This wouldn't have been possible without Patreon supporters! Consider joining us behind the scenes and get ad-free, early, and extra content. Patreon: www.patreon.com/thenandnow Follow: twitter.com/lewlewwaller Sources: www.thenandnow.co/2024/04/26/how-ai-is-being-stolen/

My guy is competing with Christopher Nolan for longest film

Reminds me of "If you steal from one author, it's plagiarism; if you steal from many, it's research"

Honestly, the best independent documentary channel on youtube

The problem isn’t AI in isolation. The problem is AI + current economic reality. The corporate world lost its mind during the pandemic when the unemployment rate reached 30% because they couldn’t profit from it at that time. AI has made it possible for corporations to profit from unemployment. So if the unemployment rate rises to 60%, corporations simply won’t care. AI protects CEOs from the consequences of their own actions.

Holy shit. 3 fuckin hours? My boy is puttin in that work.

Unfucking believable achievement my dude. Regardless of the monetary outcome: this video will age like wine. You ought to be deeply proud.

The entire stolen labor section giving me “pay no attention to the underpaid exploited labor behind the curtain”

Takes a lot of courage to make an Oppenheimer length video about AI. Masterful.

"The question is not what AI can do, but who it can do it for."

Honestly as far as research publications are concerned, they deserve zero money for anything aside from hosting, and the salaries of their editors, it's not like scientists are paid much or anything for peer review either. Scientists don't get paid by journals to research, they do so with public or private grants, there's no rationale to justify them holding copyright when most scientists would prefer as many people read their paper for free. If a service like scihub is required for even most scientists and students to do research on the subject, and it's an open secret that everyone uses it especially in developing countries where many institutions with limited funding can't afford jacked up pricing of publishers, then the system is broken and needs restructuring. So unlike other copyright holders, a publisher like Elsevier has very little moral claim to any compensation from AI using their paywalled content. I'm also highly critical of the concept of copyright and IP laws in general, they're all instruments of monopoly with ever expanding scope given by judicial diktats, and have to be reined in to a large degree if not radically rethought. This doesn't mean I'm against AI companies profitting off of other's work giving fair compensation, but that shouldn't be an excuse to further strengthen IP laws to the detriment of all, and should instead come from some new legal mechanism.

Wow. My comment on the “How the Internet Was Stolen” video was “Watching this and seeing what's happening with Al right now is so eerie. All according to the playbook, rinse and repeat...” Seems you felt the same exact way, haha.

In an age where hordes of channels are exporting their entire production pipeline to AI in order to deliver low brow, low quality, frankly insulting content to monetize people's desire for long form video essays, you stand out as an exceptional example of how this format should be used. I am in awe of the quality of every single aspect of your work, and I cannot imagine how hard you work to get this out in the time frame that you do. May your curiosity and strive for excellence long continue

They stole all that data. Once they finished training those models, all the data was gone. They are still looking for the missing data to return it to the owners.

To me I feel like we should be more weary towards the people who are going to abuse the technology, than technology itself.

Bezos not paying out those tiny amounts of cash (that often mean life to many, still - and mean NOTHING to him) is one of the vilest stories we have 'on our books'. [puke emoji]

Kurt Vonnegut repeatedly asked: “What are people for?” No definitive answer but his first novel Player Piano explores that question in the context of workplace automation.

This topic could easily take three hours to even be a general introduction to the topic of AI

Only half way, but wow, this is absolute premium content and production. Thank you! Sharing widely.

Amazon naming its ghost work platform ‘Mechanical Turk’ is diabolical 🤣

That's an insane effort. 3 hours of this quality!!? That's a definite sub. Haven't finished watching yet but so far extremely well presented.

AI + dark patterns = Hell on the internet.

I highly disagree with that sociologist (at 2:23:51 or so)… to me, knowing and understanding why lightning strikes, how hugs communicate, etc IS wondrous and I love knowing, it doesn’t take any of the awe away from me whatsoever. The natural world is incredible and it doesn’t lose its majesty just because you understand it. Excellent video, mates. AI already made me uneasy and I wasn’t sure exactly why (besides the obvious theft issues) but this video does a great job explaining all kinds of things I both did and did not ever expect about the way AI works. I am both terrified for and excited by the future… AI could be used for greatness, but until it’s legislated I view it as unsafe and chaotic, as well as obviously thieving.

