Pop Culture
49:48 Share

Welcome to decision's decisions. The podcast for boundary or push and conversations get handed. Join your favorite host, me. We see WTF and me mad as we dive deep into the world of non traditional relationships and exploit the autou topics surrounding dating sex. Every monday in wednesay, we vote invite to unlearn the outdated narratives dictated by traditional patriotic norms tunin enjoying the conversation, listen to decisions, decisions on the black effect podcast network, I heart radio, APP apple pocket. Wherever you get your podcast.

looking for the .

perfect Christmas playlist will check out ninety three point nine light FM chicago s Christmas music station playing your favorite holiday artists, including bin crosbie, mari, Carry for sinatra, michabo bay and Moore. Turn on the holiday light to get in a festive mood, whether at work, in traffic or at home, putting up the decorations, we can wait to celebrate the season with you. Listen to ninety three point nine light F M in chicago or anywhere in the world on the I heart radio at.

Cause media, well, I know the Oscars lds grow on trees, but I made you to trim the leaves. This is Better rough line. I'm your host that they on.

It's bullshit. We care a Better offline headquarters. The thousands of elves that work for me have been finding me things that say, bulls ship for no reason when I wrote the episode weeks ago.

I don't know what to do with them fit. Anyway, we're starting with one of the largest financial institutions in the world calling bullshit on generate AI as usual. Don't take my word for him. Check the episode notes for links that map to everything i'm talking about. I've tried my best to map them to exactly what i'm saying too.

And you can feel free to yell about IT like yellow, the words like the bcd boys, or just follow along because I think it's important for you to know where i'm getting all this from. This is just assuming that I made IT up somehow. That is an accusation i've had made.

I don't really know how I would do that anyway. The episode, very sorry. At the tail end of june, goldman sacks, one of the largest global investment banks, put out the thirty one page report titled G.

A. I. Too much spend, too little benefit. That's with a question market of the end that includes some of the most damming literature on genre, avi, that I ve ever seen. And yeah, that V P sound in the background, you hear the slow, painful deflation of the bubble i've been warning you about since march. The report covers A S productivity benefits, which golden reMarks are likely limited.

A S returns, which are likely to be significantly more limited than I anticipated, and a is power demands, which are likely so significant the utility companies will have to spend nearly forty percent more in the next three years to keep up with the demand from hyper scales and raw economists like google and microsoft. The report is so significant because goldman sacks, like any investment bank, doesn't care about your feelings or your happiness or emotions or anything unless doing so is profitable. It'll gladly hy pop anything that thinks that will make a bug.

I can make IT publish a report which claimed the AI not just generate A I was showing very positive signs of eventually boosting GDP and productivity, even though said report buried within its constant reminders that A I had yet to impact productivity growth states only five percent of companies report using generate vi in regular production. The golden to suddenly turn on the AI movement suggest that it's extremely anxious about the future of generate A I, with almost everybody agreeing on one core point in the report that the longer the general AI takes to make people money, the more money that it's gonna to make. The report also includes an interview with economist dymock of which can be found, by the way, on page four of the document, which you can find in this episode spread an institute professor who published a paper back in may called the simple macro economics of A I that argued that, and I quote the upside, the U.

S. Productivity and consequently GDP growth from generate A I will likely prove much more limited than many forecasters expect a month has only made as a ogly more pessimistic declare that truly transformative changes won't happen quickly, and few, if any, will likely occur within the next ten years. And the generate capability to affect global productivity is low because, and I quote again, many of the tasks that humans currently perform are multifaceted, require real world interaction, which A I won't be able to materially improve anytime soon.

What makes this interview and really of the whole report so remarkable is how thoroughly and aggressively attacks every bit of marketing collateral that the artificial intelligence movement has as a mogue specifically questions the belief the A N models will simply get more powerful as we throw more data and GPU graphics processing unit. The thing that branches the number's capacity, add them and specifically ask the question, what does IT mean to double a is capabilities. How does that make something like, say, a customer service read Better? Seriously though, what does Better really get them less errors? How do you quantify less errors without factory in the current state introducing them? It's not great.

And this really is a specific problem with the whole J I I fantastic bushi spill. They heavily rely on the idea that not only will these large language models, elms like ChatGPT, get more powerful, but getting more powerful will somehow granted the power to do something. As a smoggy says, what does that mean? The double let. What does that do? No, really, what does that more actually mean? We've heard people talk about this frighten months in.

The more powerful ChatGPT gets, the Better IT will get. But what does Better even look like GPT for o the latest version of ChatGPT, other than accepting more inputs, is kind of the same thing. The capabilities have not really changed.

And well, one might argue that more powerful will mean faster generative processes. There really is no parameter for what Better looks like. And perhaps that's why ChatGPT, claud and other elements have yet to take a lep beyond being able to generate pictures of gold field and long serial we do with a gun.

Anthropic cladel might be, and I quote, to best in class according to take running, but that only means that is faster and more accurate, which is cool, but not really the future or revolutionary. Al, even necessarily good. In all cases, I should add that these are questions that I and other people about A I kind I should have been asked in the whole time. Generally, if A I generates outputs based on text based inputs in request, and eventually multi model will mean we have to look at something and Jerry and answer to, and these requests can be equally specific and intro, yet the answer is always as obvious as IT sounds generated fresh, meaning that there's no actual knowledge, indeed intelligence Operating in any part of the process. As a result, it's easy to see how this gets Better ish faster, but much, much harder, if not impossible, to see how general F, A, I leads any further than where already out.

