I'll lay out my bias up front here, I'm the guy in the second panel on this one - so definitely influenced by how goddamn cool I think the tech is, but I was also working with it a good decade...
Exemplary
I'll lay out my bias up front here, I'm the guy in the second panel on this one - so definitely influenced by how goddamn cool I think the tech is, but I was also working with it a good decade before anyone (including me) saw it as much more than an obscure academic tool and I'm largely disdainful of the hype cycle here.
If the main point of the article is intended to be that "AI" is overhyped, companies are throwing giant buckets of money at it with no clear plan, and the media are fuelling the whole thing with nonspecific and emotive reporting, then yeah, I'd say pretty much the same. But the author is flat out wrong on enough specifics along the way that it undermines the credibility of the whole, and incoherent enough on other points that I'm genuinely not sure whether that's even what he's intending to say.
A few key snippets for now:
...one problem that has poisoned reporting on generative AI is an unwillingness to clearly describe how these things work, instead referring to its actions as some sort of magical process done on a big, scary computer.
Agreed on this in isolation, but it seems to form a broader undercurrent to the rest of the piece that's conflating "I don't know" or "reporters don't explain" with "nobody knows". You can go read the research papers, the LLaMA or Mistral or Stable Diffusion source code, the mathematical underpinnings of the whole field - most people can't or don't want to do that, and frankly I don't blame them, but the information is there. The big players have a definite edge, but we've seen open source catching up within a few months of most major breakthroughs, leading the way on some others, and the theoretical academic papers are often published from a university lab somewhere before the techniques make it into commercial models anyway.
Science reporting has always sucked, this field isn't unique in that. If anything it's more open than the average - the vast majority of what's important goes through arXiv and GitHub rather than commercial journals and closed labs.
...OpenAI's models continually prove themselves unable to match even the dumbest human beings alive
Seriously? Text model output is far from perfect, but we're already way past Turing test territory here. AI detection is a hot topic because people can no longer be certain whether they're seeing human or machine output. And that's just text - how many average humans can create images comparable to DALL-E or 3D animation comparable to Sora? And that's just considering models from a single largely closed source organisation.
There are sometimes telltale signs of machine generated content because the failure cases are often distinct from the way a human would fail, but that doesn't make even the bad outcomes inherently worse than human: the fact we're even comparing the models to the best of humanity's work, to the famous artists and authors rather than to the mediocrity that most of us would manage when working outside our own fields, is a sign that it's already well above the average human in many verticals.
These models do not "know" anything. They are mathematical behemoths generating a best guess based on training data and labeling, and thus do not "know" what you are asking it to do. You simply cannot fix them. Hallucinations are not going away.
Large language models are by far the best data parsers we've ever come up with. You need factual accuracy, you point it at the database and tell it to sift through and link back to the primary information when it's got an answer - exactly as you would with a human. We don't expect a person, even a subject matter expert, to have every fact and figure memorised; we expect them to filter information and cite sources.
While Joanna Stern may have said that Sora's generative video clips "freaked her out," much of what makes them scary is the assumption that OpenAI will fix hallucinations, something that the company has categorically failed to do
This is just a weird statement - hallucinations are an issue for data accuracy, that's important if you're doing something like using a text model to compile statistics, but conflating them with creative output and then calling it a blocker doesn't make sense. Images and videos don't need perfection, and when glitches are visible or distracting you'll by definition notice that and hit regenerate.
If you stop saying things like "AI could do" or "AI will do," you have to start asking what AI can do, and the answer is...not that much, and not much more in the future.
I can't draw, I'm a shit photographer, the last time I did any 3D modelling was in the late 2000s, the list goes on. Those are all things I can now get a model to do for me, in high quality, in the space of seconds, at a cost of pennies. AI gives me capabilities I simply didn't have before.
Is that a good thing? Does it have implications for the value of the skilled professionals who would otherwise do those things for me (if I needed them enough and had the budget to pay, neither of which is a given)? How much will that shift the commercial or economic landscape, if at all? What does it mean for understanding of truth and misinformation? Will this all shake out to be profitable for anyone? Is copyright law going to implode? Those are interesting and meaningful questions. But to ask what the systems can actually do, and then claim that generating content on demand is "not that much", just seems absurd to me.
This is just a nit, but the actual Turing Test is a game that's much harder than people pretend it is. We are not "way past Turing Test." We're not even close to building an AI player that can win...
This is just a nit, but the actual Turing Test is a game that's much harder than people pretend it is. We are not "way past Turing Test." We're not even close to building an AI player that can win against human opponents who trained for it.
In the more casual sense of generating text that looks like a human wrote it, though, it's pretty good, if you cherry-pick.
For reference, from wiki: The 3rd party evaluator who's aware there's a machine in the conversation is the part that often gets dropped, and there's still wiggle room on the knowledge of the...
