This is the typical AI booster’s game: make vague claims about what AI can already do today without showing one single example, then extrapolate to an imaginary future and say that we’re all too stupid even to debate him.

Instead of writing some post like this, I’d love to hear ONE SINGLE EXAMPLE of AI doing a core work task faster, better, and cheaper than a person can do it, now, today. Here’s the task. Here’s how people did it before AI. Here’s how they do it now. Here’s how much time or money they save.

SHOW ME OR STFU

Funny to see people think AGI is coming quickly... While autonomous driving is stalling far short of projections made a few years ago.

If an AGI is really G(eneral), then we should be able to give it the wheels of our cars (and the sticks of our planes, etc).

Now, what is happening in real, actual life? Very impressive sudden progress around 2010-2015 to L3-L4, after decades of underwhelming prototypes (DARPA grand challenge & co), a big drop in costs at the same L3-L4 level demonstrated by people like comma.ai, and... Every prediction for the emergence of L5 has since been blown.

Why would the general case of AI be better than the particular case of self-driving?