And by that, I mean, what does he do other than what it's doing today? What does ChatGPT do today? That's radically different than eighty months ago, the jump between the two generations of iphones from the first iphone to be the iphone 3 G I think I would be someone's gona email and say i'm wrong and say to them you have very rude anyway。 But that jump is huge.

Faster internet meant you could do more with the phone. There was the ap store. These were new functionalities added two years after the first iphone, two years after the first ChatGPT. 这样 啊， i guess, I guess you could talk to IT in a response that's not really gray.

How does GPT a transformer based model that generates answers, probabilistic cally, as in what the next part of the generation is most likely to be based on what's been inputs based entirely on training data? How does IT do anything more than generate paragraph of occasionally accurate text or images, maybe a picture of schools doing laundry? I don't know.

But how did any of these models even differentiate from each other when most of them were trained on the same training data that they're already running out of? And what's crazy is I mentioned the eyes should have also this. I really should have every single episode mentioned the eye.

What does Better even look like? What does more powerful and even look like? Because when we say Better, does that mean faster? Does that mean 呢？ Quick to generate things mean you can generate more things.

And the answer is usually not. It's not that I can actually function in a different way. It's just that you can do more, you can grow more.

Hey, remember the idea of growth or cost eases. But seriously though, this feels like the question to ask sam altman, al miron, the co of OpenAI. Like, what's next? What's next? Or you can generate videos.

When can I use that? cool. What does Better a look like that other than not horrible looking? okay.

But what? What, what more can GPT do is this thing? This thing can't think it's generating stuff IT doesn't know anything. IT has training data. IT eats up and then crap out an answer. How is that going to lead to even automation? Hi, just and then how do you even deal what the fact that IT really his running out of training data, and i've argued before, the training data crisis is one that does not get enough attention, but it's sufficiently die now that that has the potential to halt or at least dramatically slow any AI development in the future.

By which I mean general of A I as one paper published in the journal of computer vision and pattern recognition found in order to achieve a the improvement in model improvements, you even expands tiary large amount of data. So it's not like they read something and then work something out from there. They need so much more to learn one thing kind of right? H it's not very good.

And maybe is another way to put in each additional step of training becomes increasingly and exponentially more expensive. To take to IT requires you to get a bunder training data and process more of IT. But IT also requires extremely expensive technology GPU and a ton of energy to actually do so. And this inferred a steep financial cost, not mely just obtaining the data, but also, as I mentioned, the computer required to process IT and anthropic CEO.

Dario amo del said that the A I model les currently in development will cost as much as a billion dollars to drain, and within the next three years we may see models that cost ten or a hundred billion dollars to train, which is just insane as that's roughly three times the GDP of estonia, a beautiful, quite cold country. The facts of mister darras, a mobile who danced the elms, can even become super intelligent and that even his most conservative estimates of productivity gains, and I quote, may turn out to be too large if AI models prove less successful in improving upon more complex tasks. And I think that's really the root of the problem.

All of this excited, every second of breathless beat off hype has been built on this idea that the artificial intelligence industry LED by generate avi, will somehow revolutionizing, automate everything from robotic to the supply chain, despite the fact that generate F A, I is not actually going to solve these problems because IT is not built to do so. I won't a calculi drive my fucker and car. Jesus Christ.

And while as a mogue may have some positive things to say, for example, that AI models could be trained to help scientists conceive of test new materials, which actually already happened, thanks to google deep mind researchers. His general verdict is kind of harsh, that using general A V I N quote too much automation too soon could create bottle x and other problems of firms that no longer have the flexibility in troubleshooting capabilities that human capital provides, in essence, replacing humans with A I might break everything if you're one of those bosses that doesn't actually know what the fact is they're talking about. But you know what? I'm sure the following advertisements from people who know exactly what they talking about.

This is an antonia section with you hear, because of our job. B, we need to be connected, tiny.

over connects by more than us, right?

Gen, but it's important, like connection with your friends, your family, every connections are your life. And that gives you something that provide high speed internet, just in the context of this show is so important, will be talking about something here.

And it's easy to forget when you have a big parts of the country still don't have high speed internet.

So A T N T knows this. There are ware of IT at A T, and they are making a huge effort to change that. What they're doing on trying to cover thirty million plus locations currently who don't have IT to get IT with fiber by the end of twenty twenty five.

Millions of people obviously will benefit families, businesses, schools, spitals. There is a place called odom counting in kentucky, where they're now providing high speed internet to more than twenty thousand customers. A T N T, join a lot of great things.

Connecting changes everything. A T N T, even if you .

think it's a bit overhyped, A I is suddenly everywhere, from cell driving cars to molecular medicine to business efficiency. If it's not in your industry yet, IT is coming fast. But A I needs a lot of speed and computing power.

So how do you compete without cost spiralling out of control time to upgrade to the next generation of the cloud, oracle cloud infrastructure, or O C I. O C I is a blazing fast and secure platform for your infrastructure, database and application development, plus all your A I and machine learning workloads. O C I costs fifty percent less for compute and eighty percent less for networking.

