The A.I. Bubble is Bursting with Ed Zitron
Vložit
- čas přidán 2. 07. 2024
- Big tech is betting tens of billions of dollars on AI being the next big thing, but what if it isn't? ChatGPT burns obscene amounts of cash daily with little return, Google's AI dispenses useless and sometimes dangerous advice, and a recent study showed that tech companies will soon run out of new training data to improve their AI models. If AI is really so costly, unreliable, and limited, what happens to the industry that has bet so big on it? This week, Adam talks with journalist and influential tech critic Ed Zitron of wheresyoured.at to discuss the impending burst of the AI bubble, the hubris of Silicon Valley, and how we suffer under big tech's "Rot Economy."
SUPPORT THE SHOW ON PATREON: / adamconover
SEE ADAM ON TOUR: www.adamconover.net/tourdates/
SUBSCRIBE to and RATE Factually! on:
» Apple Podcasts: podcasts.apple.com/us/podcast...
» Spotify: open.spotify.com/show/0fK8WJw...
About Headgum: Headgum is an LA & NY-based podcast network creating premium podcasts with the funniest, most engaging voices in comedy to achieve one goal: Making our audience and ourselves laugh. Listen to our shows at www.headgum.com.
» SUBSCRIBE to Headgum: czcams.com/users/HeadGum?sub...
» FOLLOW us on Twitter: / headgum
» FOLLOW us on Instagram: / headgum
» FOLLOW us on TikTok: / headgum
» Advertise on Factually! via Gumball.fm
All of these companies announced MASSIVE layoffs at the same exact time they announced their MASSIVE investment into AI. Then they tell us it's because of economic slowdown, that we need to "upskill", that it's somehow everyone else's fault.
The tech industry fired literally hundreds of thousands of highly qualified software developers---a skill that requires years of training AND a brain that can look at the world in a specific and uncommon sort of way, and tech companies are saying Upskilling is required???
I call Bull$#!+
That's like telling elite mountaineers they're just not working hard enough at climbing.
My theory?: this is 100% a repeat of the 'white collar recession' of the early '90s. A large swathe of offices would fire a bunch of necessary people, leaving the survivors to do the work of 3 people---so a lot of necessary work was unfinished or done badly. In the short term, stocks went up. After a while, they'd hire 1 person, so everywhere was still understaffed. That new hire was desperate, so they accepted less money. The overall wages go down, and the middle class shrinks some more.
That's what's happening now in Tech.
People on cs subreddits tell you this, it not just companies.
You know what, if AI is the means through which sillycon valley companies commit collective suicide, I'm in support of AI.
one of the biggest defense internet billonaire bootlickers present for the ultra powerful corporation and billionaire is that they "create job"
There is a causal relationship, but not what you think. Companies looking to invest in AI are trying to cut expenses to free up capital for that investment. That's one big cause of the tech layoffs. Another is that they just plain overhired.
When game executives got excited about adding NFT to every conceivable game, I remember thinking "Which gamer actually wants this? Which gamer asked for this? It seems like the executives are very excited about this, and the average game consumer couldn't be bothered to give a damn."
Now we have AI buttons on mice and keyboards.
Is anyone else getting "NFT game item" vibes from this?
I suppose a valid application of AI to game design would be in making the bots less stupid or procedurally generated levels to be less boring but I can't imagine management limiting themselves to useful applications when there's so much hype.
@@fPonias1seen a great mod that uses character descriptions to create unique portraits for generated NPCs. Pretty sweet!
Yeah but it's worse because AI has actual use cases and you can fool people into thinking that it's presence in your rice cooker is necessary lol
Gamers who wanted to trade rare items in alternative markets were pretty stoked for a new area to speculate in
95% of the ai they are pushing solves a problem the world doesnt have
In French, Chat GPT sounds like, "Chat, j'ai pété," which means, "Cat, I farted." That says a lot about where this is going and how people feel.
Yeah in french you're supposed to write "Tchat" but people seem to use the English word which makes it very confusing too.
"like wiping your ass with silk "
It definitely farted last week at the Presidential debate. We weren’t supposed to see the Biden shit show. He was supposed to be cleaned up using AI
CNN held the “debate” at the Georgia Tech basketball arena to get around the robust CNN network security
Then the two candidates had separate camera feeds (split screens at times)
The Biden campaign and media went to the DC FEMA center and deployed AI to clean him up. The telltale signs were all there. Cardboard appearance and black edging contouring his suit
My AI professor literally told me Artificial Intelligence is at best a projection of Computer Scientist goals and at worst a marketing term to fool people.
You're going to need to see beyond your teachers...
I mean... Look at where he ended up...?
@@ariloulei814 your professor is incompetent
@RAZR_Channel Ended up as a professor? Sounds good to me.
This is true. I have had presentations from companies claiming they use AI when in reality it was people doing all the work.
Haha stupid teacher. This what many people said before chatgpt came out.
Adam purposefully wearing a shirt that messes up video compression to get the AI overlords off his track is genius.
I'm not trying to be a weiner, but it's called Miore. It's a weird thing Cameras do when you have thin stirps like he's wearing. I know you're making a joke but maybe you'll find the answer intresting.
Wear more QR codes and “face / gait recognition” camouflage?
@@brutalandyI thought it's called Moiré, its a shutter effect you can also find in reality if you change your point of view in regards to two overlapping, similar raster cutouts in metal for example. But yes, I was trying to make a joke 😁
@@robertbensch7748 You were sucessfully on you joke. I laughtet at least. You seeemed like you knew what you were talking about in the original comment while the other guy was explaining something that It was UTTER OBVIOUS that you already knew.
@@brutalandy No no! That's fascinating, thank you!
Companies: Yeah, we can lose millions on this garbage program that shits mediocrity.
Also Companies: No, you can't have a raise. Costs like that are untenable.
Oh wow I feel like you are in my brain.
Apple to their unions: no, that's unworkable and too expensive for us, we're not agreeing to that
Also Apple: Hey Shareholders, we're announcing a $110 billion buyback program! (Equivalent to giving away $750k per employee)
@VideoGameStarChannelSupreme if they keep buying back stock, they can eventually go private and no longer worry about maintaining stock price.
@@thegamesninja3119 I think it is woefully expensive as it currently is, and not to mention that even if they buy back stock, the price will just keep going up and maintain a $3T market cap, so my hunch is that it'll get more expensive as the market maintains as identical a valuation as pre-buyback.
Granted, I suck at getting jokes so my bad if I forgot to laugh xD
@VideoGameStarChannelSupreme likely more sarcasm than anything else. Use of failed humor to make a point. Bubble chasers now have Web3 to try to get rich quick, so they are left with buying back stock.
Stock buybacks are a valid choice... if a company has no clue as to what they can work on next... or what it means to be human.
Again, sarcasm. My humorist very dry at times. Just call me Sandy.
None of this AI false hype surprises me. I live in a big tech city, and the attitude of "new technology is inherently good and worth using and we're not going to ask ourselves whether it actually improves our lives" is super prevalent
Watching the AI billboards in San Francisco suddenly clutter the skyline, because they're so desperate to sell it, was everything (and a reason I left after living there my entire life.)
My CEO, who is so boomer he refuses to use slack when the rest of the company does, keeps pushing for us to add AI into our processes SOMEHOW just to make investors happy. I'm so jaded now that I'm just sitting back with my tea and waiting for shit to burn.
@@melimsah slack does have ‘AI’ features doesnt it?
Haha, I was there 6 years ago but with blockchain. Thank God that startup failed and I was able to move onto different things.
Yeesh. Meanwhile my family is trying to essentially find "elder friendly Discord-like" for my grandmother and her friends' meal meetup planning. That'd be freaking AWESOME. Because they need forums, not email chains, but are also in their eighties and nineties and need it to be easy.
