Will GPT5 show clear signs of diminishing returns?
When GPT5 comes out, plot a graph of its capabilities according to all quantifiable metrics, and chart that against the cost to train it and operate it. If it's clearly getting more expensive faster than it's getting better, or if the increases in capabilities are a significant slowdown from previous jumps (GPT2 to GP3 to GPT4) this resolves YES. CLARIFICATION: RE: the last line about capabilities, what I meant was, "if there is a significant slowdown in capabilities from this generation to the next sufficient to mean the cost curve still shows diminishing returns," which in hindsight is redundant. I spun off a separate market for those who want to bet on JUST whether the capabilities jump will be less than we've seen before, regardless of cost curve: https://manifold.markets/LarsDoucet/will-the-performance-jump-from-gpt4 If there's no CLEAR signs or it's ambiguous, this resolves NO.