So you're saving a pile of money. Thousands of businesses have already upgraded to O, C, I, including M, G, M resorts, specialized bikes and fireworks. Say, right now, oracle is offering to cut your current cloud bill in half if you move to O, C, I for new U S, customers with minimum financial cost. December 3， see if your company qualifies for special offer. Slash rac, that's oral .

dot com slash rac. The present very much on kids need both personal plus.

Function on these play bosses in the mud, he father, that the boast corporation and we're back. The report also includes a palot kenser for the quirked up A I hype fines. And you'll find out on page six where golden sex is, Joseph bricks argues that generate F, A, I will, and I quote, likely lead to significant economic upsides based.

And I shit you not entirely on the idea that A, I would replace workers in some jobs and then allow them to get jobs in other fields. Brags also argues that. And I quote again, the full A I automation of A I exposed tasks that are likely to occur over a longer horizon could generate significant cost savings, which assumes that generate A A I or A I itself will actually replaced these task.

But also that such a funny thing to say, oh, you know, all amain tasks over a long time could save money, fucked in alaska, drop golbin sex, any? Anyway, sorry. I should also add that unlike every single other interview in the report, bricks continually mixes up A I and generate A I, and at one point suggests that recent generate AI advance are for shadow wing, the emergence of our super intelligence.

This is like suggesting that cause I freezed my washing up liquid and it's prayed across the wall that i'm one step closed to become a picasso. I included this part of the report because sometimes, very rarely, I gets somebody suggesting i'm not considering both sides. The reason I don't generally include both sides of this argument is that the A I hype side generally makes arguments based on the assumption that things will happen, such as transformer of all the probabilistic ally generates the next part of a sentence or a picture, somehow gaining sentence.

I wish my calculate a game sentence. During high school, I was a very lonely child, did not have many friends, but not all that would do is just let me, and to make IT look like its boobs, then I get trouble. But anyway, france was surely A I researcher at google recently argued the large language models like ChatGPT can't lead to average general intelligence. The sentient A I that everyone's excited about explaining in detail in an interview with a podcast i've linked in there, the models like GPT are simply not capable of the kind of reasoning and theories zing that makes a human brain work.

Chali also argues that even models specifically built to complete the tasks of his abstraction and reasoning corpus, a benchmark test for I skills and true intelligence that he invented, are only doing so because they've been feed millions of data points of people solving the test, which is kind of like measuring somebody's IQ based on them, studying really hard to complete an I Q test. Accept, even duma. But the reason i'm suddenly bringing up the super intelligence is, or A G I are official general intelligence, average general intelligence.

Lots of people say different things. The reason saying that is because throughout every single defensive generate AI is a really nasty delivery attempt to get around the problem. The generate A A I doesn't actually automate many tasks well. It's good at generating answers, sometimes even correct ones, or creating things based on a request, sometimes with the right number of fingers.

There's no real interaction with the task or the person giving the task or considering what the task needs at all, just the subtraction of things said to output, generate, albeit in quite a complex way, tasks like taking someone's order and relying IT to the kitchen at a fast food restaurant might see elementary to most people. And I went right easy, because working in food is a hard and horrible job. He might seem elementary though, but IT isn't for an A I model that generates answers that really understanding the meaning of any, the words, and really I shouldn't t the word, really, that IT doesn't understand anything.

Last year, windy, a burgers chain here in america, announced IT would integrate its generative fresh AI ordering system into some restaurants. In late june, IT revealed that the system requires human intervention on fourteen percent of the orders. On one redit ors post, they noted that when these A I regularly required three attempts to get IT to understand them, and would sometimes cut you off if you weren't speaking fast enough.

I is two A M, I I connecting the world's chicken sandwich. The diet codes, difficult for me now, the burger, a place here in america. White castle, which implement IT, a similar system in partnership with samsung g and soundtrack, fed a little Better with a remarkable ten percent of voters requiring human intervention last month, card aldis.

This continue its own A I ordering system, which had built with IBM in the point to more than hundred restaurants, likely because IT wasn't very good with one customer, ran up for literally hundreds of chicken nugas like the, I think you should leave sketch. However, to be clear, mcDonald system wasn't you based on generate A I and if nothing else, all of these examples illustrate this disconnect between those building A I systems and how much of, really how little they understand the jobs they wish to eliminate. Little humility or doing a real job goes a long way.

Another thing to know is that on top of genera A, I generally coking up these orders when he still requires human beings to make the fuck and food. Despite all of this hype, all of this media attention, all of this incredible investment, the supposed innovations don't even seem capable of replacing the jobs that all of these horny capitalism have been planning for them to do. Not that they think they should just entire being told that these future is inevitable, or indeed, here it's going to really accept what the problem is with generate A I though IT isn't good at replacing the kind of jobs that actually affect the economy, but commotio ing distinct acts of labor.

And in the process, the early career of jobs that help people build portfolio OS to advance in their industries, the freeLance is having their livelihoods replaced by bosses using generative A I aren't being replaced so much as they're being shown how little respect many bosses have for their craft or the customer they allegedly serve, copy editors and concept as provide far more valuable work than any generate A I can. Yet an economy dominated by managers you don't appreciate or participate in labor means that these jobs are constantly under assault from large language models, pumping out stuff that all looks and sounds the same. To the point of the BBC reports, the copywriters are now being paid to help to make the A I sound more human.