At the very least, these companies engaged in piracy of millions of items. Many people have been prosecuted and fined to the maximum extent allowable by the law and these companies being allowed to get away with it is one of the clearest examples of a double standard and selective prosecution. These companies should be fined BILLIONS of dollars if the law was applied evenly.

This was like an entire college module on LLMs compressed into 3hours, so much info and well rounded

Great mix of truths / philosophy. I just listened to 3 hours in normal speed? Must be good!

Great stuff, like many have said. Friendly reminder that there is a playback speed option on YT, if you find the pace of the narrator a bit too slow.

The thumbnail is giving me deja vu

another Then and Now video? awesome way to start a day

"AI is more important than even electricity!" *Flips switch*

294,000 titles at the generous average cost of $100 per title is $29,400,000. Basically 30 million dollars. It is possible that openAI bought a copy of each book and proceeded to let their AI “read” the books. They could argue that their AI has read all the books and is good at remembering the text.

As a former computer scientist and current philosopher, thank you so much for this video! I'm impressed by how well researched and thoughtful this is.

Awesome work! Very well researched and done. This documentary deserves to be seen by many. You could easily compete with pieces that have millions of views. The KZfaq algorithm has suggested the docu to me and I hope it will direct many more to this address. One thought - maybe dividing the 3 hours into 3-4 parts would attract more people to watch. All the best.

Amazing content as always. As a computer scientist (non AI primarily) this all rings true

This is the best material on the subject of AI I have seen on KZfaq for the past few years. The way in which you explain some concepts is brilliant. Hats off to you sir. You've really outdone yourself with this one.

When our school received our first computer it was at Tandy TI 1000 & somehow we ended up with a commodore 64 as well neither were programmed and that was what we were doing is seniors in high school trying to get it working. Once it was working we use the term synthetic intelligence since it was a plastic box and did a lot of things that were taking a exponential amount of time to do on paper. Started as a CIS major at Arizona State in 1983 but after my junior year got frustrated every time he would take one semester of classes the next semester follow-up class was not available because the language was already obsolete. Got tired of the lack of consistency and objectives in programming at the time. It is great to see it finally starting to come along. Will still be a ways away, has grown exponentially to finally be able to do what we would hope it would do all way back in the 80s

Hello Then & Now Team... I know you guys may never see this comment, but I just wanted tontake a brief moment to expel my thoughts and ocerwhelming feelings upon watching this documentary...which I firmly believe is the single most important one of its kind available on the Internet right now... Thank you from the near dear bottom of my heart for such a profoundly insightful, intriguing and thought-provoking content. I had aspirations of wanting to go into the AI field upon my current transformative phase of self-development/improvement, and current unemployed status. However, this video has surely made me revise my prior intentions, and made me look deeper into our current AI conundrum and prompts (pun intended!) me to re-evaluate my strategic plans for the long term, while helping me refocus on taking more measured next steps into my future career plans AND contemplating the future outlook of the world with AI prevalence the contemporary benchmark and one of the most important pressing issues of our time. This video has made me pledge to work on AI ethics, governance, critically reviewing the re-usage and reliability of LLM's and other AI technologies. I look forward to further videos from you guys in the meantime, and supporting your guys' work further. Please keep these kinds of top-tier quality videos coming guys, and keep actively disproving the Dead Internet Theory with life-changing online content like this.🙏🏾🙏🏾

Very well done. Had Choamsky and Lenat etc paid more attention to Turing’s last paper on morphogenetic systems instead of obsessing over symbolic systems they would not be also rans. Hugely satisfying.

1h20m : This isn’t an AI problem, this is a corporate law problem.

My contention is that this documentary requires multiple views anyways so speed it up if the duration seems daunting. Great content sir! Thank you for the hard work.^^

I've been waiting for you to talk about AI and I always knew you would take your time and make it a great one! I've been around a long time and there are not many KZfaqrs I really respect, but you are among my heroes.

