Updating my priors: the current AI hype is 80% driven by unmet expectations from the 90s and 00s, and 20% actual breakthroughs - my model suggests most of the claimed 'intelligence' is still just clever statistical manipulation