Feels like an incredibly shallow take on this technology. I predict this article will age horribly within a decade.
IMO if you can’t see the difference between crypto and LLMs, you aren’t paying enough attention. I think it’s much closer in impact to the internet than to crypto. Just because something is wildly overhyped, doesn’t mean that what’s underneath the hype isn’t huge.
There’s tons of value in AI, the only questions are how far will this improvement curve go before leveling off and how long will it take to usefully integrate the technology into various domains. The tide will go out and we’ll realize that many emperors have no clothes, but there will be many AI companies that have strong understanding of the reality of the tech, have strong technical and business fundamentals, and will be well positioned to become Fortune 50 companies.
The short answer is that the technology as of today works and is useful. Text summarization. Text rephrasing. Image, text and video generation. Intent extraction. Voice generation. AI brings structure to previously unstructured data. AI allows us to solve problems we couldn't solve before where we aren't smart enough to write down the rules but we are smart enough to collect and label examples. It's imperfect but it works well enough to be useful - the dumb hype is driven by usecases that haven't found the right use of the technology given current accuracy realities. If you have enough data and can frame a problem as a sequence problem, these models can help you solve it. The hardware and software is rapidly improving and costs will go down. As the productization of this tech gets better we are better able to leverage today's technology in useful ways - consider chatgpt vs copilot vs cursor as ways of leveraging AI to help with coding.
Cold take, assuming because completely AI output is not amazing that nobody is going to use the AI more intelligently than prompting for the finished end product in one go.
Artists use stable diffusion like a magic paint brush to transform sketched compositions into exceptional art. Writers use it to quickly trial ideas, generate fairly good dialogue and create a feedback cycle even when they don't have readers. Just because completely AI generated images/writing aren't amazing _NOW_ (and that deserves the emphasis), doesn't mean that they can't be mined for nuggets of gold.
Anybody remember when speech to text was hyped up in the 90's? Dragon Naturally Speaking and IBM Via Voice? It was going to change the world. Ended up being a few mundane things like 'hey siri', annoying car voice commands, and really annoying customer service. I think AI is going to end up like that, a few niche uses (maybe really good, but really niche), and with that fade into the background.
Yep, that was like three AI hype cycles ago (in between we had the self-driving cars era of CV, the prior Microsoft Tay-era chatbot mini-boom, and arguably also an expert systems bubble in the noughties, though that never really went anywhere). Meanwhile, 30 years later, voice recognition is finally at the point where it can be used for low-stakes scenarios where correctness doesn't matter.
IMO if you can’t see the difference between crypto and LLMs, you aren’t paying enough attention. I think it’s much closer in impact to the internet than to crypto. Just because something is wildly overhyped, doesn’t mean that what’s underneath the hype isn’t huge.
There’s tons of value in AI, the only questions are how far will this improvement curve go before leveling off and how long will it take to usefully integrate the technology into various domains. The tide will go out and we’ll realize that many emperors have no clothes, but there will be many AI companies that have strong understanding of the reality of the tech, have strong technical and business fundamentals, and will be well positioned to become Fortune 50 companies.