This is just a nit, but the actual Turing Test is a game that's much harder than people pretend it is. We are not "way past Turing Test." We're not even close to building an AI player that can win against human opponents who trained for it.
For reference, from wiki:
... a test of a machine's ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. Turing proposed that a human evaluator would judge natural language conversations between a human and a machine designed to generate human-like responses. The evaluator would be aware that one of the two partners in conversation was a machine, and all participants would be separated from one another. The conversation would be limited to a text-only channel, such as a computer keyboard and screen, so the result would not depend on the machine's ability to render words as speech.[3] If the evaluator could not reliably tell the machine from the human, the machine would be said to have passed the test.
The 3rd party evaluator who's aware there's a machine in the conversation is the part that often gets dropped, and there's still wiggle room on the knowledge of the evaluator (are they an AI expert, a layman, someone in-between?). How long is this conversation? Is the human in the conversation supposed to be aware they're talking to an AI, and if so, are they supposed to help it pass or make it harder?
I think in many cases a layman probably can't tell the difference, but I do think that with even a passing knowledge of the field it's going to be pretty obvious still under most cases. Clearly we've come a long way from basic chat bots, and the Turing Test continues to be a useful metric for progress, but I'm pretty sure we haven't passed it by most definitions.
of course, like many factors in life, we do hold bots to much higher standards than humans on the same tasks. Self Driving cars is the most obvious factor as of late. I'm sure they are already at...
We don't expect a person, even a subject matter expert, to have every fact and figure memorised; we expect them to filter information and cite sources.
of course, like many factors in life, we do hold bots to much higher standards than humans on the same tasks. Self Driving cars is the most obvious factor as of late. I'm sure they are already at a point where they are better than the average driver (becasuse the average driver isn't super competent to begin with), but every accident will hit the news. The news has meanwhile given up reporting most small accidents for human error.
I can now get a model to do for me, in high quality, in the space of seconds, at a cost of pennies.
For now. That's my biggest fear out of everything long term. I already heard a small story in China of a game studio that kicked out all its artists for some mobile game, and then the AI studio jacked up costs to the point where it was more economically viable to hire back the artists (which makes sense. Remember we're talking about China here for artist labor, a very popular outsoursing location. So the prices probably were still pretty low from a Western standpoint).
The game isn't anything noteworthy but it was a funny microcosm of what will happen in 10,15 years. another instance of market capture. The actual labor will be cheaper, but those savings will barely pass to the developers, let alone the consumers. It will just homogenize towards entertainment mush (for many reasons, but I could double my word count describing the logistical and artistic issues here) with less talent behind it, while conveniently pricing out most indie ambitions (the people who would seem to gain the most benefit out of it). And with years of refusal to train talent the market there will stagnate outside of relatively few "old school" creatives who still take the time to learn the craft.
I would love somebody with more knowledge to chime in on this. This is a strong argument (9 months away), but I can see that there are many places where it could fall apart if certain points are...
I would love somebody with more knowledge to chime in on this. This is a strong argument (9 months away), but I can see that there are many places where it could fall apart if certain points are not accurate. Especially the potential false equivocation with the metaverse that seemed more an appeal to emotion (among the various other emotional appeals).
I have yet to finish the article, but a few things stand out: The intro very heavily does focus on models creating images and videos. Less so on LLMs. In fact, for an intro to the greater point I...
I have yet to finish the article, but a few things stand out:
The intro very heavily does focus on models creating images and videos. Less so on LLMs. In fact, for an intro to the greater point I feel it was a bit long winded as it was more or less about openAI as a company being problematic about their training data and less open than they could/should be. Which is a valid article in itself but doesn't say much about the greater point of the article.
The author at some point mentions they only know one person who uses chatGPT or ai in general. From there they draw the conclusion that it is less useful or productive than made out to be. Which might be train for certain fields of work the author is active in themselves. If you look at software development and IT in general you are very likely to run into people who find a real benefit in the use of AI powered interactive rubber ducks.
The author seems to also focus on AI only being a success if it can wholesale replace jobs. Which I think is a bit silly.
I'll finish reading the article in a bit. But, so far this really feels like an article more based on anecdotal experience to me.
Is the current generation of AI as revolutionary and game changing as some people claim it to be? I don't think so either. But, at the same time, it is already a very useful tool in the tool belt of many people so it also isn't a fad in that regard.
I agree that his approach is very anecdotal, but I think his general criticism of the hype versus reality has some fair points. The AI companies with their trillions of dollars in investment money...
