't even read works of Hubel and Wissel, Rodolfo Llnás, Mountcastle, or even basics knowledge about how the neuron was discovered using "La reacione nera". And how biologists lacked basic knowledge to understand a simple model as Hodgin-Huxley. I don't believe that without understanding how evolutionary pressures created intelligence (that time in the strict sense of a system capable building world models) it is possible to understand it. And I believe that at least in principle AI can be achieved, but not with any approach we have today (as all are just glorified prediction systems) and not even by a long shot in near future. This lack of desire for understanding neural networks in a more broad interdisciplinary sense that made me disappointed with academic world: everybody just wanted to study a tiny very specific thing publish papers about it and progress in the career, and then at the slightest attempt to talk to someone of another discipline or understanding something in the broad sense was a mortal sin (this reminded me of a comment Johnathan Schaeffer about the advises of colleagues when he tried to solve checkers game). That was when I got disappointed and departed from the academic world.

[@EricSiegelPredicts]

I believe the word you are searching for is "sparks" of AGI rather than echoes. :)

[@agranero6]

@@EricSiegelPredicts Yes...I misquoted...this wording makes me cringe even more.

[@agranero6]

@@EricSiegelPredicts I edited my comment to add a few more things and deviated to a rant about other thing in the area.

[@webgamer3587]

In fact, the things we humans create are all illusions, not reality. If you want to create AGI, I think you must at least create self-awareness. But this cannot be done by a single discipline. In essence, IT technology is just a tool , it’s not that fantasy. The so-called AI in recent decades is just “simulating or imitating” the results or functions of certain intelligent activities of humans.

[@kkjhn41]

Show a human a truck one time and they can differentiate a truck from a car or anything else 100% of the time. Human intelligence is unique.

[@allesok7499]

The problem is not machines becoming more "intelligent" but people becoming more and more stupid.

[@johnshortridge]

I'm not so worried about the AI.. More worried about the business guys who want to lay off people to automate and save a buck.. That path leads to economic disaster with this tool .. But corporations don't care about workers and that they are the source of revenue. That said, it would be a better world having AI "Help" our management and owners get replaced with AI.

[@kyriosity-at-github]

It works now opposite. You want to lay off staff (e.g. to exchange with younger people later), you have a great excuse with AI, which gonna replace them.