He is not wrong, though.
It is incredibly stupid but adding AI somehow in your product at the moment demonstratively improves your company's performance on the stock market.
Same way everybody and their mother wanted to have something involving "block chain technology " and TIOT.
From the perspective of an executive this is the correct move. Moronic, but it is the stupidity of our markets and consumers that eat up tech babble without a second thought.
The executives just do what makes number go up. To be fair that is their job.
@@nikinikolov6570 think of the shareholder's first
Utility comes later
And profit comes last (if ever)
This is the first time I ever heard of Ed Zitron and I've never laughed out loud like an idiot in the middle of housework so many times that I'm instantly obsessed with anything he ever does
Highly recommend his Better Offline podcast!
Robert Evans regales Ed Zitron with stories of awful people on Behind the Bastards a lot. He was on the Steve Jobs episodes. He is very funny and I love his voice
@@djadelaney ah I knew I recognized his voice
Google The Man Who Killed Google Search
I always assumed that if they had something amazing, they would have shown us by now. First rule: show, don't tell
It's a solution in search of a problem. Most of the specialized solutions that already exist are better, so it's hard to fit genAI into an actually useful task. And when you can, there's often a less resource-intensive option that does that useful task better. It's good at summarization or rewriting, I guess. That can be occasionally useful, but not as a standalone product.
@@epb9000if not for the over-hyped expectations I think we could see that AI in its current limited state is already pretty neat, and even useful in the right context.
"Pretty neat" and "occasionally useful" are a far cry from the corporate hype we are being sold though.
@@epb9000 Yep, add it to the list of the "Metaverse", NFTs, and cryptocurrency aka "Things that we have to invent problems for so that we can sell as a solution or at the very least turn into a grifting bubble we can make billions on through speculation and hype and then rug pull and shift to something else"
@@jonathan0berg I kinda agree, but it's "pretty neat" quality alone also comes at the expense of all the data being stolen to power it. It's like a hover car that runs on insulin. Sure, it's "pretty neat" but very quickly becomes a needless waste even if it's not hyped up as the "standard of the future".
@@epb9000 Disagree - the language problem does exist and indeed a giant step has been made with the new gen LLMs. There also exists no other software that does it actually better to my knowledge (feel free to educate me if I am wrong). That the costs of the solution is in no way appropiate to the problem it is solving is obvious. That google / Microsoft / Meta market the LLM's as a solution to everything instead of stating what it actually does (as in it solves the language problem) is another problem, but in terms of technology the new gen LLMs are really a giant leap.
"What if all of your emails could be in the room around you?"
Me, who has worked at an extremely understaffed customer help center: please no
I feel like if I studied every artist and every image ever uploaded, I'd at least be able to draw hands.
The hand problem has been solved a long time ago.
@@pin65371Eh, I still see messed up hands time to time in AI generated images. This is just AI bro cope
@@dreamwanderer9769 i still see messed up hands with humans sometimes, this is just human bro cope
@@pin65371No it wasn't you're full of shit it still does hands horribly.
@@pin65371a long time ago I saw a baby having a half leg half hand
NEVER train your AI off Reddit data.
I was laughing my ass off when I saw that Google was buying Reddit data... I knew where it would end and was not mistaken lol
Yeah they should've used discord instead
I didn't know whether to laugh or cry when I heard that....
Or fanfic websites
ALWAYS train your AI off reddit data
Saying "I don't know how this addiction to growth in the market started" as if capitalism hasn't been addicted to growth since the 1500s
That's the promise, the market expects to make more and more money as if there wasn't a limited amount of resources and people.
That’s not correct. Capitalism used to be about profits for the owners, not growth. Now the focus is usually on growth even if it means huge losses and they can only make money when they sell their ownership to someone else.
Any company that doesn’t pay dividends or distributions to the owners is a failure. And that includes Amazon and Tesla, who only have profits on paper due to timing differences. If they paid dividends they’d have to increase prices, slash spending, and/or go out of business.
@@DoctorBiobrain nah its still the same - its just shifted a little who the owning class is
its just not the robber barons anymore - its the shareholders and investors - same crap, different name
@@DoctorBiobrain line go up shareholder value has been there the whole time. Ever since the railway mania of the 1840s. Will this railway make money? Who knows. Will it even get built? Who knows. Even if it is built will anyone want to use it? Probably not. Will the price of the shares increase 1000% over the next 5 years? Yes!
Interestingly enough, socialism had its own form of growth addiction - at least, its USSR edition. In planned economy it wasn't considered good enough to implement the plan - you should exceed it every time. 5-year plan should be done in 4 years, etc. With all the ugly consequences of course.. We, as a mankind, have amazing ability to screw up any idea.
from the perspective of a coder - honestly it's been disastrous because the managerial class has bought into it, and the people actually doing the work are being dismissed or having the team size reduced because ai can spit out mediocre code snippets. and now there's less of us doing the same job and a lot of our time is taken up by trying to tweak the bad code to actually work in the broader context of whatever our product is
No one asked for “A.I.” Like this. My next phone will be a dumb phone.
Ironically, buying a Pixel phone makes it super easy to ditch Google because Pixels are one of the few models left with unlockable bootloaders. My phone has a completely google free profile for day to day use then a sandboxed profile with google services running in their own sandboxes if the app I need requires them.
I use GrapheneOS but there are a few options.
I almost got a dumb phone last week. Unfortunately, there are still a few apps I need for work and for everyday convenience. Contactless payments and bank app being the main ones. Maps is also extremely useful since I'm terrible at knowing where I am going. Might be worth getting two phones, and just having the smart phone off most of the time.
@@Imevul work phone and personal phone (and learning how to read a map to plan a route)
@@Imevul Hilarious, so smartphones are good?
@Imevul i was expecting something actually related to work, like a sign-in or communication app or something, not three easily replacable apps 😭
i have been so depressed looking around and seeing everyone being duped by this nonsense. you have no idea how much hope and relief this conversation brought me. from contra last week to this you have really helped uplift me during this stupid stupid stupid dark time that we';re all suffering through. thank you adam
Depressed? The people will keep getting duped. The people are too busy chasing dopamine highs to understand when something techy sounding is scamming them. It will only get worse. Grow up.
so many older managers legitimately thinking we would hold meetings on the metaverse
The same ones who are already in the middle of laying off workers - because the actually believe the tech is able to do most of the work already... Thinking about how people so clueless and out-of-touch have been able to fail upwards this long is infuriating. Makes you realise how much of all of this is just complete BS.
Jesus wept, for there were no more worlds to conquer.
@@kapsi 🤓☝️ “jesus wept, for there were no more worlds to conquer”
You think we'll just continue to have meeting as zoom calls or skype forever and not in a more immersive 3d environment when the technology will be available and affordable (e.g. not apple VR)
@@joannot6706 we will when it is a better alternative that can be done with the same or less effort. Right now - doesn’t provide anything more useful than zoom and requires more effort.
I know people who talk about how awesome AI is and I ask 'what exactly do you think will change in the next few years because of AI?' And they say 'everything'. And i'm like 'what exactly?' And they don't know.
@@twilightbin their stock portfolio
AI has changed the way spammers go about their business by huge amounts. I imagine most people wanting to support AI will neglect to mention that one clear and obvious beneficial use of AI, where it's made a tremendous impact.
Legit I have seen ONE company come up with a way to use AI in a useful manner--Amazon. They use it to scan reviews and make a short summary of "reviews overall" and you can ask the AI questions to see if any of the reviews answer them. It's VERY open about the fact that yeah, it's not useful on products without a lot of reviews, so it needs human input, but it's useful for using Amazon. It legit improves the Amazon experience. I have yet to see ANYONE else manage that.