I will give it a listen. I hope your three hour long video is more worth it (as it usually is) compared to Scorsese's last two movies. :D

First vid. I'm an AI researcher myself with an MS in Computer Science. This was incredibly well done.

The real time demo had some moments where the narrator commented “that was right” and expressed relief at times that the response was correct. I think they did the recording multiple times and kept the best take

Thank you for this great video and for all the hard work you put into these essays!

Whatever AI means, you and I will still be working more hours than is good for us.

Just wow.. speechless. This is pretty much all I ever wanted to tell people about AI. And I've been telling people way too much about it already. Incredible essay

Bellos and Montagu convincingly argues in their book "Who Owns This Sentence? A History of Copyrights and Wrongs" that copyright law was not created to protect copyright holders, but rather to explicitly limit their control over works to a short number of years and expand the public domain. It had nothing originally to do with rewards and incentives, and everything to do with stripping monopolies from, e.g., publishers The printing press was a major driver for this, because publishers were claiming perpetual copyright over works they printed.

Best comb-over in the 21st century

being excited at a long video, but also wondering when/how you'll make time to watch it all

I used to work in AI. You absolutely nailed it on this video. The depth of your understanding of the topic and problematic near future is spot on.

Im only halfway through this work of art and its filled with so much information and such rich CONTEXT that im in awe. For real this is changing how ill think about tomorrow.

whoa. there goes my sleep tonight.

Better make myself some popcorn 🍿

I think this doco shows clearly the first most important step to a resolution on the question of theft, plagurism, and copyright is regulation that requires all AI answers to include/reference sources.

Amaaaazing job. Summing this up so thoroughly and clearly in 3 hours so so impressive. One of the all time KZfaq gems

This video is amazing. I'm still in the first hour and it's gold!. Keep it up my man!

BEST DOC OF ALLLLLLLLL TIME!!!!!!

The level of quality you guys put out every time is incredible. A lot of food for thought. Thanks!

The worst part about all of this is that when AI is used to implement some global big brother type stuff. You will literally have the entire world against you (from voicing anything online & irl). And you can't do anything about it.

You have all the data in the world, all the processing power in the world, practically unlimited storage and algorithms to statistically identify patterns in the data that allows iterative deduction from trillions and trillions of cycles; thats the automation of automation. That's not intelligence (especially not AGI) but it can look like it if you don't think about it very deeply and listen to people who are pimping it to you non-stop.

So excited for a documentary from you. Click and watching straight away.

A long time ago, i was a young undergraduate studying AI, and we were discussing the Turing test and whether or not it was a good test of judging AI. My argument was: 'No. With enough raw 'horsepower' you could fool a human they were talking to something intelligent.' - here you are Professor. We've found what that amount of computational horsepower looks like.

Excellent video! Amazing work, easy to understand but still deep enough to cover all aspects of the topic.

OpenAI is likely wrong. Copyrighted work without explicit permission cannot be used for AI training without breaking copyright.

"stolen intelligence" me when I make stuff up

That was one of the most beautiful "call to action" requests for support I've ever seen... but hey... you might think "imposter syndrome" is a joke that everyone has, but it's a self-esteem thing you can breakthrough. Remember it's okay to do advertisements when you keep it touching while explaining it's hard for you but you're doing it to improve documentaries like this... and because you're a human supporting family members with your hard work, then when you ask them join Patreon to support you, or explain you're advocating a sponsor quickly and tastefully... (people don't know sometimes companies are doing charity to sponsor people... but they can tell if it's the opposite where you have to spend several minutes advertising a product... partner with companies that allow you to keep you real and you'll both find it's mutually beneficial with you and the other business entities)... I dunno, I could be wrong

Thank you. I got a lot from this - you’ve spanned past to future - and with this context - given pause for thought on the biggest technology driven topics of our time.

This is great, but I'm not sure it's such an absolutely certainty as you state that AI will out perform humans at EVERYTHING, that's an article of faith.

Another excellent video essay, I was one of your early followers and patrons and I've been following you when you were "just a little channel"

Outstanding video, thank you. AlphaGo especially has stuck with me philosophically since I watched the 2016 documentary. Amazing how quickly ai has developed.