One of the most fundamental misunderstandings of the buses replacing these workers with generate a AI is that you are not just asking for a thing, but outsourcing the risk and responsibility for delivering them. When I hire an artist to make a logo, my expectation is that they're listen to me. They have their own flare.

They will go back and fourth with drafts until we have something I like, and then they are proud of. Two, i'm paying them not just for their time, but for their years, learning their craft and the output itself. And so the ultimate burden, production, is not just my own, and that their experience means that they can adapt the circumstances that I might not thought of.

These are not things that you can train in the dataset, because they are derived from experiences inside and outside of the creative or process. Well, one can teach a generative. A, I, what a billion images look like.

A, I doesn't get hand crap. So I call that A P M. Saying that something needs to pop more IT doesn't have moved, nor can you infer them from written or visual media.

Because human emotions are extremely weird, as our moods, our bodies are are general existences with disgusting and weird, beautiful. And I realize all of this is a little flurry, but even the most media a copy ever written is, on some level, a collection of experiences. And fully replacing any creative is so very unlikely if you're doing so based on copying a million pieces of someone else, his homework.

Now, if I was looking for something creative, I get IT from one of the following advertisements. S, which I am sure aligned perfectly with the things that I am talking about, won't make me look weird, won't have people email me. No, you're gonna love the advertisements, and then you're not going to tell me about them.

Welcome to decisions, decisions, the podcast where boundaries are push and conversations get handed. Join your favorite host, me V Z W T F and me mad b as we dive deep into the world of non traditional relationships and explode the offer tw topics surrounding dating sex. That's right.

Every monday and wednesday, we vote in by two to unlearn the outdated narratives dictated by traditional patriarchal norms with a blends of humor, vulnerability and authenticity. We share our personal journeys, navigating authorities, tacklin the complexities of modern relationships and engage and thought provoking discussions that chAllenge social expectations. From ground breaking interviews with diverse guests to relatives, stories that are resonate with your experiences, decisions, decisions.

Is gonna be york goat to source for the open dialog about what is truly means to loving connect in today's world. Get ready to reshape your understanding of relationships and embrace the freedom of authentic c connection. Tune in enjoying the conversation, listen to decisions, decisions on the black effect podcast network, I heart radio APP apple podcast, or wherever you get your .

podcast looking for the perfect Christmas playlist will check out ninety three point nine light F M chicago Christmas music station, playing your favorite holiday artists, including being crosby, mari, Carry for insana ra, michabo bay and more. Turn on the holiday light to get in a festive mood, whether at work, in traffic or at home, putting up the decorations, we can wait to celebrate the season with you. Listen to ninety three point nine light F M in chicago or anywhere in the world on the eye art radio APP.

All right, we're back in the room. The most fascinating part of the golden sex report and you find IT on page ten is an interview with jim kail. o. Godman sax is head of global equity research. Cavo isn't a name you've ve ve heard, unless you are, for whatever reason, a big semiconductor head, but he's consistently been on the right side of history.

Named is one of the top semiconductor analyst by la research for years, successfully catching the downturn in fundamentals in multiple major chip firms far before others did at times mocked for, quote, be in wrong, and then turning out to be so very, very right. And jim, jim cavell, in no uncertain terms, thinks that generate A A, I, the whole bubble is whole generative monsters city, kind of flush. IT cavo believes that the combined expenditure of all parts of the generate A A IBM data centres, utilities, applications will cost the trillion dollars in the next several years alone.

And he asks one very simple question, what trillion dollar problem will A I solve? He notes that replacing low wage jobs with tremendously cost, the technology is basically the polar opposite of the prior technology transitions that he's witnessing the last thirty years. Just to be consistent with my A I international GDP rubra.

One trillion dollars is roughly half the GDP of italy or the entire danish economy, multiplied by two and a half times. Yes, denmark, an E, U. In nature member state that gave the world, among other things, as epic, lego and aqua, the artist behind nineteen and ninety seven security song of the year, barbi o one amazing country.

And apparently their GDP is just a foot now in comparison to how much we have to spend on generating pictures of, well, the fielding laundry ry, shooting at school. We do at this point, i'm just gonna be best laying one particular myth that can allow dispelled. And this was my favorite part, is that many people compare, generate A I to the early days of the internet, and he notes, is the even of its in its infancy.

The internet was a low cost technology solution that enable things like e commas to replace costly in combat solutions, and that A I technology is exceptionally expensive. And to justify those costs, the technology must be able to solve complex problems, which IT isn't designed to do. And that is a quote, by the way, I was a beast.

He also dismisses the suggestion that takes stance of expensive gets cheaper over time as revisionist history. That's a quote. And that quote, again, the tech world is too complacent in the assumption A I, possible declines substantially over time.

He specifically notes that the only reason that mos law was capable of enabling smaller, faster, cheaper chips was because competitors like A M, D forced into and other companies to compete. I think that doesn't really seem to be happening with video, which is a new strangle hold on G, P, S required to handle generate A I. And indeed, all of these other companies are really making competitive products.

They just kind of making the same thing. Google has their own video and text in image. General sodas lama, the metal ones sodas, ChatGPT is all the same.