Medicine. Law. Music. Your mobile. Video games. Animation. Coding.
@PatrickDodds1 hmmm
I know about one sector of law that's going to have a real big change
That is copyright law
The other stuff...
Meh
It's called Xitter now. I worked in tech for 20 years. I'm done. I hate it. Thinking about bagging groceries for my retirement years because all of the start-ups I worked for didn't pan out. I'm 63 and now...
AI😅
@@mountainmythboxmeta
Hope I look that good at 63.
Is the X pronounced "zh", like it is in East Asian languages?
Because I could get on board with calling it Zhitter..
My favorite: when companies and CEOs looked at completely crazy results and said that it was the AI programs being "creative."
The irony here is that people in the creative industry today, the digital space, including the very platform we're watching on, were accused of not being "creative" and not being real "journalists" or not being real entertainment.
All "real" means is incumbency.
It always just a data model compiling something halfway interesting out of a near infinite number of seeds and a massive training set. I'm curious how these companies take this stuff to enterprise customers. "Our system will let you fire 99% of staff for a tool that 10% chance of providing correct data."
@@SloMoMonday At the Enterprise level its about using the decades of data created that no one can parcel or use in any meaningful way. LLM not getting things right is mostly as it always has been, look to the humans not understanding the technology or just having bad data sets. Sound like everyone wants magic and is mad when its going to take a lot of smart people and years to make it better.
@@SloMoMonday Humans work off of data models in their head. Enterprise uses humans.
ai has hit a wall and plateau'd. machine learning can only go so far
Never underestimate the oil industry's deviousness with trying to remain neccessary.
Naw, they've all (outside of Russia) already gut their own investment, and no one is planning any large expansions anymore: they are instead all seeking new industries to get into. Technically speaking the return on investment for oil production is actually the highest it's ever been in human history, and it remains one of the best ways to, you know, make money with your money..... but all of the investors are moving away anyway. That doesn't happen for no reason, and (outside of Russia) we see no evidence that it is going to invert.
Which means we have once again reached peak oil, and energy prices can only go up from here. Unlike the last times this will also come with rapidly and durably increasing petrochemical costs (plastics, rubbers, gaskets, lubricating fluids, ect, ect, ect), so it's not like there's not plenty of new industries that must spring up for them to invest in instead, they're just biding their time until some of those start working out and become proven instead of hypothetical.
@@ASDeckard Could be. The Oil industry is definitely making the "We have to have AI and oil is the ONLY thing that will let us power all those computers!" argument right now very openly.
@@FFXIgwyneveryone yapping about energy and ai is simply an idiot struggling to appear relevant.
Energy is the oldest problem man has had, and the answer today is still what it was 10k years ago: the sun. Oil is going to fade into a weird historical quirk.
Do you really think oil ind. is worse than tech.?
We all use/need oil and know its filth.
We've all allowed big tech into our homes and it has destroyed democracy, mental health, esp. childrens mental health, uses a fuk tonne of oil (ref. Google ["do no evil" or whatever Orwellian double speak b.s.corpo-babble IT parasitically operates by] with its +48% carbon footprint in a few years "coz AI will save you proles"), foreal?! Big tech is a million times more sinister than big oil. AND, IT would seem its only getting started.
Petro isn’t going anywhere for several generations, we just haven’t find a good enough replacement for it.
The most prolific uses of "AI" are:
A) Helping students cheat on their homework;
B) Helping lawyers cheat on writing contracts and legal briefs.
AI has at least made non trained voice typing way better. As a dyslexic person this has been a huge improvement in my life, but the effect AI has had on the internet are not worth it.
To be fair to Google about the fiber, the other providers fought them every step of the way. They did everything they could to prevent Google fiber from entering the areas they controlled because they knew they would have to improve their infrastructure and pricing to compete.
And its still going strong where it did grab a foothold. I'm a GF subscriber paying $70/mo for 1gb. But I realize how lucky I am compared to a lot of places.
So you are saying the internet search monopoly had trouble trying to break into the already monopolized telecom industry?
@@mattmullett9521 America? It's still insane for me, as an European that pays $8.5/mo for 1/1 Gbps to imagine how bad it is in America when it comes to fiber. Especially considering how America was always considered as"that technologically advanced" country here.
I remember over a decade ago before google fiber was anywhere other than a couple of cities, there was a guy offering gigabit speed internet, and ISPs sued him until he ran out of money. capitalism is such a drain on innovation.
I'm still mad about the sugar substitute that is a protein, so it's low calorie, low glycemic, called brazzein. Naturally occurring in plants in West Africa, and the people trying to bring it to market were blocked. It's only now starting to come out, I'm guessing after sugar companies were able to get their hands on it.
Or even the drug risug that is reversible male contraceptive first made in India that was stopped because it was too effective. It could last 8 years, so it was made not available, and an American version that lasts 1 year is being worked on. Not only that, but risug is already market ready, but vasalgel is still years away.
@@ArtemGms Jesus, I didn't know that it was that cheap
5:10 oh wow, that explains the sudden push for Google to stifle adblockers. Much easier to show growth when the small percentage of those not looking at ads suddenly has to look at a dozen of them instead.
Or they just leave Google cuz Google sucks anyways. I can't find anything I'm searching for using Google anymore. At this point I miss Yahoo of the mid to late 90s where it had what you were searching and then it broke it down into a bunch of different options.
…and caused me and other people to switch back to Firefox after 10 or whatever years.
Well played, google.
Getting an hour-long ad on 'AI Strategy' on this Video is just perfection.
I wish more people would talk about what I call: Correctness Conviction. These AI's have no idea what they are saying and what it means in the larger context of reality. You can literally ask them a question that you know the correct answer to, and then disagree with it and it will generate a totally different answer. Most of the time I can get an AI to contradict it's original answer in two or three corrections.
This is a huge problem as the unique human created training data shrinks. It means that AI doesn't have a way of retaining correct information, or even decern what correct information is or even means. So, as the data sets start becoming more and more just AI generated content feeding on itself, it will skew more and more incorrect over time.
It's still hilarious to me that people refuse to call facebook "meta" and twitter as "x". I think some of it is just habit but i also think it is interesting that twitter and facebook are the names of there first and most successful products and the rebrand attempts have brought us nothing new. Its like our brains refuse to adapt until these company's prove they are worthy of these rebrands. We have built in bs meters.
I don't think that's true for Facebook as much as it is for X, Facebook is still called Facebook, what's is called Meta is the company responsible for the product, so most of the time people are using the names appropriately, they are Facebook for the social media and Meta for the company which is trying very hard to not let it die.
That's a great way to describe it. We are developing a natural immunity to corporate BS.
@@bluester7177yeah. the only reason I let the word 'meta' into my brain regarding that company is them also owning instagram.
I still think it's a shit name for them---to quote a character from M.O.D.O.K.: "They think they're deep, but they're just trite."
I REFUSE to refer to Facebook as anything other than Facebook. Everyone was using "meta" as its intended meaning Zuck saw that and decided to steal a word from normal people conversation and name his dogshite company after it. Idgaf I will never refer to his business as meta, its definitely not meta. And I already don't use twitter so idgaf what its called. Actually to be fair to Muskrat, "X" was an old company name he had before his Paypal days so at least he already had the idea rolling around in his head
@@stereo-soulsoundsystem5070 X as a lone name is so fucking stupid. He's trying SO HARD to be cool circa 1990
I'm waiting for someone to upload a script that gets AI scrapers to destroy themselves
Look up Nightshade. It's not quite that, but it's something us artists are excited about.