Wasn't Riaz himself effectively running an outsourced sweatshop? Even harder to track for fair employment practice than the already shady services themselves.

I think one key thing to consider regarding LLMs that is easy to slip by is that LLMs are statistical models. When you ask it a question, the result you get is merely the most likely expected string of words that follow the string of words that you put in as your question. There is zero understanding of the semantics involved. It is trivially easy to get most LLMs to contradict themselves by asking the same question in a different way even in the same context ("conversation")

This channel is a refreshing break from the sea of cheap, insipid mediocrity that KZfaq has become. Good stuff!

I like that you used 70s footage of Greg Weaver skateboarding in a pool at the moment you mention "knowledge"

I found this presentation absolutely informative (and entertaining). Thank you for taking the time to create it.

We're likely to enter into a new age of AI.

I’m so excited to listen to this at work later!! ❤❤

The exploration of AI's philosophical and ethical dimensions in your article is incredibly thought-provoking. It's crucial for us to continuously evaluate how advancements in AI align with our human values and the impact they have on intellectual property and creativity. 🤖

This is some really fantastic reporting and I'm so happy to have stumbled across this channel! If I had one thing to comment on though, I do wish this video had touched more on the (non-human) resources and inputs that are necessary to make these very early and somewhat unsophisticated models work. This is going off the idea that we continue to build on the foundation created by large language models like chat GPT, but if these models are to become more complex and the models get larger (think GPT-3 to GPT-4) the amount of resources needed to run the systems also increases significantly. GPT-4 already uses massive quantities of water and power in order to function, significantly more than GPT-3 - which already used an uncomfortable amount of water imo. Personally, my thought process is more centered around whether or not these models will reach a level of self sufficiency before the very complex, expensive, and difficult to maintain infrastructure that allows them to exist decays to a point of non-function. Maybe they will get to a point where they can develop their own ways of gathering water, power, etc without a need for human systems/institutions and will just suck the planet dry trying to maintain themselves. But maybe modern society will just collapse in on itself before this can happen. The deeply complicated and fragile modern human processes that have created an environment where a AI system can even be developed and run in the first place will cease to exist. Obviously this is a real doomer way to think about it, but I don't think we take into account enough how resource use can affect the viability of these massive super-computer systems. To many of us software and even hardware can feel very disconnected from real physical resources in a way that something like a chair doesn't. But they do take many physical materials to function, even the internet is really just a bunch of fiber-optic cables along the sea floor transferring data. What's to say the continuation of these social systems is guaranteed?

wow. This is a time for a wow, very wowed

I appreciate the larger point you're building towards, but as a computer science major, I have qualms about the phrasing around neural networks. Although originally inspired by biology, neural networks were quickly determined not to be a facsimile of the brain as the field of neurology advanced beyond that simplistic conception of neurons. The technology was historically abandoned as a presumed misstep until Deep Mind demonstrated the power of back propagation. The value of neural networks is they just so happen to be a structure that is well suited to linear algebra. What every neural network does is k-means clustering: dividing the input space into k groups. This means all _any_ neural network does is iteratively apply multidimensional linear regression to reduce some loss function and we humans assign assumed meaning to the output groups. AI is nothing more than a method of calculating particular statistics on 'big data' scales that is performant on current computing technology. I think this is significant and worth distinguishing because it more intuitively highlights the fundamental misunderstanding of modern AI: they don't derive 'knowledge', they derive statistics. We're all familiar with how easy it is to lie with statistics; that something can be true yet worthless or downright misleading because it disconnects our abstract representation of a complex problem from the relevant context those statistics summarize. In research we know that a table of statistics is not knowledge in and of itself and the scientific community is exceedingly careful to not draw erroneous conclusions from otherwise arbitrary correlations, yet I feel your presentation gives too much credence to the conclusions of neural networks as if the statistics they calculate are an actual understanding of the data trained on. Moreover, unlike traditional statistics where humans apply true knowledge of the thing being studied to determine what relationships are even valid to investigate let alone the specific process used to derive particular statistics, neural networks are not given _meaningful_ independent and dependent variables instead being trained on relatively raw data with effectively hopes and dreams the 'features' it discovers are worthwhile hence why academia uses that term over 'patterns'. This is why those in industry refer to neural networks as black boxes: the math is straight forward and all the numbers are available, but claiming one understands the _implications_ of those statistics is impossible when we cannot understand the framing of the input data. Google's Deep Dream demonstrates we can even reverse the algorithm but it's of no value when the inputs do not symbolically represent a tangible concept. I think it's imperative we reiterate that current AI is just a new method of calculating statistics hence its conclusions have all the same problems of traditional statistics. Many of the complications of AI discussed are rooted in the fact that all it does is statistical prediction, it does not possess 'knowledge' that isn't a linear regression and it cannot 'understand' the context of the training data it learned from thereby giving capitalism a new avenue to abuse the public's misunderstanding of what wisdom statistics can realistically derive. ELIZA was the first AI said to have beaten the Turing test against real humans yet all it did was rephrase the input and agree with the user---we should not be so quick to assume AI has worthwhile insight let alone independent thought because the most important aspect remains subsequent human interpretation of those predictions.

