Kinda, more accurately the current goal for labs is to produce economic value to justify high valuations / capex, thus having to RL on correct answers rather than interesting ones.

If you could access the current models as base models I guess you'd still get very interesting hilarious stuff.