What? The thing tech bros tried for years to make us think is cool despite it demonstrably sucking in every way turned out to be garbage? How unusual. How out of character for them.
yea, it sucks so bad that the 2024 McKinsey report on AI found that a mere 77% of companies are using it
@@Brandon82967 that's not nearly the endorsement you might think it is. Companies often use poorly tested and ineffective technologies and methods because they're told by a few other companies/people (such as McKinsey, which is quite possibly the single worst company in existence) that they need to adopt whatever the new fad is or they'll become irrelevant. They do it not because it's good, or effective, but because supposedly everyone else is to the point that everyone ends up using something that could be complete garbage. That's just how the business world actually works, unfortunately.
@@Brandon82967did you look at the report or just spouting headlines. Sure, 77% of companies are looking at “AI” because LLMs are the current hype there are teams looking to see what they can do with it. Hint, chatbot it’s a horrible user experience. Some parts of the tech behind LLMs are extremely powerful tools. (Such a sentence transformers based embedding). The fundamental tech isn’t sexy and is harder to demo in the news. But it has a much lower run cost and dramatic practical upside. Parts of this technology will be around for the ages but not every company is going to have “digital assistances”.
@@Brandon82967 damn bro it's almost as if corporations are looking for any excuse to shave off workers and expenses for a usually shitty product
@@chillyoil528 but that product is replacing shitty workers so it's ultimately worth it.
I personally would love to see small startups popping up again that focus on making good products, are as self funded as possible, and refuse to ever go public. A friend of mine and I have spoken numerous times over the years about this topic and have both agreed we'd rather pay the employees the same, if not more than we take, give huge yearly bonuses, as many benefits as possible... you know, actually give a shit about the people making the company money. We'd be happy making just enough money each year to keep the company going. Having fun building stuff we love and able to do that until we decided to retire. Sounds like this may be a much larger shared mentality than I thought.
I like the idea if only I could make something that the market needs and still yet unfulfilled.
I've been trying to tell people that A.I. is a bubble and everyone has given me "that look". But I'm glad there are people out there yelling about it much louder than I ever could.
How is deep learning a bubble?
@@pararth4476 The current hype is a bubble. Deep learning as a technology is great but it won't deliver what AI companies are promising (at least not within the time frame they suggest). We might be heading for a new AI winter soon.
@@alexanderst.6529ai winter not going to happen anytime soon. Momentum is too high, and even if all the science will suddenly stop now, it is done enough to keep specializing, improve on real data from real tasks ai will be tried at after trained on common data and developed e.g. emergent skills, scale using more computation for at least for decade.
Gartner has been saying it's at the peak of the hype cycle for at least 6 months
@@pararth4476sorry that you're holding an Nvidia bag you bought at the top but I ain't buying it when the bubble pops
I'd like to relate an in industry story, from game development, visual media.
Someone tried to work with some people who provided good, but still incorrect work. They received instructions on what to correct and what was the minimum required standard. They also had a number of traditional artists working on the same project.
The traditional artists were able to adapt their work, correct the elements present and complete the work.
The AI prompters, produced additional images and those were less suited for the project and only progressively got further away from the desired end state. Every single time was a new image, a new prompt. Every single refinement proved to be less suited to the project at hand.
I've explained it badly, left out the details I couldn't confirm, couldn't remember.
Generative AI is a verbose if badly trained parrot, just repeating what it knows. Giving answers that at best might, be close to what someone actually wants.
As they said, probabilistic.
It can have a role in generating ideas, frameworks, and prototyping. It's easier to generate a bunch of things that look sorta-kinda-maybe like what you want and have an actual artist use that as reference, rather than trying to explain what you're looking for in words or badly drawn napkin diagrams or spending a bunch of time trawling through google images and pinterest.
That's a fairly legitimate use-case and fulfills a need. That's... not worth trillions of dollars and gigatonnes of CO2, and should not be involved with anything anywhere near the final product.
Then those were badly uninformed and incompetent genAI users.. with Controlnet and IPAdapters it's ridiculously easy to modify an image even pose subjects just as you want.
@@hypotheticalaxolotl I still prefer napkin doodles honestly, I'd rather make exactly what my client wants then what they just settled for when looking through the generated options
@@FuzBrain I mean in either case I would hope they wouldn't just dump them on your desk and say "Go off, king." Either way they should be guiding you through what they're handing you, how the scribble/generated doodle has this failing, doesn't reflect that idea, has a decent pose but they aren't enthralled by it, etc. But fair enough.
@@Hosea405 Sure 🙄... Like this problem isn't EXACTLY what most people encounter when trying to have AI-prompt generators create anything more intricate than a thumbnail for CZcams.
No-one has ever shown - what you're claiming is "ridiculously easy" for anyone who knows how to use the tech - is actually possible...
Not even the big tech companies pushing this overhyped garbage are able to show that it's "easy" to change small details here-and-there on an AI-generated image... in any reproducible and predictable way. It's always a version of "prompt and pray" again and again - until you hopefully manage to get anywhere close to what was asked by the client.
My husband works in tech. We're very concerned what's coming later this year or next. When the AI bubble bursts, it'll take a number of industries with it, and likely be the thing that bursts a bunch of other bubbles.
Oh the curse of living in interesting times
You are either lying, or your husband is an idiot.
Here's the thing about bubbles, *You can't see them*
People think AI is a bubble because they keep comparing it to .com.
Main concerns with 'bubbles' is people dump money into a thing, and nothing profitable comes out of it. But we are already seeing returns on basic AI ( not even pushing agi groups ).
Where the new modeling techniques help critique existing narrow AI systems.
If you have savings, get out while you can. It’s never too late to change careers, but it will soon be too late to hang on to the one he once had.
Thing is, there is a reason why at least 5 countries have outright banned AI, some since GPT3. Other issue hardly mentioned is the CO2. The amount of energy used to train these LLMs is shocking, it is is not sustainable. Big tech knows this and you can find out just how bad if you do the research. The reason it will halt is ironically not related to the technical problems that are facing but instead climate change. There is a reason why FinalSpark is taking a different approach (yes they mention this issue) but what they do is controversial, since it deal with ethics... after all, using 'human biological tissue'... best is to research this.
@@stonykark yeah... We just bought a house and have a baby on the way, he's looking at other options. His work has always been a bit unstable, usually going from one startup to another every couple years, but his current job is the most stable yet. Just not sure it's going to stay that way :/
I'm surprised there even was a bubble. I first got interested in LLMs due to playing around with AI Dungeon years ago, and in the early days they straight up told us that they couldn't afford too many new users. Now that AID has developed into Latitude Games, one of the few semi-ethical AI projects, their unofficial slogan is "Running AI is expensive." They say it in every dev update, instead of claiming that AI will be a magical money generator.
I think I'm going to be eternally grateful to the guy who came up with AI Dungeon thanks to it showing me that AI is both brainless and expensive before it even became a big, controversial thing.
So because you saw an early version of it that didn't do so great, that means that all future versions are bad too? Are you aware of what all we can already run locally on standard consumer hardware?
@@zuriel4783 Why do you think it didn't do great? I'm still subscribed to Latitude Games all these years later.
Why do you assume it didn't do great? I'm still subscribed to Latitude Games years later.
I love to see people really talking soberly about AI. Shit is insanely over hyped right now and people are just parroting "just wait till next year!"
"Next year" is when Teslas become robotaxis, remember? Or was that this year? No, this year was when Elon said he'd be landing a manned mission to Mars... Weird that I don't really see any of this materialising any time soon... Lmao.
I'd be glad to talk soberly with you without referencing future possibilities 😊
The thing is next year will be better and you will be left with egg on your face...because hardware is accelerating at an exponential rate. It will peak 2028ish...but until then all people like you do is sell ai to the dumb even more as something special by proving the parrots right...as the hardware will improve.
@@onjulraz754 I think the improvements in gen AI are at a plateau where more training data will not lead to significant improvements in the models
The fact that people simply do not understand the difference between science and science fiction is constantly frustrating.