I agree that his approach is very anecdotal, but I think his general criticism of the hype versus reality has some fair points. The AI companies with their trillions of dollars in investment money and massive running costs are still trying to sell the concept as something everyone should be using. AI is already giving plenty of value, but is it enough for the current valuation and running cost? I doubt programmers having a rubber duck is enough for it to be worthwhile the current amount of money being thrown around. Especially with how they are trying to push AI into everything where no-one asked for it, at least give the impression that they desperately need to drive up demand.
I think it is fair to say that there is a hype. I also think it is fair to say there is some sort of bubble that will at some point burst. However, the author does seem to approach this from an...
I think it is fair to say that there is a hype. I also think it is fair to say there is some sort of bubble that will at some point burst.
However, the author does seem to approach this from an "all or nothing" point of view I can't really agree with. That, and the article really feels like two different articles mushed together as they conflate the discussion about training data with the practical applications of these models.
There are plenty of technologies we use today that came with a lot of hype and promises that technically speaking have not been achieved. At the same time, these technologies still exists and are in use today. This simply is often what happens when something is new and people scramble to explore the possibilities. So, you get all sorts of wild investments and experimentation. Then the initial bubble burst, and you are left with those experiments that turned out to be actually viable longer term.
It's those viable implementations and usages that are interesting. But those are also the ones you'll easily miss or ignore when you view it through the earlier mentioned "all or nothing" lens.
To be clear, I get where that view comes from. If you go purely by reporting and promises of these companies, you'd have highly inflated expectations. But if you have paid any attention to any reporting about new technologies in... well any point in history you realistically can already know that these are marketing promises as with any new technology. The feeling I get with this article is that the author only started to realize this very recently.
In short, the article touches on some valid points. But overall it didn't make a cohesive argument and didn't seem to really know what it wants the subject to be.
He doesn't really say it's all our nothing. Just that unless something dramatically changes and they can somehow fix the hallucinations as they politely call it (really it's just trying to soften...
He doesn't really say it's all our nothing. Just that unless something dramatically changes and they can somehow fix the hallucinations as they politely call it (really it's just trying to soften the blow that the reality of LLM is just extreme pattern matching and has no ability to determine fact from fiction, nor tell you how it reached that conclusion ) that it will likely provide some productivity savings for companies bottom line but no where near the hype they are selling. We've seen this with pretty much every new revolutionary product that's come out. Blockchain, RPA etc
Let me preface this with the fact that I don't intend this reply to be combative. I was about to draft a reply to your remark about them not saying that. However, after I finished reading your...
Let me preface this with the fact that I don't intend this reply to be combative. I was about to draft a reply to your remark about them not saying that. However, after I finished reading your comment, I also realize that you probably did skim over what I said. Specifically your last line did stop me, as I talk about that exact thing in my comment you are replying to.
A while ago I have set myself a goal that if I do engage in conversation on the internet, I want it to be meaningful. Specifically, I don't want to engage in conversations where it ends up a back and forth between half read messages.
So I'd like to invite you to reread my comment before continuing this conversation. Or not, of course, there is no obligation to read everything random people say on the internet.
The article is one more from the "writers" who write for the online audience. Writers who spend time online, trawling through social media. Looking at trends, what gets upvoted, what gets lots of...
The article is one more from the "writers" who write for the online audience. Writers who spend time online, trawling through social media. Looking at trends, what gets upvoted, what gets lots of engagement.
One of those subjects has become AI, and a lot of people who know nothing about AI except for what they read in articles like this one, have "decided" (based on other poorly informed folks telling them to) they hate AI.
Everyone's seen scifi movies and shows, read scifi stories, where the computer is sentient. Or semi-sentient. Where the characters will say stuff like "put the (problem area) up on the screen" or "analyze and report to me the exact specific thing we're plot-talking about."
And the computer just does it. Like a person. Intuitively, intelligently. Everyone wants that. Anyone who says they don't, who says "no, computers should always remain stupid and dumb and need to be painstakingly lead by the nose only when a programmer with a decade of training is on hand" is lying, because no one wants that.
So when you say "no-one asked for it", I feel like you're just on the bandwagon of "AI BAD AI BAD GRRRR RAAAWWWWRRRR."
I hate corporations as much as the next guy. I especially hate greedy people, like those who often go into tech. People who sometimes learn to talk a nice game about discovery and all that, but who really just want the cash. They want it so they can use it to make more cash that's theirs and theirs alone.
So there are a lot of less than great people involved in tech, and in AI, and that's regrettable. But that doesn't change that AI is coming, and that it will end up taking us to these scifi places. Not in a story, in reality. In the real world. AI is coming so the computer will be usable by people like my parents, who still get confused over device remotes.
But it's not here yet. Which is not cause to write it off. Things take time. That's what development is.
I see it constantly with SpaceX, where article writers, and thus social media posts, instantly seize on the fact that a rocket didn't immediately spring into existence, leaving the surface of the Earth, heading off to Mars, in one-and-done in-and-out perfect swoop as proof positive that SpaceX suxxxx.
