the irony is if any of these labs actually developed AGI "first" the incentive to shutoff their consumer services and dedicating all available compute to solving multiple "trillion dollar problems" would be overwhelming why worry about making $20 a month from a bunch of consumers at that point when you can dedicate your compute to identify massive ROI solutions to very valuable problems? that is if they are not nationalized by the government immediately maybe you'd see old hardware which isn't good enough still be used for consumer inference, but then you're still trading off energy use other scenarios worth thinking about are if one lab gets AGI, it's highly likely other labs will be very close behind or maybe the transition to AGI is not a binary jump from not having it to having it, but rather is a continuum we slide into- starting now as recursive model self-improvement may already be driving the bulk of new capability gains at this point