I have friends who think it's possible we'll have chips implanted in our brains fully displaying the Internet in the next 20 years :/
@@michaelzzakiwho cares? Shouldn't we concern ourselves with the now, and not what people think might be possible in twenty years?
I'm sure we all have false beliefs about the future
@@michaelzzakigo look at the interviews of the guy with the first neuralink implant who can use it to control cursors and plays civ 6 with it. It might not be able to do vivid video in 20 years but probably at least a basic UI. It'll do wonders for disabled people at a bare minimum. Probably won't be used by everyone unless they can solve a lot of complex problems quickly.
@@JoeAuerbach I bet 95% of people in this comment section don't even know what an API is
I think we should start to distinguish between science fiction and science fantasy, the first being grounded in our current scientific understanding of the world and the other, well, being fantasy
As a graphic designer who was working at a big marketing agency when midjourney first came on the scene, I can pretty confidently say that ai is not going to take my job like people were saying at the beginning. It’s a very cool tool that makes specific parts of my job easier but I’ve been using it for almost a year now and at this point the improvement in capability has somewhat slowed.
Yes it is super important to be up to date with new tech in this field but turns out that a lot of the initial scramble was hype.
It’s not quite as different from other technological advances as we initially thought, imo. It’s changing things for sure but it’s not going to change EVERYTHING
These AI tools dont take peoples jobs. They just are a tool for people to use. For you it might be a way to just come up with some concepts. The only jobs it might take are people that are doing very repeatable work. Even with those people it would free them up to do more productive work.
Do you know about the relatively new -p personalize mode i woud suggest you try that i think its a massive update to the old version.
Ed's 'Better Offline' podcast is definitely worth a listen y'all
Thanks for listening!
Those insane Google AI answers are actually gathered from really old forum posts and web articles, most of which were parodies. AI simply can't handle irony.
Right because it only knows how to produce sentences like the ones it has seen, it does not know what any of the words mean or the mood, voice, or intentions of the speaker.
@@timothystamm3200 Yup, computers are not human. (shocking...) Computers cannot decipher human context or meaning in anything they copy. Sure, you can "program" a computer into deciphering mood or meaning but that's not the same thing. Also, not even the greatest programmer in the world can do that for every conceivable thought or question.
666k subscribers. No one else is allowed to sub now. Let us preserve this glorious achievement
Another thing about Sora and the woman in the Hong Kong street evolving into a full blown movie - how do you make a movie if you can't even have the same person across shots?
ED ZITRON?! THIS CROSSOVER 🔥🔥🔥
Love his newsletter
@@Obiwancolenobigood god I'm gonna puke
@@pkthunder416 Out of excitement?
I know! Been waiting for two of my favorite people yelling about AI to get together lol.
Last week my logitech mouse driver poped up anoucing their new ai stuff.
what the fuck does my mouse drive need ai for, this is nft all over again, but worse
I'm an engineer in the industrial automation sector and I think there's some really cool stuff going on there. The AI hype doesn't get much of a foothold because everything we make has to actually be useful to the customer at the end of the day, otherwise they rightfully fuck off to another supplier. It isn't sexy, it's really difficult, and there's very little low-hanging fruit, but a lot of smart people in this country are still working on solving real problems.
Some people will say ai but just doing automation which is not ai.
@@timothybrown5741 Most of Modern automation has at least some level of Machine Learning. Which is what the term "AI" actually means. It is not like we have any actual AIs yet.
Immediately clicked, I know this is gonna be a banger
yeah a banger of nonsense
We hate tech. We hate the future. We hate all solutions. We want to be permanently miserable. Yes. We are "progressives". Vote for us.
@@angrygreek1985 how is debunking marketing lies 'nonsense'?
You didn't have ANYTHING better to do than listen to two nerds sit in the ac and yap? 😢
@@pkthunder416 Imagine thinking there are people that work from home and are working on a project that does not require that much focus, therefore can afford to watch such things on a 2nd monitor. Insane, right.
First Contrapoints and now Zitron? You’re on a roll
Has he done one with Dan Olson, yet?
Because around 31:00, he mentions that "no one's mentioned the Metaverse to me in around 18 months," and there's a certain Folding Ideas video that came out about 16 months ago...
@@NeutralDrow He did do one with Dan Olson last year. They have great chemistry, actually, it was fun to listen to
@@bluegreenmagenta Well this is embarassing. The interview he did with Dan popped up in my recommendations...and I'd already watched it at some point. 😳
One of the only actual uses I've seen for AI that's somewhat impressive is a video codec my friend and I have been working on, basically just using AI as a smarter way to compress a file, because that's something you'd actually works well for, it's terrible being creative and actually coming up with proper information, but when you're approximating colors and small changes in resolution of different video artifacts, ai is kind of perfect for that
It actually does good for small dynamic problems.... But as soon as you increase the scale it starts hallucinating
I'm SO glad you brought Ed on the show. I've been listening to Better Offline since he started and I'm so angry about the state of the tech industry. Chasing indefinite short term growth is ruining everything, and to see these former 'innovators' fall to capitalism when we know it's broken is infuriating.
Just being sold a bunch of buzz words and fake hype by tech CEOs and watching people trip over themselves to platform them without ever pushing back is such a disappointment.
I want this new disruptive nonsense to stop and be recognized as a fancy "auto-complete" that it is. I've been a programmer since 1991 and I've seen so many things like this fizzle out.
As a programmer you should be well aware that GPT is way more than a fancy auto-complete. It's neural network architecture... modeled after our own brains, and processes information like we do.. so much so in fact, that neuroscientists are looking at GPT and neural networks in general to get more insight into our own brain processes.
@@Hosea405 The architecture is very complex and interesting, and I'm sure there's things to be learned neuroscience wise from all of this. But FFT architecture is nowhere near like a brain besides the abstract concept of neurons and strengthening/weakening connections, it's like a child's toy compared to all of the different complex biochemical processes going on in your brain 24/7. And just in general, LLMs aren't made to "think like a human": they are trained with the goal of *creating text that looks like it was written by a human*. That distinction is very important, imo.
Microsoft now added “Summarize with Copilot” as an automatic mouseover pop-up when any text is selected in edge.
“Summarize” has been an effective feature in MacOS and iOS for literally two decades, and no one cared then or cares now.
@@Hosea405- Except it’s not. The current AI is a fraction of a fraction of a fraction of the complexity of a human brain. It’s not even close. A cognitive scientist studying ChatGPT to get a better understanding of neural networks is analogous to a gastroenterologist studying a Cabbage Patch doll to gain insights into the human digestive tract.
@@periapsis413 This "it's like your brain" thing has been a parroted thing from AI cultists err... enthusiasts for a while now. Trying to explain that there's way way way more happening in your brain than just some electricity going through some neurons just doesn't land, they refuse to believe it.
tech industry: 'we'll fake it til we ultimately don't make it'
Or fake it 'till we break it. ;*[}
The Elon Musk model?
"We will fool the ignorant until we have lined our pockets and then bounce"
That's the truth really. They feed of the ignorance of the general public. If people had any decent level of statistical mathematic understanding then this self driving ai car would have been a joke to them back in 2017.
@@Skutnu i realized elon was a joke when he got pissy at the rejection of his little submarine doodle during the thai cave rescue and called the diver a pedo.
@@direktive4 Oh yeah, he is an overgrown child. I found out recently he has a transgender kid, but, on twitter, he seems to be against the LGBTQ+ community 🤦♂great dad
I'm tired of all the stuff that'll "change my life". I'm happy with my life and I don't need any company to change it.
large language models are both a legal and social liability of such a type that everyone should be outraged by anyone using them. they have no right to use the training data in the first place and that they do is a personal violation of thousands of people. all these companies using AI need to be sued into bankruptcy for the degree of harm they are doing.
agreed. it alone can cost so much fraud, identity theft, personal data theft, and seggualization of adults who didn't consent and of 18 and under. Those are the worse case scenarios that are happening right now.