There's no competition here. Can look a cartel, almost good day for an episode. And while there are companies making these graphics process units aimed at the A I market, especially in china, where U. S. Trade restrictions prevent local companies from buying high powered cards like the a one hundred that the n video makes for fears that there will be diverted to the chinese military, they're not doing so at the same scale as in video and cavo notes that and I quote, the market is too complacent about the certainty of cost. The clients, he also notes that costs are so high that even if they were to come down, you'd have to do so dramatically.

And that the comparison to the early days of the internet, where business is often relied on sixty four thousand dollars from sun microsystems and there was no life cloud stories like i'm some web services or they know or as yeah, those days paled in comparison to the current cost of generate A I. And that even before you include the replacement love, the power grid, SHE says, is a necessity to keep the boom going. I could probably just read you the entirety of cavell's interview because it's so I want to roll around on the floor on IT.

It's amazing and he's so good and he even attacks one of my least favorite things where he believes that when people say all people didn't really think the iphone was a big deal, the internet was a big deal, but specifically, the iphone and smart phones either think IT was going to be big and thus generate vala will be big. And he just says it's complete nonsense. He says that he SAT through hundreds of presentations in the early two thousands, many of them including road maps, that accurately fear how smart phones eventually rolled out, specifically knowing things like GPS.

So when G, P, S. Technology came down, of course, that would be in a small phone that makes perfect sense. Just to be clear, this guy is in he's a software hardware analyst.

He sits through presentations about what the future might look like all the time. And he said, there's no such road map for generate A I, and there's no killer rap either. He also notes the big tech companies now have no choice but to engage in an art official intelligence arms race given the hype, which will continue the trend of message spending.

And he believes that there are, and I quote, low odds of A I related revenue expansion because he doesn't believe that generate AI will make worker smarter, just more capable of finding Better information faster, which is not great, by the way, and that any advantages that generate A I gives can be arbitrary because the take can be used everywhere. And thus you con, as a company, really raised Prices, in fact, in bridging for what he said, that what might happen is a race to the bottom. Or maybe when things get too expensive, everyone will get more expensive too.

None of this is good, by the way, for them. But in plain english, just saying IT, i'm just gonna put IT out there. You'll be surprised by what I am about to say. Generate vi isn't making any money for anybody because IT doesn't actually make companies that use IT any extra money. Efficiency is useful, but it's not company defining.

Cavo also adds that hyper scale is like google and microsoft will also gana incremental revenue from A I, not the huge returns that they perhaps counting on given their vast A I related expendable over the last two years and the ridiculous things they're been doing like putting genera I in search jees Chris sunder. And this is damming ing for many reasons, chief of which is that the biggest thing that our official intelligence is meant to do is be smart and make you smarter. Being able to access information faster might make you Better at your job, but that's sufficiency rather than allowing you to do something you you're not actually really even being enhanced.

This is one step above google, but maybe it's an internal google search. I guess if you can generate really crappy looking on, i'm just not sure where IT is. And I don't think jim is either.

He ends with one important and brutal node that the more time that passes without significant AI applications, the more chAllenging, and I quote, the AI story, will become with corporate profitability likely floating the bubble as long as IT takes for the take industry to hit a more difficult economic period come like we saw in the middle to twenty twenty two. He also had his own prediction, investor enthusiasm may begin to fade if important use cases don't start to become more in the next twelve waiting months. And I think he's being a little optimistic. C. While I won't recount the rest of the report, one theme grow up repeatedly is the idea that america's power grade is literally not ready for generate A A I.

In an view in an interview with former markets of V P of energy brian janus on page fifteen, the report details numerous nightmarish problems that the growth of genera A I is causing to the power brid such as hyper scale is like microsoft, amazon and google increasing their power demands from a few hundred megawatts in the only twenty tens to a few giggles ts by two thousand thirty, enough to power multiple american cities or the centralization of data center Operations from multiple big tech companies in northern Virginia, potentially requiring a doubling of grid capacity over the next decade. Utilities, they've not experience a period of load growth as in a significant increase in poetry in nearly twenty years, which is a problem because power infrastructure is slow to build and involves numerous owners permitting a biocon tic measures to make sure it's done properly or a toll. And the total capacity of power projects waiting to connect to the grid grew thirty percent in the last year and wait times of forty to seventy months.

Expanding the grit is no easy or quick task. A monk, socker, said that these power constraints are the biggest thing in the way of A I, which is sort of true, remarkable, from a guys so often full of shit. In essence, on top of general F, A, I, not having any killer s, not meaningfully or hopefully increasing productivity or GDP, not generating any revenue, not creating any new jobs or massively changing existing industries, IT also requires amErica to rebuild its power grid, which is a good thing, which brian janus regret ly heads, that the U.

S. Is kind of a forgotten how to do. us.

Doesn't really do big infrastructure anymore. It's not a, not a great thing. I don't know. Perhaps sam alton's energy break through could be these fucking AI companies being made to pay for new power infrastructure.

And I don't mean the morning IT, I mean tax the tax the rosses they are burn the world may can pay for IT. The reason so agonisingly picked apart this report is that if golden sands is saying this, things are very, very bad. And IT also directly attacks the specific hype tane tics of A I freaks.

The sense that generate A I will create new jobs really hasn't in last eighteen months. The sense that costs will come down, they haven't in there doesn't seem to be apart than to do so in a way that matters. And then there's this incredible demand for these products they claim exist and there really isn't.

