it's already known that from spacex to nvidia, from google to openai, it's all a circular ponzi scheme. the most important factor determining whether or not all that speculative capital investment will pay off is the productivity trend of the flagship model. but that's a long-term prediction. by the time that horizon arrives, you've taken the company public, collected money from the public, and inflated the share price as much as you want based on public expectation rather than experts. as long as the derivative of productivity sits at a certain slope.
anthropic is already used to doom-marketing along the lines of "now this model is gonna mess everyone up, don't forget to ask for it at your dealers tomorrow." the well-known ipo is at the door too, and this has time-bombed reasons like the market collapsing once trump turns into a lame duck. but fable was probably measured at a performance that's a harbinger of the structural model plateau, meaning that the trillion dollars of capital buried into it won't turn over quite that many trillions, and a receiver had to be appointed to public perception.
they had already started giving away some of the problems by changing their pricing and mechanisms. then before fable came out, they prepped the ground in advance by going on and on about oh no, cybersecurity, some object or other is approaching. that way, when the model was released with such a badly excessive amount of guardrails that it was unusable, nobody would get suspicious. i'm sure they got some marginal improvement out of all the user data and such, but the rest of the model's visible productivity, in my opinion, mostly amounts to flooring the engine right at release time, internalizing the market's harness stubbornness into their own harnesses. (making it burn tokens like an ox with a supervisor prompt along the lines of "try ten times until you find it, don't ask the user" instead of saying "i couldn't find this, what should i do" isn't a skill, but it isn't productivity either. people who use openclaw and the like will get what i mean.)
if the amount of profitable work that can be done per unit token or dollar or time isn't rising as fast as the capital investment being made, those shares are going to fall there. normally this is a healthy thing. but if what you've got on hand is bloat and circular pumping, then this truth coming to light has very unpleasant consequences. especially when you consider that the american economy has nothing else whatsoever that promises any hope of growth. and this national security matter in question is exactly that.
top: all
|
today
search in topic