OK, you're wrong, but this should be obvious even to a genius like you.

How did you learn to read? Did you ingest a billion different texts? No.

How did you learn to distinguish the color red? Did you look at a million different labeled pictures and classify colors accordingly, or did someone show you a crayon and say, "this is red"? People learn concepts differently.

People don't produce words in the same way LLMs do. The reason you think they do is because you are ignorant: you haven't done your due diligence to understand the problem well enough to justify your strong opinion.

I was surprised they hadn't already done this. It seemed like an obvious target given the large test suite available. The result is interested in that it helps show the possibilities and limitations, the contours, of AI coding.

Now it's obvious why no one did it: $20,000 of compute time.

People are capable of writing original words in a way that LLMs are not. Covfeferetter.

The source code is available online if you want to look at it: https://github.com/anthropics/...

In any case I agree with your point that craftsmanship matters, and that has been lost because people are attracted to programming for the money, not because they enjoy it. (Of course, someone who wants money can be a skilled programmer too, but often they don't).

Using a helmet as a battering ram is absolutely illegal in the NFL. https://operations.nfl.com/the...

The rule changes have already happened. Head-to-head contact is illegal in the NFL. Hardly any spectators noticed, and fewer cared.

Soft outside helmets are used in the NFL, but not always.

That's why it's against the rules in football to knock heads together. https://operations.nfl.com/the...

If it's blatant the player who does it will be ejected. Both players may be checked for a concussion.

The NFL has made major rule changes over the years to avoid concussions.

For example, here are the rule changes that show proper tackling technique of the QB. I think even a non-player can tell the difference.

Congrats, your score is now two.

Neural networks are reasonably good at image recognition, so it should be able to recognize a school bus most of the time. The computer has more than milliseconds to recognize the school bus, just as humans do, approaching the bus over several seconds (unless it's stopped around a blind corner).

It seems likely Waymo didn't train their cars for this case, or made a mistake in the logic somehow.

Wow, nice!

There was a study years back that found ultramarathoners continue to find more health benefits the longer they run (all else being equal).

It's easy to think of a mechanism behind 1 minute of exercise improving health. For example, the lymphatic system needs daily muscle movement to circulate. Even a little movement will be enough to remove waste products. Furthermore the flow of the intercellular matrix requires muscle movement.

It might, actually. Of course attention will need to be paid to nutrition, but that's true no matter how much you exercise.

I don't think this can work without having an employee monitor the car feed 100% of the time. Otherwise they won't have enough context to advise the car in real time.