And I don't see a path to IT even golden sex. When describing the efficiency benefits of A, I added that I was able to create an A I that updated historical data, its company models, more quickly than doing so manually with one problem, IT cost six times as much to do. I love the A I future.

I love our official intelligence. I think it's so good that it's saving so money. But I feel crazy. I feel crazy. I feel sorry.

Moving on now the the remain defense, and it's also one of the most annoying people sometimes argue that perhaps OpenAI has something we don't know about, some sort of big, sexy, secret technology that will break the bones of every hate, that will bring me to my eyes and bag the machine god for mercy. And I have a counterpoint. No, they don't.

They don't have share seriously, mya morality C, T, O of OpenAI said in early june. The models the OpenAI has in its labs are not much more advances than those that are publicly available. And that's my answer to all of this.

There is no magic trick. There is no secret thing that sam altman has. This is gonna revealed to us in the next few months that makes me IT crow or some magical tool the microsoft google pops out that makes all of this worth.

There isn't. I'm telling you that isn't. We would have seen a sign, is not even a tinkle, is not a rumor, is not a league.

There's nothing generate A A. I, as I said a few months ago, is peaking. If he hasn't already peaked, IT can't do more than it's currently doing, at least not much more. Other than may be doing IT faster with some new inputs. IT isn't getting more efficient.

So koa hytner David can't gleefully oned in a recent blog that invidious b one hundred chips, the kinds used for training and running these models, will have two point five x Better performance for only twenty five percent more cost, which doesn't mean a god damn thing because general isn't going to gain sentience or intelligence or consciousness because it's able to run faster. I don't even think it's gonna able to do more things. Generating F, A, I is not going to become A G.

I, nor will IT become the kind of our official intelligence you've seen in science fiction. Ultra sa assistance, like javis or mind man, would require a kind of consciousness that no technology currently or may ever be able to produce, which is the ability to both process and understand information, flawless ously and make decisions based on experience, which, if I haven't been clear enough, are all entirely distinct things and generate A I and A I in general, doesn't have experiences. They don't have shades of gray.

They only have shades of Brown. Pop is pop. jog. Generate A A I, at best, processing formation when IT trains on data.

But at no point does IT learn or understand, because everything is doing is based on ingesting training data and developing answers based on a mathematical sense of probability, rather than any appreciative comprehension of the material itself. l. ams.

Are entirely different pieces of technology to that of an artificial intelligence in the sense that the A. I. Bubble is helping.

And it's disGraceful that the A I industry has taken so much money and attention with such a flagrant offensive lie. The jobs market isn't going to change because of generate A A, I. Because generate A, A, I can't actually do many jobs.

And its media Operate the things that is capable of doing, which is quite so shocking. There are even people who replace real paper with IT. While IT is useful efficiency tool in specific context said efficiency is based off for technology that's extremely expensive.

And I believe at some point, A I companies like anthropic and OpenAI are gonna have to increase Prices will begin to collapse under the weight of a technology that has no past profitability. If there were some secret way that this would all get fixed, wouldn't microsoft or material google, maybe amazon who's see of the cloud platform compared generate vi to doc com bubble and fever when they have taken advantage? Ship of this big, sexy sek crem why am I hearing the open eyes is already trying to raise another multibillion around after raising indeterminate amount of the eighty billion dollar valuation in february.

Isn't open a eyes analyzed revenue three point four billion dollars. Why do they need more money? How much are they burning? Because i'm guessing that if it's more than three point four billion dollars, it's got to a lot more if they still try to raise capital and if they're still going out a line about what ChatGPT could do in the future.

But i'll give you an educate, I guess, because whatever they OpenAI and other general AI hogg's have two days, obviously painfully, not the future. Generate aai is not the future, but a reggina of the past, a useful, yet not ground breaking, way to quickly generate new data from all the cost far too much, to make the computer and the energy demands worth IT. Google grew its emissions by forty eight percent in the last five years, chasing a technology that made its search engine even worse.

And IT already was, and i've got nothing to show for IT other than a bunch of very funny headlines. It's genuinely remarkable how many people have been won over about this insane con, this unscrupulous manipulation of the capital markets, the media and brainless executives disconnected from production. They're all buying there, and it's all thanks to a take industry that disconnected itself from building useful things. I've been asked a few times when I think the bubble will burst, by the way, and I maintain that part of the collapse will be invested punishing one of the major providers, microsoft or google, for a massive investment in an industry that produces little natural revenue. However, I think we'll really get the bubble pop in.

Will be a succession of bad events, like a figura pausing its new A I feature after IT immediately pleasure ize that s weather likely because he was trained on IT as part of a training data Christo by one large nasty ones, such as a major AI company like chat box company character eye, which raised one hundred fifty million dollars in funding a few years, going the information republication claims might sell to one of the bigger tech companies. I think someone like Carried to A I could collapse under the way of non sustainable business model and unprofitable technology. And just also, do you really need a chat? This is what this company is.

You can speak to legs. Tora, go. Jo from G G. Suki m, which is an an and manger by the wife in the markets, take IT too long and it's weird.

It's one of these weird companies can talk to historical figures. You can like OK, I assume you have a vae enterprise product. Oh no, it's all just non IT. It's just when I see these companies want to raise them just like what the what you we ve been doing all that.