And we see it alot lately with AI. The fact that they can't open a GPT session, and type "write me a magnum opus on (obscure subject)" or even just "hi, I just got home, do all the things I want done right now" and have it do it means it suxxx.
Plus, of course, a lot of content mill "writers" consider themselves writers, and feel threatened by an AI that can put words together into coherent sentences. The same as a lot of content mill artists (graphical, whatever) are starting to feel threatened by an AI that can take a prompt and in an increasingly coherent fashion kick back coherent images.
Those content mill hacks have been pumping out AI BAD AI BAD stuff for months now. It gets easy engagement, easy clicks, which puts money in their pockets and makes the people they report to happier.
AI is coming. Pitching a fit about it is just juvenile. Arguably wrong headed. Nothing's perfect right out of the gate. Yes, some of the shit megacorps are starting to try it with isn't there yet, and is going to be less than great for consumers.
That doesn't mean we try to un-invent it, which is impossible anyway since scientists and data researchers worldwide know it exists, know it's possible, and are intrigued by the possibilities.
Maybe it's a better use of time, and online space, to stop dooming and glooming over AI. Is it a panacea? No.
Not yet.
But at some point, it'll do all the things. Not today, and probably not tomorrow either, but thinking it's never coming is as foolish as thinking humans can't fly because they don't have wings.
Humans found a way. And right now, humans are finding a way for the computer to become sentient enough to be useful without a hands-on programmer there live-writing the code to make it so. The AI will just make itself so.
I get where you are coming from. But I also feel like the way you put down your argument, your comment is also an over the top response. Just one on the other end of the spectrum. To be frank,...
I get where you are coming from. But I also feel like the way you put down your argument, your comment is also an over the top response. Just one on the other end of the spectrum.
To be frank, your comment reads along the lines of "NO AI AWESOME AND IT WILL BE AWESOME, JUST YOU WAIT!".
Which I also don't think is a healthy attitude. I am not sure if that was your intention when writing this, but it is the vibe I get from reading it.
The current models have a fair amount of hype surrounding them, there is a lot of mud being thrown on the wall to see what sticks, and not everything will stick.
As I said in other comments, it already has proven useful for many people. But a lot of it still depends on how these models develop further as to what sci-fi feature we will actually see.
Unlike AI company PR flacks, I'm not saying it's going to be great tomorrow. Or next quarter. And that this is why you should immediately order, buy, and subscribe to us now so we have all the...
Unlike AI company PR flacks, I'm not saying it's going to be great tomorrow. Or next quarter. And that this is why you should immediately order, buy, and subscribe to us now so we have all the moneys.
Which is what corporate PR flacks are paid to do; convince you to give their corporation money. It's no different than Boeing's PR flacks making the rounds saying "we take the (crashing planes) situation very seriously and are dedicated to ensuring our customers' safety is our number one priority." Of course the PR people are going to say stuff like that; they're being paid to say that, only that, always that. However untrue it usually is.
I take a long view because that's how technology works. And also, because I've seen, over and over and over and over how people, when something new appears, don't understand and thus flame it. Distrust it. Denounce it. Declare it verboten, useless, and a waste.
Motors, industrialization, flight, computers, the lists just are endless. Most technology was thought to be a complete waste of time, utterly stupid, pointless and fanciful, when it was in its early stages. People always do this. Very few humans can look at a proposal and think "hmm, maybe." Most just shake their heads and flame it. That's what humans do; they're bad at change, hate change, and are terrible at looking forward.
I'm not super great at looking forward, but I've learned to think. To consider. To pause and ponder, rather than just reacting. To examine possibilities. And with AI, the possibilities are amazing.
Bill Gates famously said "640K of memory should be enough for anybody." Now consumers are needing 16G of system memory just to get their OS off the ground most of the time, and a lot of "casual computer enthusiasts" will have, between system and video and CPU cache, north of 40G memory. Commercial data centers are installing individual racks that use hundreds of gigs of memory. Data science researchers are working with chip manufacturers to get production lines that'll churn out factory assembled components by the truckload that have hundreds of gigs each.
What Bill said made sense at the time. There were people who agreed with him, when he said it. And yet technology marches onward.
People hate that they're bad at it, but the fact is most people are bad at the future. At looking forward, at thinking, at analyzing possibilities and positioning to take advantage of them. But they are. Humans trash the new, always. It's usually not until the new is so obvious even the blind can see it that the tune changes, and suddenly the content mills are churning out "how did life ever exist before X?" Where, usually, a year or two prior the articles were "X should die screaming in flames."