Only 42-minutes in and already have to comment about how spectacular of a job you’ve done with this! Brilliant work!❤

Wow, good work. I watched it all. So much good stuff here, especially about sketchy labor practices and misappropriation of training data.

Good video. Now I have to watch it.

corps ALWASY want regulation in THEIR favor... sooo... no.. there has to be EXTREME caution with what and how we regulate. the regulations need to be on THE CORPS.

The sequel we didn't deserve to how the internet was stolen, I love your work thank you for all those videos

@ThenNow Hey have you tried using the serial numbers, or barcode numbercodes or something else besides the titles and authors to target a book? It might be using the check for the author/title but the barcode data might give you a backdoor. What about translating the titles to other languages or morse code etc? I thnik you're right, the data is still there, still being used but has a lock on verbatim copy/paste type options and will only use the data to form something that perhaps runs a 'plagerism check' after the reply is formulated but before it's given to the user. Does google have its own plagerism check platform?

Babe wake up, a new Then and Now vid dropped about another broken thing in society!!

In a world of labour-replacing automation and increasing inequality, the largest and most consequential Ai solutions of the future should be non-profit maintained, and community-owned with profit sharing. That's what we're trying to achieve with Zenate.

This is outstanding--even for someone like me who doesn't really have much interest in computers. The writing is excellent and the path the narrative takes was cleverly devised.

Excellent work! I think we may soon question the assumptions around intellectual property itself and whether its protections are still advancing civilization's interests.

Which model and prompt did you use to make this video?

This episode repeatedly reminds me of a faux documentary called "Ghosts with Sh!t Jobs." It's about Westerners with low paying high tech jobs doing things like looking after simulated human babies, or being paid to promote products in social situations.

To be fair, the corporate slave scenario of AI is pretty much possible in the western world where corporations control the government. This actually makes me admire Asia where corporations are still controlled by the state.

6:06 the "imitation game" Turing proposed was different that it was stated on this video and usually stated on media. The original Turing test was on the paper is a game where we have a woman and a man on separate rooms communicating by writing and each one must convince the people that they are the other one, and then we change one by a computer, in the same paper Turing changes the test to one having to computer and a man and a person communicating by writing only and you must say which is which. None of those is precisely the same as the definition usually described. Making the same questions to a person and a computers shatters the mirrors and dissipate the smoke. MIT used to make events and one of the things on those events was a series of terminals where you talked to someone or something but restricting the subject to a predetermined one. Many terminals were connected to a professor specialized in that subject (this fact unknown by the public) and other to programs. Usually the professors were dismissed as a computer because no one believed that someone could know so much about a subject. So being bad at being good was an advantage to the programs in a kind of reverse apophenia.

How Human knowledge was stolen

Three hours, bro?

Many other problems that you mentioned seem justified as true issues; but the copyright/IP rights matter is complicated by the fact the content was handed over freely and en-mass, and often from sources where even an attentive human would not see any mention of a third-party author retaining rights; it's like as if you inherited a junkyard and had no control over what people had thrown away there before.

One of the BEST videos I’ve seen lately and I spend hours on KZfaq everyday