But IT could be a slightly more useful sounding one, like cognition AI, which raised one hundred and six seventy five million dollars, two billion dollar valuation in April to make an A I soft engine with one problem. Well, I mean, it's just a small one. IT was that they had to fake a deo of IT working.

They had to fake the demo. And i've included the link in the notes to a youtube of a really great software engineer sitting. I'm actually a fan of A I, but this is crap. He just goes through the line by line. Basically, there's gonna be a moment that spoke the venture capital firms and it's going to smoke them in to pushing one of their stops to sell, which will lead to a sun and unexpected, yet very obvious collapse of a major player. Evaluations drop, and people get this from for open.

And on anthropic, there really is no path profitability, only one that includes burning further billions of dollars in the hope that they discover something, anything that might be truly in innovative or indicated for the future rather than further reiteration of generate AI, which is at best and extremely expensive. New way to process data. I just see no situation where OpenAI nand thropps continued to iterate on large language models in perpetuity is at some point microsoft, amazon, google will decide that cloud compute welfare isn't a business model.

And by the way, all of those companies are listed micros. Amazon and google have either invested in open eye or anthropic. In the case of amazon, google theyve only invested anthropic.

And a lot of that is in cloud credits, meaning the it's just welfare is arias very right wing people in the tech industry. I think that this kind of welfare bother them. I know what the differences anyway without a real, tangible breakthrough one that require them to leave the world of large language models entirely.

It's unclear how generate AI companies can even survive. Generate a AI is locked in the red queen race, burning money to make money and attempt to prove that they one day will make more money, despite there being no clear path to making more money than they spend. So a little while, a little bit upsetting and a little bit crazy making, and I feel a little bit crazy every time I put together one of these episodes, because it's also painting ridiculous.

Generate vi is unprofitable, unsustainable and fundamentally limited in what I can do, thanks to the fact that is probabilistic ally generating an answer. It's not learning anything. IT doesn't know anything.

IT isn't intelligent. It's only artificial. It's been eighteen months since this bubble started inflation.

And since then, very little has actually happened, involving technology doing new stuff, just iterative explorations of the very clear limits of what N A I model that generates answers based on training data can produce, with the answer being something that at times is sort of good. It's obvious. It's well documented.

General A A I costs far too much. IT isn't getting cheaper. IT uses too much power and doesn't do enough to justify its existence. There are no killer apps and no killer apps on the horizon, and there are no answers to these problems.

I don't know why more people aren't saying this as loudly as I am, and I don't know why everyone's not freaking out. I understand that big tech desperately needs this to be the next hypo growth market. He have got any others pursuing this cause I useful.

Environmental disaster causing cloud efficiency bone doggo will send shock waves through the industry when IT all collapses. And what's frustrating to me about this is it's becoming office. You seeing people lightly kind of come out their shells.

And so not sure this is gonna be good and not sure we hate P. K. I is a great piece in the new york times by someone recently saying that, K, I used the bunches of my links.

Thank you earlier. Anyway, put inside a bit ness, the problem here is all of the money and power going into this, but also the big lie being perpetrated by people. In my next episode, i'm going to go into some more of that.

I'm going to talk about the fact a lot of these companies are not really doing anything that their kind of doing marketing exercises is a means of manipulating the media in the markets. It's all just a disGrace for waste, the people getting hurt by generate A I people that are working contract jobs because it's harder than never to get a full time job. They're being replaced by shitty software that cost too much money that will invariably go away when this all collapses.

And I feel that the media has some responsibility here, the markets too, but definitely the media. I understand that it's difficult to look at the tech industry and think and they all kind of leading us on. But i've done IT so many times and this is the worst one i've seen.

This is ridiculous. Billions of dollars wasted, so much time, so much energy. And yes, we can cover this.

Yes, we can talk about this. We don't owe them both sides. Why do we owe them that? They wouldn't give us one.

Have you ever seen a tech entrepreneur that was a nice, fair interview? Or do they usually pison moon the moment anyone says anything negative? no.

The right way to look at this now is with suspicion, and Frankly, most innovations, but in particular this one, because despite an alarming amount, people in the media saying things like, oh yeah, well, generate, I will of course, lead to super intelligence. It's never going to do that. And we all, I don't know if you consider me a reporter or journalist.

This is a conversation to have with me by my email, I guess. But whatever I am, I should not be the only one that's really going at this hard. And I know i'm not.

There are other people, some people, right? Bright merchant on the machine. fantastic.

Ya know, a great guy. Paris Marks, not a great guy. Don't great work. An alan hugh with bloom bucker on a few episodes o great coverage and but nevertheless IT is time to start saying when we cover generate A I specifically OpenAI and anthropic, hey, where are we with super intelligence? And I don't mean what level wear out.

I mean what is the technology that you have that is going to lead to that? And when they refused to give an answer, what should be covered is that they don't have one. Make them dance.

Why do we have to do the work to make A I important? Why doesn't generate A I impress us? Right now? It's taking money out of people's hands.

It's killing free Lances. It's really horrible. And the thing to cover is that you can cover fun apps. I don't have a problem with that.

But when IT comes to google, microsoft, amazon, OpenAI, anthropic, all these companies, we need to treat them as kind of interlopers at this point because they need and show us what they're working on, then they need. I don't care about the safety side. They don't care about IT either.