I just want people to calm down and use their heads for more than just holding up their perfectly coifed hair as they take a selfie. AI isn't bad. There are bad people, and some of those bad people work in AI, but that's true of everything, everywhere, in the world. Bad people suck, and they're bad, but they're also just how humans are sometimes.
That doesn't make AI bad. Or mean it's not going to continue advancing. History books talk about "the rapid pace of development" as it relates to the transition into industrialization. Meanwhile, we're smack in the heart of the information age, where about a decade ago no one envisioned that just about every human on the planet would have a computer in their pockets that's more powerful than what corporations in the 70s would have paid a hundred grand or more to own.
And that everyone mostly uses those smartphones to take selfies and fire off emoticon posts that decry the crushing advancement of technology. Never noticing the irony. Which would be amusing if it just didn't get kind of tiring sometimes to hear them not noticing it as they declare the latest wave of advancements to be stupid and impossible.
I don't disagree with your message. I am just saying that you're wording earlier came across as something you might not have intended. You are still doing it to some degree. Like, I am not even...
I don't disagree with your message. I am just saying that you're wording earlier came across as something you might not have intended.
You are still doing it to some degree. Like, I am not even sure what you are even responding to in my comment. As your entire reply feels a bit much in response to what I said.
I think AI is just a calculator on steroids. It's just a tool. The tech behind it is very cool and it will allow workers to be way more productive the same way spreadsheets revolutionized...
I think AI is just a calculator on steroids. It's just a tool. The tech behind it is very cool and it will allow workers to be way more productive the same way spreadsheets revolutionized accounting and factories revolutionized the way we build things.
Right now, people are expecting it to do everything and maybe one day it will, but right now it's just a handy companion to help people use computers.
I guess the author doesn't know any programmers. Charging a subscription for better autocomplete and advice that you have to be somewhat wary of might not result in all that much revenue, but it...
I guess the author doesn't know any programmers. Charging a subscription for better autocomplete and advice that you have to be somewhat wary of might not result in all that much revenue, but it does change the day-to-day experience.
I expect efficiency improvements and competition to lower the subscription costs. The only way to maintain them would be to significantly improve the results.
That might happen. There are hints that OpenAI will release something new this year.
It's frustrating, though, because these hints are indistinguishable from hype. Similarly, there was a fair bit of hype about GPT-4 before it was released, and none of it could be substantiated.
I'll lay out my bias up front here, I'm the guy in the second panel on this one - so definitely influenced by how goddamn cool I think the tech is, but I was also working with it a good decade before anyone (including me) saw it as much more than an obscure academic tool and I'm largely disdainful of the hype cycle here.
If the main point of the article is intended to be that "AI" is overhyped, companies are throwing giant buckets of money at it with no clear plan, and the media are fuelling the whole thing with nonspecific and emotive reporting, then yeah, I'd say pretty much the same. But the author is flat out wrong on enough specifics along the way that it undermines the credibility of the whole, and incoherent enough on other points that I'm genuinely not sure whether that's even what he's intending to say.
A few key snippets for now:
Agreed on this in isolation, but it seems to form a broader undercurrent to the rest of the piece that's conflating "I don't know" or "reporters don't explain" with "nobody knows". You can go read the research papers, the LLaMA or Mistral or Stable Diffusion source code, the mathematical underpinnings of the whole field - most people can't or don't want to do that, and frankly I don't blame them, but the information is there. The big players have a definite edge, but we've seen open source catching up within a few months of most major breakthroughs, leading the way on some others, and the theoretical academic papers are often published from a university lab somewhere before the techniques make it into commercial models anyway.
Science reporting has always sucked, this field isn't unique in that. If anything it's more open than the average - the vast majority of what's important goes through arXiv and GitHub rather than commercial journals and closed labs.
Seriously? Text model output is far from perfect, but we're already way past Turing test territory here. AI detection is a hot topic because people can no longer be certain whether they're seeing human or machine output. And that's just text - how many average humans can create images comparable to DALL-E or 3D animation comparable to Sora? And that's just considering models from a single largely closed source organisation.
There are sometimes telltale signs of machine generated content because the failure cases are often distinct from the way a human would fail, but that doesn't make even the bad outcomes inherently worse than human: the fact we're even comparing the models to the best of humanity's work, to the famous artists and authors rather than to the mediocrity that most of us would manage when working outside our own fields, is a sign that it's already well above the average human in many verticals.
Large language models are by far the best data parsers we've ever come up with. You need factual accuracy, you point it at the database and tell it to sift through and link back to the primary information when it's got an answer - exactly as you would with a human. We don't expect a person, even a subject matter expert, to have every fact and figure memorised; we expect them to filter information and cite sources.
This is just a weird statement - hallucinations are an issue for data accuracy, that's important if you're doing something like using a text model to compile statistics, but conflating them with creative output and then calling it a blocker doesn't make sense. Images and videos don't need perfection, and when glitches are visible or distracting you'll by definition notice that and hit regenerate.
