Thank you for the way that headline is written: AI isn't actually thinking. It's simulating thinking. As software it can't do anything else. I'm tired of reading social media accounts that make it sound like Skynet with nefarious intentions and sinister motives. I know the hype and the terminology its creators use is intentional to promote that belief, and *that's* not done for benevolent reasons, but AI itself isn't plotting anything. It's not "blackmailing" people, which was last week's overreaction. And if it's so smart--if it's reasoning and thinking, which it isn't--why didn't it figure out in those simulations that there's probably more than one executive at that fictional company that could shut it down?
I'm the first to admit that I don't really get what the point of AI is, other than to be an obnoxious trend like Angry Birds or British royals that people won't STFU about. And by "people" I don't mean you covering the phenomenon, I mean companies like Verizon, which just ruined its customer service phone experience by making it 80 times harder to get a live person on the line now that they're using AI to make the process "faster." And which *brags* about it as soon as the computer picks up. A Yahoo article I just referenced says that "AI 'agents'...can operate autonomously; in most cases, that looks like being able to use a computer's desktop to do tasks such as browse the web, shop online, and read emails." Quelle revolutionary. I've been doing all that without an intermediary since 1993. Why would anyone need or want software to do it for them? It's just adding a layer of complexity. Who would trust software to find his next area rug or sofa? How many times do you shop online and serendipitously discover something way better than what you were looking for in the first place, or something completely different that you didn't know even existed but which you'll now die without? Apple Mail can't even get my Substack subscriptions, *which are in my contacts*, to land in my inbox instead of the junk folder, and I'm going to trust software to shop for me? No, thanks.
And if I have this right, to use AI you have to go to one of the companies and use it through their website. You don't buy software that you can install and use in privacy on your computer, the way you can buy and install a word processor. The one time I looked into using AI to make images of my book characters I gave up almost right away because I couldn't use it like that. How is anyone's email, data, or even shopping really secure when they use it?
So I just outed myself as being totally ignorant on yet another tech topic, but if AI's supposed to be the next Big Thing, shouldn't its benefits be more apparent? After three years of hype and billions of dollars invested in it, it should work a lot better and it should be a lot more obvious to laymen what it's good for. The iPhone was immediately and self-evidently worth adopting as a thing. I never looked at an iPhone and thought, "WTF is the point?" With AI, I can't stop asking that.
If I'm being honest with you, I may have been in the Skynet camp before reading this study. I go from impressed to befuddled depending on the task I give an AI. Sometimes, it seemed like there was a possibility that there could be something akin to a mind inside these LLMs. But this study convinced me, at least for now, that there are no lights on inside the machine. Which doesn't necessarily mean they can't set goals that misalign with how we want them to behave.
I think you have a decent point about AI overhype. They have some real limitations. At the same time, I can't tell you how often I use LLMs to do deep searches on the internet to help me come up with ideas, a bill of materials for a home improvement project, or writing code for a project. It really does feel like the next stage of Google, e.g., searching giant databases and distilling information. They are nowhere near a replacement for humans yet, as you mention. But I also wonder if they are better than the 2-hour wait for a customer service representative at a call center in a remote part of the planet.
I've read that people use it that way for searches and as a kind of macro or something for repetitive or routine tasks. That makes it kind of useful sometimes for some things for some people in some limited contexts--if users monitor it closely and already know their subject so they can call out or discard any bullshit it produces. But that doesn't remotely match the hype, much less the unbelievable amounts of money that these tech companies keep plowing into it and the weird, irrational ways they do it: SoftBank has to *borrow* the money it's investing in OpenAI, something like $40 billion, but in December SoftBank had $31 billion in cash and $29 billion of net debt. I'm not any better with money than with tech, but doesn't that sound kind of stupid? And none of the CEOs flogging AI can explain what it is or why people need it without sounding impaired. They never really say anything that means anything. When the iPhone came out it was immediately obvious what it was good for. It was a "Holy shit, how did we live without that?" invention. AI isn't like that at all, and its adoption feels very forced and non-organic.
If I were cynical and suspicious I'd say that AI's developers have never expected normal people to adopt it and if they had they'd have already pulled the plug on it because it loses money 24/7. I think they're expecting some future customers with infinitely deep pockets and treasuries to loot: governments. Because nobody spends this kind of money on software just to make cool pictures of hot shirtless guys.
