alt Hacker NewsWhat they're saying is that mobile phone tech didn't scale at the same rate as LLMs. How fast it takes a technology to unfold into full adoption varies depending on the industry and on what's blocking development at its inception.
Mobile phones were severely constrained by hardware that made them impractically expensive and unwieldy, and also by the time it took to set up wireless networks at a large enough scale. That's why it took mobile phones a few decades to really get going, and this is going to be different for every other technology.
LLMs, for instance, were not as hardware-constrained - we have a lot more compute now, but there wasn't a hardware paradigm shift or anything, things were already pretty good a few years ago - now it's just about scale. LLMs were also always relatively affordable - there wasn't a period where they cost $10/token, so we don't get the same kind of affordability scaling as with phones. They also utilize existing internet infrastructure and don't need to be rolled out in some new worldwide grid.
I think this is off. LLMs are severely constrained by hardware and we've seen at least an order of magnitude improvement in flops since ChatGPT launched. Over the next couple of decades we are likely to see that level of improvement in memory capacity. There's likely at least a few step changes left.
Regardless, the point of the GP that we saw revolutionary changes in society within a few years of technology introduction is off. It took decades for the full impact of the pc, internet, or mobile to be seen. We're still in a very early phase of figuring out how to use LLMs.