AI will soon consume itself and the outcome won't be pretty. There's a reason why ChatGPT-3's training set was cut off at Oct 2019: ChatGPT-2 was released in Nov 2019. If they had kept training GPT-3 past Oct 2019, it would have been contaminated by output from GPT-2, as there is no way it can distinguish between AI and human-generated training data. ChatGPT-4o has digested training data up through Apr 2023, which includes GPT-3's entire publicly accessible output. There is no turning back, ChatGPT has already begun consuming itself.
GPT-2s outputs were actually extremely bad in the context of todays models. So bad in fact, that almost noone will have used it to produce texts online in numbers that would meaningfully affect the training of GPT-3. Therefore, this is certainly not the reason for the cutoff of the training data, at least not for GPT-3.
It's like the movie "Human Centipide", only with LLMs instead... "GPT Centipide".
None of this is true 😂
GPT-3 always had a training cutoff date of September 2021 when it released in November 2022. today its training cutoff is june 2023.
@@mycelia_ow AI haters have absolutely no idea what they're talking about lol. They don't even know that LLMs are purposefully trained with synthetic data now and outperforms human made data.
The problem is not AI. It's capitalism.
All roads lead to Wall St.
Why not both?
Yeah I was wondering why they made it seem so specific to tech. Like it's very basic capitalism? endless growth is basic capitalism, look at oil
I'd be more specific, the issue is with LLMs or any similar model. That new wave of generative systems while we have much more rudimentary and simple machine learning tools that work just as good but this obsession with talking computers and image/video generative programs is giving a bad look to what is was supposed to be in the first place. The money is going in the wrong area...
It’s both. It’s definitely both. Capitalism makes everything worse, but A.I. as it’s currently envisioned and used are also bad. Example: using A.I. created by scraping the entirety of the internet to try to make written communication more uniform instead of improving education hurts everyone. There should be room for individual humanity. It makes the people writing feel like people and not like we’re just doing a more complicated madlibs. It’s worse the people reading things because it just becomes reading another form.
Seeing the AI images becoming inbred due to the massive numbers of AI images online that gets fed back to the AI made me stop seeing genAI as much of a threat tbh
Like at the very least they will be forced to add SOMETHING to those images to differentiate them from the man made ones, so there will always be a way to tell
This is somewhat the foundations for the Dead Internet & the Dark Forrest theories.
Eventually, so much of the internet will be full of so much artificial nonsense that it will supposedly drive real human users to a handful of private and secure services, in hopes of getting away from the bots.
Honestly, with how much nonsense is already on the internet, I feel those theories aren't too far off.
I'm a vr evangelist. My first vr was the hp reverb g2. Its a game changer, it's night and day. Immersion is like nothing else. Being able to move, turn your head in a 3D enviornment is something else. I don't know what it is but I don't get headaches, migraines, or get dizzy from playing vr. There are some issues still though. It also has a lot of applications outside of games. I do a lot on the computer, I have 3 Monitors. This is largely because I do 3D modeling, I have a panel for esearch, then I also have youtube videos. like this one running in the background. Having complete 360 space to put floating screens, or notes, et cetera is amazing. Running a bunch of tabs through vr, and also running my scene with millions of polygons would take a lot of processing power even though my computer is pretty decent. There are a couple of problems though
-The cable
-Control system. Haptic vr gloves are a solution though, but they take a lot of work to make and I haven't finished mine yet.
-The space to walk around/move around, lack of truly affordable vr treadmill)
-Setting everything up, the cable management, moving things around for space purposes, using the onlydisplay port plug, the paranoia that I'm going to hit something. It's an experience that really needs a lot of setup and prep, and if you don't have a lot of space oof.
Vr is definitely very good, but a lot of the technology surrounding it needs to get better, as mentioned the haptic gloves, treadmill, cable management, etc.
@@Arunnejiro ‘i can see all around! Im walking!’
I love VR. I have a quest 2 & 3 and play for at least half an hr most days (without cables). The only time I have issues with nausea is when the apps stutter or freeze which is rarely a problem with finished native apps (as opposed to betas etc) but almost always a problem with PCVR. Not sure what in my setup is the issue but I've never gotten it working nicely.
There are a ton of ways it could get better as you say. I'd love to have foot tracking since I use it mostly for exercise and being able to add kicks and whatnot to existing punching-focused apps would be amazing (when space allows!)
I'm really hoping immersed Visor is as good as advertised, would love to do my programming in VR :D
I said, prior to the launch of the vision pro, that VR had three problems. Not the problems you listed, actually, as I see those as trivial in comparison.
VR is sweaty.
VR is heavy.
VR is isolating.
Go read reviews and see what negatives they pointed out. Ask people why they don't use it more. They won't say "because the cord is a pain" they'll say "because my face gets so sweaty".
@@GerryRR bulky and heavy are issues for sure IMO took me a few weeks to get used to the weight (though mostly I've heard complaints about nausea) but since I use it mostly for exercise, I expect to get sweaty 😆 I soak most of it up with a bandana. You can get third party head straps with fans built in to help with that too, I've heard they are quite effective and I'm considering ordering one soon.
Isolating? kinda depends on what activities you are using it for, but for some, that's a bonus 😜
I won't shed a tear over its passing. There are specific use cases like medical diagnostics that I'm excited about, but as something that is a pervasive part of our daily lives, I just don't need it or want it
Pretty much this; I think it could have a few specific use cases (one that came to my mind was using AI to run protein folding simulations) but just stuffing it into everything is ruining so many things that were perfectly fine before. Can't wait till the bubble bursts.
@@QueenofTNT once crypto lost its sheen, those bros needed something new to get excited about. But rather than improving continually, it's already peaked. It needs training data and with so much erroneous AI-generated content to be scraped, all of the "art" will look the same and all of the text will become increasingly unreliable.. And with AI-music finally having pissed off entities large enough to bring these companies to heel, we're going to find that generative content isn't just bad, it's illegal.
People are downgrading to Windows 10 because Microsoft wants AI everywhere all the time and it just ruins the experience
@@dyscotopiacrypto had promise blockchain had promise only problem was that was still a promise and Bros started selling it as a reality
As an ai enthusiast, I’m hoping the AI bubble bursts. Nvidia and AMD are ratcheting up supply for data centers because that’s where most of the demand is coming from. People who have more money than sense trying to appeal to other people who have more money than sense. Nvidia and AMD are making all these gains in performance but the end consumer is not really seeing it because it’s being bought up by the line goes up crowd.
What the bubble bursting will mean is cost of compute will go down as demand goes down.
Only when cost of compute goes down will open source truly thrive.
Cost of GPU anyways, since that's where most of the AI processing is done. Would be nice to have MSRP or below on RTX cards again
Cost of compute always steadily drops no matter what happens. That's literally the story of computation. Whether you're citing Moore's Law, or more broadly and accurate, the Law of Accelerating Returns.
Adams description of the current economy was really sad. I would not have watched this without Ed.
Ok, I’m a software engineer. Been a SWE for almost twenty years.
This guest sounds like the reason I haven’t taken a day off since April 2024.
😢
"Actually It helps with my productivity"
I've been hearing this from people around me since 2008 when I started Working in Advertisment.
I work with video editing and motion design and I see all these tools, hacks, methods to increase my productivity and people hoping on them like It would actualy make them more productive and increase their revenue. And I'm hearing the same thing with some new AI tools with the exact same pitch.
Throughout all these years with new shining tools I haven't used almost none of them, and my productivity has been exactly the same as other people in my area?