I can't draw, I'm a shit photographer, the last time I did any 3D modelling was in the late 2000s, the list goes on. Those are all things I can now get a model to do for me, in high quality, in the space of seconds, at a cost of pennies. AI gives me capabilities I simply didn't have before.
Is that a good thing? Does it have implications for the value of the skilled professionals who would otherwise do those things for me (if I needed them enough and had the budget to pay, neither of which is a given)? How much will that shift the commercial or economic landscape, if at all? What does it mean for understanding of truth and misinformation? Will this all shake out to be profitable for anyone? Is copyright law going to implode? Those are interesting and meaningful questions. But to ask what the systems can actually do, and then claim that generating content on demand is "not that much", just seems absurd to me.
This is just a nit, but the actual Turing Test is a game that's much harder than people pretend it is. We are not "way past Turing Test." We're not even close to building an AI player that can win against human opponents who trained for it.
In the more casual sense of generating text that looks like a human wrote it, though, it's pretty good, if you cherry-pick.
For reference, from wiki:
The 3rd party evaluator who's aware there's a machine in the conversation is the part that often gets dropped, and there's still wiggle room on the knowledge of the evaluator (are they an AI expert, a layman, someone in-between?). How long is this conversation? Is the human in the conversation supposed to be aware they're talking to an AI, and if so, are they supposed to help it pass or make it harder?
I think in many cases a layman probably can't tell the difference, but I do think that with even a passing knowledge of the field it's going to be pretty obvious still under most cases. Clearly we've come a long way from basic chat bots, and the Turing Test continues to be a useful metric for progress, but I'm pretty sure we haven't passed it by most definitions.
of course, like many factors in life, we do hold bots to much higher standards than humans on the same tasks. Self Driving cars is the most obvious factor as of late. I'm sure they are already at a point where they are better than the average driver (becasuse the average driver isn't super competent to begin with), but every accident will hit the news. The news has meanwhile given up reporting most small accidents for human error.
For now. That's my biggest fear out of everything long term. I already heard a small story in China of a game studio that kicked out all its artists for some mobile game, and then the AI studio jacked up costs to the point where it was more economically viable to hire back the artists (which makes sense. Remember we're talking about China here for artist labor, a very popular outsoursing location. So the prices probably were still pretty low from a Western standpoint).
The game isn't anything noteworthy but it was a funny microcosm of what will happen in 10,15 years. another instance of market capture. The actual labor will be cheaper, but those savings will barely pass to the developers, let alone the consumers. It will just homogenize towards entertainment mush (for many reasons, but I could double my word count describing the logistical and artistic issues here) with less talent behind it, while conveniently pricing out most indie ambitions (the people who would seem to gain the most benefit out of it). And with years of refusal to train talent the market there will stagnate outside of relatively few "old school" creatives who still take the time to learn the craft.
I would love somebody with more knowledge to chime in on this. This is a strong argument (9 months away), but I can see that there are many places where it could fall apart if certain points are not accurate. Especially the potential false equivocation with the metaverse that seemed more an appeal to emotion (among the various other emotional appeals).
I have yet to finish the article, but a few things stand out:
I'll finish reading the article in a bit. But, so far this really feels like an article more based on anecdotal experience to me.
Is the current generation of AI as revolutionary and game changing as some people claim it to be? I don't think so either. But, at the same time, it is already a very useful tool in the tool belt of many people so it also isn't a fad in that regard.
I agree that his approach is very anecdotal, but I think his general criticism of the hype versus reality has some fair points. The AI companies with their trillions of dollars in investment money and massive running costs are still trying to sell the concept as something everyone should be using. AI is already giving plenty of value, but is it enough for the current valuation and running cost? I doubt programmers having a rubber duck is enough for it to be worthwhile the current amount of money being thrown around. Especially with how they are trying to push AI into everything where no-one asked for it, at least give the impression that they desperately need to drive up demand.
I think it is fair to say that there is a hype. I also think it is fair to say there is some sort of bubble that will at some point burst.
However, the author does seem to approach this from an "all or nothing" point of view I can't really agree with. That, and the article really feels like two different articles mushed together as they conflate the discussion about training data with the practical applications of these models.
There are plenty of technologies we use today that came with a lot of hype and promises that technically speaking have not been achieved. At the same time, these technologies still exists and are in use today. This simply is often what happens when something is new and people scramble to explore the possibilities. So, you get all sorts of wild investments and experimentation. Then the initial bubble burst, and you are left with those experiments that turned out to be actually viable longer term.
It's those viable implementations and usages that are interesting. But those are also the ones you'll easily miss or ignore when you view it through the earlier mentioned "all or nothing" lens.