We don't need to worry about the safety issue right now because the only safety issue is the problem we are facing today, which is that millions of people's works being stolen, and millions of other people are having their jobs taken in the most media away by city buses who don't do any work. When sam mult man gets on stage and claims the A I will solve all the physics. The appropriate answer is what a fucker you talking about.

You sound luck in idio when a bread light cap C O O of open a eyes says, I D I done, trained on youtube, you should say, why don't you know? And then when he bubbles the way as all he just, i'm not sure, say, is really worth that. The company with eighty billion dollars as A C L O, who does not know what's happening, saying goes from merali when SHE can't answer whether the train, new trip, these people must be treated, if not as criminals, as a special types, as con artists, as people that have absorbed so much attention, so much money, so much time, so much adulation, they should be held accountable.

And holding them accountable starts with a real evaluation of generate A I. And as i've shown you today, this bug is full of shit. Now when I don't want a happy another, I looked IT up yesterday.

And as of speaking this out on you been four and a half months of the show. I thought I had been six months or a year. I have time dillaway issues in my brain, and I will be seeing a doctor, but I genuinely want to thank all of you.

I want to thank everyone who's been there from last episode or the first episode, everyone whose come through through different ways. This shows evolving and IT will continue to evolve and angry and i'm py, but I think at this time in history is a reason to be i'm not feeling myself. I get it's naturally there.

It's naturally there. When I look at an industry of taking industry that I genuinely love, that genuinely made me the person I am today, and I see what's being done to IT and IT makes me so pissed off because technology is everything we do. To discard tech as something that happens just on the computer and then there's real life is kind of silly.

They're both the same thing now. We are all influenced by offline and online culture, at times far more by online culture. As the show progressed is i'm onna get more into that and the upcoming episodes I have going to a lot of fun.

The second epo of this week is just a real laugh, and he kind of show me how the show can grow into a more conversational fun. Interesting thing, right? Talk to all sorts of focus about all sorts things in taking out affects them. I'm really grateful for you're all listening.

You'll my email after that bother its easy in letter a, letter a or ed from a british listener, email easy about offline 点 com message me on twitter or facebook instagram are always happy tear from people I want to try to make this bear but I know one ways just keep doing in and i'm really excited for the future。 I'm very, very thankful to have all you there, even the people who are very man at me, not like in Jerry if A I but serious ly down. One final thing, people will see this stuff.

They do contact me fairly regularly and they say what I said this at the end of the shader. The premise is, well, i'll say here as well, you can't do a tome. What you can do is, say, people's name's memorable and proper agi.

I know IT sounds silly, but a lot of these ultra rich people, they aren't care about experiences. They not that really have predators. What they have is their reputations, but also they have is their lives.

I know that sounds a bit dramatic to say their lives, but look at what sam wen saying. The only recent sammon has been able to get where he is today is a series of lies and kens. The way to break, generate A I, which IT should be broken until IT improve itself, profitable and not environmentally destructive, is to push bank to constantly talk about.

How about is to share your bad experiences with that? Share every hallucination, share them publicly on social media channels, taxi women at summer on twitter, S A M A. Do IT.

I know this sounds silly, but guess what? These people have grown so rich and powerful from nobody calling out this year. And there are people that made their jealous.

You ve done. And i'm not saying nobody does IT. I'll get mad at me.

I M saying is thousands, tens of thousands, hundreds of thousands of people listen. Why they start sucks. Why it's bad will get to the tech people.

And eventually one of those messages will get to some of these financial people. And when they really turn against this, IT will collapse. And if IT doesn't, if i'm wrong, well, i'll do an episode.

How wrong I am. Not what this is meant to be informative, is meant to be entertainment. But also, I don't mind being wrong, just kind of you had to do so. Someone could be mad d of that anyway. Thank you.

Thank you for listening to Better offline. The editor um composer of the Better offline theme song is matter sales pe. You can check out more of his music and audio projects a meta salty 点 com M A T T O S O W S K I dot com， you can, you email me, are easy at Better or off line dot com, or visit Better off line docs to find more podcast links.

And of course, my newsletter, I also really recommend you go to chat, that was your ed or ant, to visit the discord and go to ask slash betroth line to check out our ad IT. Thank you so much for listening. Better off line is a production of cool zone media. For more from cal zone media, visit our website, calls on we get a dot com or check us out on the I her radio up apple podcast or whatever you get your podcast.

Welcome to decisions, decisions, the podcast or boundary to push and conversations gig handed join your favorite house, me, V Z W T F and me as we dive deep into the world of non traditional relationships and exploit the oft w topics surrounding dating sex, every monday and wednesday, we both invite you to unlearn the outdated narratives dictated by traditional patriotic norms, tune and enjoying the conversation. Listen to decisions, decisions on the black effect podcast network I have radio APP, apple pocket or wherever you get .

your podcast looking for the .

perfect Christmas playlist will check out ninety three point nine light FM chicago s Christmas music station playing your favorite holiday artists, including being crosbie mori, Carry for sinatra, michabo bay and Moore. Turn on the holiday light to get in a festive mood, whether at work, in traffic or at home putting up the decorations. We can't wait to celebrate the season with you. Listen to ninety three point nine light F M in chicago or anywhere in the world on the I hard radio.