3 ms·
That's been a very common reaction to many deep learning successes the past few years. I lost track of how many people seriously suggested that a GPT-3 output m
by gwern 3y ago
That's been a very common reaction to many deep learning successes the past few years. I lost track of how many people seriously suggested that a GPT-3 output might be some outsourced laborer in realtime, never mind how fast the output came back for arbitrary inputs!
- TeMPOraL 3y agoIf we're talking about general reactions of serious (and supposedly serious) people, and not just government (as per GP's question), then SpaceX reusable rockets also qualify - I remember people were still arguing that physics makes it infeasible, some even saying fundamentally impossible, for some time after the first successful landing of a booster in a regular launch (and not a test). (Ironically, this briefly turning into some kind of conspiracy thing was likely fueled in large part because SpaceX couldn't get something else right: their live streaming setup. Even a a year later, the streaming still tended to cut off exactly as the booster was about to land, and resumed with it either in pieces, or safely on the pad. This was a big problem with barge landings in particular - where IIRC the primary explanation was that they just couldn't keep the antenna pointed in the right direction with all the shaking induced by the booster's hoverslam.) (And yes, I appreciate the irony of getting hung up over live streamed booster landing, as if it was common and expected practice.)