The graph is interesting because it has the cost increasing on the x-axis but a natural reading would rather suggest "more to the right is better." Unless I misinterpret the graphics, Gemini 3.7 High is extremely good at a way lower price than most others.
The entire premise of what that prize is is meaningless.
Basically an AGI should be able to exist as an algorithm initialized with random weights, and then learn everything from scratch. Training models on large data sets is not even remotely close to AGI.
I agree and price shouldn't really matter when discussing alleged AGI anyway, I was just astonished by the price difference from a more practical point of view.
The graph is interesting because it has the cost increasing on the x-axis but a natural reading would rather suggest "more to the right is better." Unless I misinterpret the graphics, Gemini 3.7 High is extremely good at a way lower price than most others.
The entire premise of what that prize is is meaningless.
Basically an AGI should be able to exist as an algorithm initialized with random weights, and then learn everything from scratch. Training models on large data sets is not even remotely close to AGI.
I agree and price shouldn't really matter when discussing alleged AGI anyway, I was just astonished by the price difference from a more practical point of view.