I keep reading this over and over and of course it's a natural impulse to just extrapolate in a linear fashion.
The problem with this is addressed tenfold in this substack; the reason Gary and others (like me) reject the scaling hypothesis is not that we can't extrapolate, it's because there are fundamental issues with it and repeating yo…
I keep reading this over and over and of course it's a natural impulse to just extrapolate in a linear fashion.
The problem with this is addressed tenfold in this substack; the reason Gary and others (like me) reject the scaling hypothesis is not that we can't extrapolate, it's because there are fundamental issues with it and repeating your point does not increase my propensity to abandon to the facts.
Your example with model T is particularly flawed because it precisely did *NOT* follow a scaling that is necessary for AGI: The model T was one order of magnitude less efficient, sure, but it was driving about the same speed (~factor 2) and cost the same money (inflation aside). You may quote comfort and convenience as major improvement points in the last hundred years, and I'd agree with that, but let's be honest: did the economic reality change because of how automobiles evolved since model T? Not at all, just like it's not a good bet to hedge for a replacement for chairs and beds that have been around for even longer. The truth is in the marginal contribution and it's just not there.
I keep reading this over and over and of course it's a natural impulse to just extrapolate in a linear fashion.
The problem with this is addressed tenfold in this substack; the reason Gary and others (like me) reject the scaling hypothesis is not that we can't extrapolate, it's because there are fundamental issues with it and repeating your point does not increase my propensity to abandon to the facts.
Your example with model T is particularly flawed because it precisely did *NOT* follow a scaling that is necessary for AGI: The model T was one order of magnitude less efficient, sure, but it was driving about the same speed (~factor 2) and cost the same money (inflation aside). You may quote comfort and convenience as major improvement points in the last hundred years, and I'd agree with that, but let's be honest: did the economic reality change because of how automobiles evolved since model T? Not at all, just like it's not a good bet to hedge for a replacement for chairs and beds that have been around for even longer. The truth is in the marginal contribution and it's just not there.