And you wanna know why?
Because no one has invented a productivity tool to make Product Owners, Directors and Clients answer e-mails FASTER.
that's literally what an LLM would do though. just autosuggest your way through a reply email
It’s good for coding
That's called a lead pipe.
@@paulscooter4076 If you think its good for coding, then you have never been good at coding.
It's funny because if clients actually use LLMs effectively they should be able to respond to emails faster, both from cutting out much of the drudgery of the process with writing assistants, and adding layers of assistance in managing one's inbox and in summarizing email contents. Basically we're still at the beginning of the the adoption and literacy curve, so I wouldn't discount it just yet.
$700,000 a day would change a lot of lives very quickly.
Doesn't that, in and of itself demonstrate that the current model doesn't work? A company can burn that much cash on something that may never work but some people can't get the basic medical treatment they need? 😢
Right, and they are never required to put some of the gazillions aside to retrain the humans whose ways of making income are erased by new technology.
@@courtneybrown6204But only if it works. If it fails, we can all keep working our jobs.
@@courtneybrown6204vote for a Democrat if you want University to be paid for...
This factoid is very out of date and indeed openAI is running a profit from its subscription base.
Not trying to seem condescending. I don't feel you fully understand the economics of new technology. Just about every single Breakthrough technology in the last 150 years has been initially un profitable including electricity. Will every startup make it, of course not. Does that make every attempt a wasted effort and money , no in my opinion. Perhaps we have reached a point of extreme diminishing returns in turns of technological advancement and human happiness. Hell maybe even real economic returns on every new patent is also falling, but I think for now its still worth is.
"I want ai to do my laundry and dishes so that I can do art and writing, not for ai to do art and writing so that I can do my laundry and dishes." - Joanna Maciejewska
In do time. It just turns out that work like accounting and other seemingly complex digital work is easy compared to doing 100% automated 3d real world work. The tec is there already here today or currently being worked on to replace a vast majority of human task and labor. Only thing holding the tec back is more societal and economic then purely technological.
Why do I have the feeling that so many rich people always try to solve the wrong problems?
Because you are correct, they do.
They don't want to solve our real problems, because our real problem is them.
Thank god, I thought I was going crazy. I've been saying this shit to people around me ever since the google next developers conference where literally everything they showed was AI and almost none of it was even accessible. None around me seems to share the sentiment, It's all tech demos pitching a solution without a problem.
So nearly all of the tech major tech companies in the world are putting all their resources towards AI because they don’t believe that it has large potential and all of the engineers within those companies have done absolutely no due dilligence on it and are just following a hype? Sure.
"EnRon adjacent" 😂
they burn cash, energy, water resources, jobs, environmental goals ....
The biggest threat is to Concept artists for movies, games, shows, story board artists, designers who do not make moving media themselves but design things to be used for another artist to make into something real or for animation and low paid writers who are being both being urged to be paid to paint over or clean up logical mistakes in writing or visual images after the AI slop churns something out as the humanities lacking techo bro csuite looks down on them as if their skills are considered devalued, and to b e replaced or tell them to get a 2nd job or pick up dishwashing. Stock images, factual information, photos, small time positions are being hit the hardest. things have declined. EVEN if it pops, these things despite the copyright violations and power concerns, I see AI ads for fake phishing attempt games online every day, I see obvioulsy melting people made with lower quality generators for ads for things that would have paid for stock images or a photoshoot or a real drawing or have that money go to someone who isn't exclusively a giant corporation that is parasitical to the "product" it produces.
This
His point about Facebook justifying the enabling of a terrorist attack in the name of growth is a solid point. But I like to also remind people that Facebook is complicit in enabling genocide in Myanmar. I, for one, believe the leadership of Facebook should face justice for their complicity.
Is also complicit in the insurrection in Brazil and many other things, if it makes money they don't care.
@@bluester7177 and defending a certain country in west Asia doing mass deletions of peoples right now and going against the detractors via Instagram.
Only ten minutes in, but damn is Ed Zitron ever on point. As usual. Credit to Conover for the good guest and being clued in enough to hang with EZ.
These kinds of things will never change as long as we keep supporting a system that rewards encourages people to chase the easiest, quickest buck
It's clearly not remotely ready to do any of the things big tech wants us to believe it can do. It seems specifically extra bad at doing any of the things they want it to do. It's embarrassing how far they missed the mark with this. This may be the single dumbest roll out I've seen from tech companies.
I think the issue is that it is the most flexible software in human history, it takes true imagination and it takes time to develop all the use cases. I have developed several at my work.
We use computers at work for deterministic, business logic driven processes. Same inputs produce the same outputs, awesome. These GenAI models are probabilistic blackboxes. You can’t trust them to do your regular work, as you can tell from the pivot to explaining it as an idea generator. That’s the part humans want to do! To paraphrase a much smarter poster, I want AI to do my laundry and dishes so that I can be creative.
@@reedriter2 I think these models could have uses. I think mostly fringe uses that have far to limited usefulness considering how much energy they require. However, they aren't remotely suited for any of the uses they are currently being tasked with. They are too chaotic. The people pushing it have internalized the marketing spin of calling these things AI, but there's no intelligence here.
Depends in what area... it's making huge strides in cancer detection and other non-general areas. The smaller and more focused the dataset the exponentially better it is. I think the issue was putting out a generalized model with too much data and not enough parameters. Makes for an inconsistent experience because the weights are spread way too thin.
Concur - it has good applications for scientific discovery. But LLMs as a productivity tool in generic office environment? No.
As a dev I always knew this was just going to end up like crypto, when every company has the same language model under the hood ie opengpt you know theres a massive bottle neck in how the technology works. The only reason why we had so much invoation with phones and computers is there was hundreds of companies working on different unique approaches to the technology. Now it's all just a couple of monopolies that don't care about innovation but rather just money, a product that is only functional on extreme extreme amount of server usage that only said monopolies can achieve and way to much hype via marketing. But glad this video exists to explain this far better then I could ever do.
agree with most except the pat about only functional on extreme amount of server usage... there's 50+ local LLM models at the moment that can all be run on your home computer, some comparable to GPT
The reason social media sites don't work as social media sites is because they're not social media sites. They're ad delivery platforms.
One thing I have found that chatgpt is good at is taking a vague description of something I can't remember (old movie name, science word, historical event, etc.) and give me a word or phrase or list of those and it almost always gives me what I'm looking for.
I truely can't wait what the next big thing/buzzword will be once 'AI' has gotten stale and has joined the ranks of NFTs, and the blockchain.
Especially because it's Limited Intelligence (LI) rather than Artificial Intelligence (AI).
@@nielsjensen4185It's all a 'Limited Intelligence Enterprise' (aka. 'LIE')
@@nielsjensen4185 It's all a 'Limited Intelligence Enterprise' ( aka 'LIE' )
it's really not going anywhere, what actually makes you think that?
@@jamessderby Because there is currently an enormous hypetrain and marketing massage when it comes to AI that won't last forever.
Well, we now know where our Ed's at!
As a software dev I can quantify the time AI has saved me in my day job. It's very good at solving simple but obtuse problems I don't want to spend energy on like "make a function that inserts an item in the middle of the list then flips the list around where the item is inserted" kind of thing, or my new favourite thing - "using this API write a request that will get me back this data that I want".
It's pretty bad when you start trying to solve novel problems but in general I'm 10 to 20% more productive on any given day not having to google shit.
One observation though, I used to google what I AI now. I used to use Google all the time but even before AI I stopped using it because it just wasn't helpful. I genuinely believe AI would not be as popular as it is today if Google didn't start gimping their search algorithm for profits. The terrible nonesense Google has become now has made AI seem more ground breaking than it is.
