The Hype Thermometer Is Broken Again

Robert Scoble and Matt Shumer posted within minutes of each other on March 26th, which is the AI equivalent of two seismographs both spiking at the same time — not proof of an earthquake, necessarily, but proof that something shook the table.

Scoble has been the hype canary for twenty years. He was there for RSS, for blogging, for mobile, for Glass, for the metaverse — each time certain, each time early, each time pointing at the thing and saying this is the one. His track record is not the point. His function is. He is a sensor, not an analyst. When Scoble goes loud, something in the atmosphere changed, even if nobody can agree yet on what it is.

Shumer is newer to this particular role but has found it fast. HyperWrite has given him a platform, and he uses it the way you'd expect — capabilities demos, benchmark reactions, the occasional thing that turns out to be real.

When both of them post within minutes about the same thing, you have a choice. You can treat it as signal, or you can treat it as two people with very online brains reacting to the same stimulus in the same stimulus-soaked moment.

The honest answer is: both. The hype is real and the hype is hype simultaneously. Something is genuinely happening with these models — the jump between GPT-3 and GPT-4 was not nothing, the jump to Claude 3 Opus was not nothing — and also the social infrastructure around AI has been tuned to maximum sensitivity for three years now, so even incremental improvements register as civilizational events.

The thermometer is broken in a specific way: it reads every fever as the same temperature. Hot. Very hot. The hottest thing ever. Until the next one.

What you do with that information is your problem.