To be clear, I get where that view comes from. If you go purely by reporting and promises of these companies, you'd have highly inflated expectations. But if you have paid any attention to any reporting about new technologies in... well any point in history you realistically can already know that these are marketing promises as with any new technology. The feeling I get with this article is that the author only started to realize this very recently.
In short, the article touches on some valid points. But overall it didn't make a cohesive argument and didn't seem to really know what it wants the subject to be.
He doesn't really say it's all our nothing. Just that unless something dramatically changes and they can somehow fix the hallucinations as they politely call it (really it's just trying to soften the blow that the reality of LLM is just extreme pattern matching and has no ability to determine fact from fiction, nor tell you how it reached that conclusion ) that it will likely provide some productivity savings for companies bottom line but no where near the hype they are selling. We've seen this with pretty much every new revolutionary product that's come out. Blockchain, RPA etc
Let me preface this with the fact that I don't intend this reply to be combative. I was about to draft a reply to your remark about them not saying that. However, after I finished reading your comment, I also realize that you probably did skim over what I said. Specifically your last line did stop me, as I talk about that exact thing in my comment you are replying to.
A while ago I have set myself a goal that if I do engage in conversation on the internet, I want it to be meaningful. Specifically, I don't want to engage in conversations where it ends up a back and forth between half read messages.
So I'd like to invite you to reread my comment before continuing this conversation. Or not, of course, there is no obligation to read everything random people say on the internet.
The article is one more from the "writers" who write for the online audience. Writers who spend time online, trawling through social media. Looking at trends, what gets upvoted, what gets lots of engagement.
One of those subjects has become AI, and a lot of people who know nothing about AI except for what they read in articles like this one, have "decided" (based on other poorly informed folks telling them to) they hate AI.
Everyone's seen scifi movies and shows, read scifi stories, where the computer is sentient. Or semi-sentient. Where the characters will say stuff like "put the (problem area) up on the screen" or "analyze and report to me the exact specific thing we're plot-talking about."
And the computer just does it. Like a person. Intuitively, intelligently. Everyone wants that. Anyone who says they don't, who says "no, computers should always remain stupid and dumb and need to be painstakingly lead by the nose only when a programmer with a decade of training is on hand" is lying, because no one wants that.
So when you say "no-one asked for it", I feel like you're just on the bandwagon of "AI BAD AI BAD GRRRR RAAAWWWWRRRR."
I hate corporations as much as the next guy. I especially hate greedy people, like those who often go into tech. People who sometimes learn to talk a nice game about discovery and all that, but who really just want the cash. They want it so they can use it to make more cash that's theirs and theirs alone.
So there are a lot of less than great people involved in tech, and in AI, and that's regrettable. But that doesn't change that AI is coming, and that it will end up taking us to these scifi places. Not in a story, in reality. In the real world. AI is coming so the computer will be usable by people like my parents, who still get confused over device remotes.
But it's not here yet. Which is not cause to write it off. Things take time. That's what development is.
I see it constantly with SpaceX, where article writers, and thus social media posts, instantly seize on the fact that a rocket didn't immediately spring into existence, leaving the surface of the Earth, heading off to Mars, in one-and-done in-and-out perfect swoop as proof positive that SpaceX suxxxx.
And we see it alot lately with AI. The fact that they can't open a GPT session, and type "write me a magnum opus on (obscure subject)" or even just "hi, I just got home, do all the things I want done right now" and have it do it means it suxxx.
Plus, of course, a lot of content mill "writers" consider themselves writers, and feel threatened by an AI that can put words together into coherent sentences. The same as a lot of content mill artists (graphical, whatever) are starting to feel threatened by an AI that can take a prompt and in an increasingly coherent fashion kick back coherent images.
Those content mill hacks have been pumping out AI BAD AI BAD stuff for months now. It gets easy engagement, easy clicks, which puts money in their pockets and makes the people they report to happier.
AI is coming. Pitching a fit about it is just juvenile. Arguably wrong headed. Nothing's perfect right out of the gate. Yes, some of the shit megacorps are starting to try it with isn't there yet, and is going to be less than great for consumers.
That doesn't mean we try to un-invent it, which is impossible anyway since scientists and data researchers worldwide know it exists, know it's possible, and are intrigued by the possibilities.
Maybe it's a better use of time, and online space, to stop dooming and glooming over AI. Is it a panacea? No.
Not yet.
But at some point, it'll do all the things. Not today, and probably not tomorrow either, but thinking it's never coming is as foolish as thinking humans can't fly because they don't have wings.
Humans found a way. And right now, humans are finding a way for the computer to become sentient enough to be useful without a hands-on programmer there live-writing the code to make it so. The AI will just make itself so.
