Rocket AI: 2016’s Most Notorious AI Launch and the Problem with AI Hype
Riva-Melissa Tez

Making up a machine learning speech on the fly with a glass of champagne in my hand was the most entertaining experiment I have ever performed (the champagne was necessary equipment: were I to stumble, the audience would of blame the drink). Although I suspect somebody is going to invent TROL(tm) for real, just because.

Machine learning/neural networks/AI has a long history of hype bubbles. While people in the field with a bit of a historical perspective are aware of it and the damage they have done, it is still far too easy to get carried away. If you get rewarded for overpromising and you know that if you are not stepping up somebody else (likely far less conscientious) will get the money, can you resist? It is very much a “unilateralist curse” situation where there will be hype despite everybody knowing it is a problem.

The irony might be that the people who have the best chance of saving machine learning from hype bubbles are the venture capital people. If they actually try to go for value and demand some evidence of solving real problems they change the incentives. Plus, it is closer to their own goals of long-term profitability too. I hope.

Otherwise we will no doubt see many more RocketAIs take off.

Show your support

Clapping shows how much you appreciated Anders Sandberg’s story.