ChatGPT and other generative AI isn't about to create superintelligence, but even if what we see now is it, the impact is on the scale of Google search and algorithmic social media, except maybe worse
AI can be very efficient. That said, I refuse to use it to write even an email. The benefit of writing is it forces one to think deep and hard, and the opportunity to read how logical or otherwise your conclusion is. If AI does even the first draft, I will have to learn how the facts and figures, along with interpretative views are pieced together, making it harder to defend any positions as they are not mine to begin with.
It depends on the use. For original writing, AI lacks the “voice” and its train of thought wouldn’t be mine. But for mere reporting purposes, it’s often
fine. Over the middle term I think people will settle on using it as a suggestion, rather than a replacement.
Captures my POV precisely. I’ve been doubting the AI mania for a year, and sold all our AI-related equity several months ago. It’s clear that “model corruption” is having the same effect that click bait has had. This has the potential to damage serious AI science for years.
Thanks. I've been doubting it for a while too, and seeing the hype stumble with the release of GPT-5 gave me the push to write all these thoughts down.
Model corruption is an interesting question, because it predicts the models will stagnate or even degrade, while most predictions assume they'll keep improving, even if they differ on how far or how fast. I don't know.
Like you, I think the AI-related stocks are in a bubble, and don't have financial exposure to them (beyond their overall effect on the stock market, which isn't small). That said, I'm not a financial advisor, I'm not giving anyone financial advice, and even if I'm right about the bubble, that doesn't mean I know when or how it'll pop. It could keep inflating for a while or crash soon. Or something else. We'll see.
I compare model corruption to clickbait in the way they both, over time, have or will give rise to doubt and disdain in the public mind regarding reliability and utility. Once people begin to distrust AI, they will mock it and disregard it, and further development will be impeded as investors hold back out of fear that there is no route to monetization.
AI can be very efficient. That said, I refuse to use it to write even an email. The benefit of writing is it forces one to think deep and hard, and the opportunity to read how logical or otherwise your conclusion is. If AI does even the first draft, I will have to learn how the facts and figures, along with interpretative views are pieced together, making it harder to defend any positions as they are not mine to begin with.
Exactly. Same.
It depends on the use. For original writing, AI lacks the “voice” and its train of thought wouldn’t be mine. But for mere reporting purposes, it’s often
fine. Over the middle term I think people will settle on using it as a suggestion, rather than a replacement.
Captures my POV precisely. I’ve been doubting the AI mania for a year, and sold all our AI-related equity several months ago. It’s clear that “model corruption” is having the same effect that click bait has had. This has the potential to damage serious AI science for years.
Thanks. I've been doubting it for a while too, and seeing the hype stumble with the release of GPT-5 gave me the push to write all these thoughts down.
Model corruption is an interesting question, because it predicts the models will stagnate or even degrade, while most predictions assume they'll keep improving, even if they differ on how far or how fast. I don't know.
Like you, I think the AI-related stocks are in a bubble, and don't have financial exposure to them (beyond their overall effect on the stock market, which isn't small). That said, I'm not a financial advisor, I'm not giving anyone financial advice, and even if I'm right about the bubble, that doesn't mean I know when or how it'll pop. It could keep inflating for a while or crash soon. Or something else. We'll see.
I compare model corruption to clickbait in the way they both, over time, have or will give rise to doubt and disdain in the public mind regarding reliability and utility. Once people begin to distrust AI, they will mock it and disregard it, and further development will be impeded as investors hold back out of fear that there is no route to monetization.
Plus there's the cost of running it: https://time.com/7308925/elon-musk-memphis-ai-data-center/