I get where you are coming from. But I also feel like the way you put down your argument, your comment is also an over the top response. Just one on the other end of the spectrum.
To be frank, your comment reads along the lines of "NO AI AWESOME AND IT WILL BE AWESOME, JUST YOU WAIT!".
Which I also don't think is a healthy attitude. I am not sure if that was your intention when writing this, but it is the vibe I get from reading it.
The current models have a fair amount of hype surrounding them, there is a lot of mud being thrown on the wall to see what sticks, and not everything will stick.
As I said in other comments, it already has proven useful for many people. But a lot of it still depends on how these models develop further as to what sci-fi feature we will actually see.
Unlike AI company PR flacks, I'm not saying it's going to be great tomorrow. Or next quarter. And that this is why you should immediately order, buy, and subscribe to us now so we have all the moneys.
Which is what corporate PR flacks are paid to do; convince you to give their corporation money. It's no different than Boeing's PR flacks making the rounds saying "we take the (crashing planes) situation very seriously and are dedicated to ensuring our customers' safety is our number one priority." Of course the PR people are going to say stuff like that; they're being paid to say that, only that, always that. However untrue it usually is.
I take a long view because that's how technology works. And also, because I've seen, over and over and over and over how people, when something new appears, don't understand and thus flame it. Distrust it. Denounce it. Declare it verboten, useless, and a waste.
Motors, industrialization, flight, computers, the lists just are endless. Most technology was thought to be a complete waste of time, utterly stupid, pointless and fanciful, when it was in its early stages. People always do this. Very few humans can look at a proposal and think "hmm, maybe." Most just shake their heads and flame it. That's what humans do; they're bad at change, hate change, and are terrible at looking forward.
I'm not super great at looking forward, but I've learned to think. To consider. To pause and ponder, rather than just reacting. To examine possibilities. And with AI, the possibilities are amazing.
Bill Gates famously said "640K of memory should be enough for anybody." Now consumers are needing 16G of system memory just to get their OS off the ground most of the time, and a lot of "casual computer enthusiasts" will have, between system and video and CPU cache, north of 40G memory. Commercial data centers are installing individual racks that use hundreds of gigs of memory. Data science researchers are working with chip manufacturers to get production lines that'll churn out factory assembled components by the truckload that have hundreds of gigs each.
What Bill said made sense at the time. There were people who agreed with him, when he said it. And yet technology marches onward.
People hate that they're bad at it, but the fact is most people are bad at the future. At looking forward, at thinking, at analyzing possibilities and positioning to take advantage of them. But they are. Humans trash the new, always. It's usually not until the new is so obvious even the blind can see it that the tune changes, and suddenly the content mills are churning out "how did life ever exist before X?" Where, usually, a year or two prior the articles were "X should die screaming in flames."
I just want people to calm down and use their heads for more than just holding up their perfectly coifed hair as they take a selfie. AI isn't bad. There are bad people, and some of those bad people work in AI, but that's true of everything, everywhere, in the world. Bad people suck, and they're bad, but they're also just how humans are sometimes.
That doesn't make AI bad. Or mean it's not going to continue advancing. History books talk about "the rapid pace of development" as it relates to the transition into industrialization. Meanwhile, we're smack in the heart of the information age, where about a decade ago no one envisioned that just about every human on the planet would have a computer in their pockets that's more powerful than what corporations in the 70s would have paid a hundred grand or more to own.
And that everyone mostly uses those smartphones to take selfies and fire off emoticon posts that decry the crushing advancement of technology. Never noticing the irony. Which would be amusing if it just didn't get kind of tiring sometimes to hear them not noticing it as they declare the latest wave of advancements to be stupid and impossible.
I don't disagree with your message. I am just saying that you're wording earlier came across as something you might not have intended.
You are still doing it to some degree. Like, I am not even sure what you are even responding to in my comment. As your entire reply feels a bit much in response to what I said.
It's a conversation, not a debate ;)
I think AI is just a calculator on steroids. It's just a tool. The tech behind it is very cool and it will allow workers to be way more productive the same way spreadsheets revolutionized accounting and factories revolutionized the way we build things.
Right now, people are expecting it to do everything and maybe one day it will, but right now it's just a handy companion to help people use computers.
I guess the author doesn't know any programmers. Charging a subscription for better autocomplete and advice that you have to be somewhat wary of might not result in all that much revenue, but it does change the day-to-day experience.
I expect efficiency improvements and competition to lower the subscription costs. The only way to maintain them would be to significantly improve the results.
That might happen. There are hints that OpenAI will release something new this year.
It's frustrating, though, because these hints are indistinguishable from hype. Similarly, there was a fair bit of hype about GPT-4 before it was released, and none of it could be substantiated.
Edward Zitron was also recently in the Tech Wont Save Us podcast