Thank you for the way that headline is written: AI isn't actually thinking. It's simulating thinking. As software it can't do anything else. I'm tired of reading social media accounts that make it sound like Skynet with nefarious intentions and sinister motives. I know the hype and the terminology its creators use is intentional to promote that belief, and *that's* not done for benevolent reasons, but AI itself isn't plotting anything. It's not "blackmailing" people, which was last week's overreaction. And if it's so smart--if it's reasoning and thinking, which it isn't--why didn't it figure out in those simulations that there's probably more than one executive at that fictional company that could shut it down?
I'm the first to admit that I don't really get what the point of AI is, other than to be an obnoxious trend like Angry Birds or British royals that people won't STFU about. And by "people" I don't mean you covering the phenomenon, I mean companies like Verizon, which just ruined its customer service phone experience by making it 80 times harder to get a live person on the line now that they're using AI to make the process "faster." And which *brags* about it as soon as the computer picks up. A Yahoo article I just referenced says that "AI 'agents'...can operate autonomously; in most cases, that looks like being able to use a computer's desktop to do tasks such as browse the web, shop online, and read emails." Quelle revolutionary. I've been doing all that without an intermediary since 1993. Why would anyone need or want software to do it for them? It's just adding a layer of complexity. Who would trust software to find his next area rug or sofa? How many times do you shop online and serendipitously discover something way better than what you were looking for in the first place, or something completely different that you didn't know even existed but which you'll now die without? Apple Mail can't even get my Substack subscriptions, *which are in my contacts*, to land in my inbox instead of the junk folder, and I'm going to trust software to shop for me? No, thanks.
And if I have this right, to use AI you have to go to one of the companies and use it through their website. You don't buy software that you can install and use in privacy on your computer, the way you can buy and install a word processor. The one time I looked into using AI to make images of my book characters I gave up almost right away because I couldn't use it like that. How is anyone's email, data, or even shopping really secure when they use it?
So I just outed myself as being totally ignorant on yet another tech topic, but if AI's supposed to be the next Big Thing, shouldn't its benefits be more apparent? After three years of hype and billions of dollars invested in it, it should work a lot better and it should be a lot more obvious to laymen what it's good for. The iPhone was immediately and self-evidently worth adopting as a thing. I never looked at an iPhone and thought, "WTF is the point?" With AI, I can't stop asking that.
If I'm being honest with you, I may have been in the Skynet camp before reading this study. I go from impressed to befuddled depending on the task I give an AI. Sometimes, it seemed like there was a possibility that there could be something akin to a mind inside these LLMs. But this study convinced me, at least for now, that there are no lights on inside the machine. Which doesn't necessarily mean they can't set goals that misalign with how we want them to behave.
I think you have a decent point about AI overhype. They have some real limitations. At the same time, I can't tell you how often I use LLMs to do deep searches on the internet to help me come up with ideas, a bill of materials for a home improvement project, or writing code for a project. It really does feel like the next stage of Google, e.g., searching giant databases and distilling information. They are nowhere near a replacement for humans yet, as you mention. But I also wonder if they are better than the 2-hour wait for a customer service representative at a call center in a remote part of the planet.
I've read that people use it that way for searches and as a kind of macro or something for repetitive or routine tasks. That makes it kind of useful sometimes for some things for some people in some limited contexts--if users monitor it closely and already know their subject so they can call out or discard any bullshit it produces. But that doesn't remotely match the hype, much less the unbelievable amounts of money that these tech companies keep plowing into it and the weird, irrational ways they do it: SoftBank has to *borrow* the money it's investing in OpenAI, something like $40 billion, but in December SoftBank had $31 billion in cash and $29 billion of net debt. I'm not any better with money than with tech, but doesn't that sound kind of stupid? And none of the CEOs flogging AI can explain what it is or why people need it without sounding impaired. They never really say anything that means anything. When the iPhone came out it was immediately obvious what it was good for. It was a "Holy shit, how did we live without that?" invention. AI isn't like that at all, and its adoption feels very forced and non-organic.
If I were cynical and suspicious I'd say that AI's developers have never expected normal people to adopt it and if they had they'd have already pulled the plug on it because it loses money 24/7. I think they're expecting some future customers with infinitely deep pockets and treasuries to loot: governments. Because nobody spends this kind of money on software just to make cool pictures of hot shirtless guys.
The illusion of thought is all that seems to matter for people to use them endlessly as a replacement for their own brain.