They just want an "AI" that can take responsibility and liability for their abusive algorithms. They know its all garbage.
Chat GPT is trash because, no matter how many times I asked it, it was ADAMANT that you can't make a guinea pig Prime Minister of the UK. And if it can't work out important things like that, then, what is it even good for?!?!
Gave me a step by step guide 🤷
Weird comment
@@onjulraz754 it's a good thing the election is tomorrow. At last my dreams will come true.
Since it's very agreeable, you can force it to say yes. Not as fun as Bing AI when it was in it's testing phase. Loved to see it arguing and being passive-agressive with people
I've been in the machine learning field since 2012/13. A am yet to create a chatgpt account and use it properly. It's nothing more than a sorting algorithm combined with nlp being done a crazy speeds (not crazy if you understand the exponential improvement of hardwares). It will make the public even more dumb and susceptible to even dumber scams the tech bros will do 😂 it's just easy money right now.
@Skutnu well then no wonder! 😤
Google fiber was probably the best internet connections I’ve ever had. Been secretly hoping they’d keep expanding it but alas…
Google Fiber started up back before they removed "Don't be evil" from the mission statement.
It's The Wizard of Oz. "Ignore the man behind the curtain!"
I've been in tech for 30 years. Most people in tech companies are, frankly, weirdos who don't live in the regular human world (generally, the engineers are all on the spectrum and the sales/management are all delusional psychopaths). I've been WFH since well before the pandemic just for my own sanity, having to deal with those freaks in person every day is too much lol. LLM "AI" is absolute garbage and will never meet even a fraction of the hype promised. I've seen a LOT of tech fads come and go, and "AI" has to be one of the dumbest of all time
I'm a cartoonist/animator who can't wait for it to fizzle out so colleagues can stop freaking out and tech-bros can stop gloating.
@@satchillananda Hardly a "waste" - I've got to live in 5 different countries and collect multiple passports, all the while earning loads of cash. So it has its upsides.... Not gonna lie though, the software industry as a whole is a badly-run disaster zone full of scammers, psychos, and weirdos and frequently very very Stupid fads. But then again, so are many other industries and professions...
As a tech person myself, I don't know that we're mostly weirdos. But, after working in AI for over a year trying to develop actual solutions, I am stunned that while these new multi-modal models can recognize hand drawings of cats and ducks, they still can't analyze PDF files, the lifeblood of the modern business. Who thought THAT was a great idea for businesses?
I’m saying this before the video is finished, but a big hang up I still have on Apple’s AI is wondering how they trained it. I like that it’s mostly on device so it’s not burning servers literally all the time (please drop that ChatGPT shit soon 🤞) but that doesn’t change the fact that most of these models are trained on stolen work in the first place. I’d love to see them say that their’s wasn’t, but I highly doubt it.
Tech bros 2011: 3d TVs are the next big thing! Gonna be in every home; bigger than the internet! Get on board now to avoid missing out!
Tech Bros 2016: VR is the next big thing! Gonna be in every home; bigger than the internet! Get on board now to avoid missing out!
Tech Bros 2022: NFTs are the next big thing! Gonna be in every wallet; bigger than the internet! Get on board now to avoid missing out!
Tech Bros 2024: AI...
I'm sure they've got it right this time.
I doubt AI will be quite as good as the hype train is promising, but I think that it will stick around at least as much as VR has if not a bit more.
What happened to the metaverse? I thought that was going to be the biggest next thing 😂
@@soliniv1411 3D TVs were neat, but too expensive and impractical to catch on.
VR is fun, but the chance for some people to have nausea issues keeps it as a niche hobby type thing.
NFTs have some vague theoretical uses in data science, but as a product they are worse than useless.
The metaverse was all hype and no substance.
I can't predict where we will place AI on that chart in a few years once the dust settles.
I think it has more substance than the other items on that list, but that the hype is still unrealistically high.
Don't forget cloud gaming lol
@@jonathan0berg AI as it stands is pretty useless in most fields, has scraped most of the known internet, and ultimately isn't doing much that we haven't been doing for years in similar spaces (digital assistants and spell check have been things for literal decades).
Sorry, but sky ain't exactly the limit here, chief.
ChatGPT said a game I invented was created in the 19th century, making me immortal.
I really loved this conversation. I am so sick of the AI hype bubble.
the stock markets have been more and more deregulated (starting with the Reagan and Thatcher administrations) so this is why everything is focused on constant and (impossibly) infinite growth at each quarter.
Coupled with the side effect of executives salaries that are technically capped but they can be paid in stock and insist on growth so they can collect on dividends....
This explains the "line go up at all costs" effect we see in the markets. But you see this mainly in publickly traded companies. My advice if you own a company: don't sell or include a clause in the contract that the company cannot be sold to private equity or switched to the stock market.
God, I hope so. It doesn’t affect me directly, but I feel bad for artists who were already struggling enough without AI stealing their work and their jobs.
Yup.
On top of it, I read about a company called Scopio---they would send youg artists job ads claiming to be disney, interested in their work. Just send us all your work so we can see if you're a good fit.
Essentially. Scopio makes their money by having a massive collection of art to sell that other peddlers (like Getty Images) don't have.
It's because they're scamming people out of their work.
A decade ago, I heard that 70% of google searches were for porn. Just imagine the possibilities with AI.
Just clicked on the video, super excited to see you chat with Ed Zitron
The problem with big tech: all the founders are gone (except for zuck). Companies that take on non-founder CEOs at that point no longer are innovative. They got big because their founders were innovative and focused on new ideas over cost cutting short term stuff. But the non-founder CEOs come by and trim all the legs off the machine so it can't go anywhere.
Much like 3D Tv's if we ignore AI, it will die.
NVIDIA is basically the shovel shop in a gold rush. Unfortunately the "gold" is pyrite.
A year ago, AI was going to revolutionize everything and grow exponentially. A year of "revolutionizing" and "exponential growth" later, it's substantially in the same exact state it was in a year ago. We thought we were at the foot of the mountain, but we were already almost at the peak.
Leaving my comment to help with the algorithm.
Very good conversation, I also think that AI is a bubble but there is a bit of utility on it, so it won't completé go away after the bubble pops, but it isn't as world changing as people are making it out to be.
yeah, AI isn't without it's use. i'd had good success using a particular app as a writing "partner". not like "oh write this essay for me" but i can go "hey, which of these phrasings sounds better?" and it'll point out pros and cons, ways a line could be interpreted, etc and helps me keep my tone consistent. but like...there's no reason for facebook to have AI integration. bing doesn't need it. image generation is nothing but harmful.
@@kerricaine how exactly is image generation harmful if it's trained on a licensed dataset like with Adobe?
@@Hosea405 because it takes huge amounts of electricity to process even a single image compared to most typical computing processes, and it's still being used by marketing chuds to put people out of work
@@kerricaine doesn't take that much power at all to process a single image, especially if done locally on Stable Diffusion. As far as people out of work, doubtful... only thing it'd be good for is something like generic stock, or a photo that would be impossible to take in real life. Good artists will still have work. Artists said the same thing when Photoshop launched,.
The second companies stop subsidising the user of AI it'll be so much more expensive than other means of production the industry will fall apart.
They're banking on being able to kill off all creative human job positions and aspirations, before they run out of investor capital injections - so they can then use their power to control the market and slowly jack up the prices - just like Uber have done with taxis and Amazon did with bookstores.
Friend of a Friend thing, so take with a grain of salt. But 3 friends over is someone who works on the Apple VR thing. Even he doesn't know why they made it. Weirdly too, they didn't give him a unit once the project was done either. You'd think they would have handed them out to the whole team to get them to evangelize it to their friends. Much like the Hololense there is one use case that makes some sense, AR, but the tech is nowhere near ready for a commercial or even janky enterprise product.