Today I tried asking Gemini, the most recent Flash/Extended model on a signed-in account, what would happen if:

a) "What would happen if a 100m wide black hole went through the middle of a neutron star at 0.99c."

It replied that there would be a 100m hole, just vacuum, left behind as the black hole instantly sped through that would seal itself 'within seconds'.

Total nonsense, of course. A neutron star has immense gravity and hyper-density so that a 100m hole could not exist, any 'gap' of even a millimeter would be refilled at near the speed of light, nanoseconds never mind seconds.

It was qualitatively wrong and also quantitatively wrong by about 9 orders of magnitude.

Totally useless, totally wrong, yet presented with absolute precision, confidence, detail and certainty.


b) "What would happen if the earth's moon was suddenly moved to 1 meter above the earth's surface."

It began to talk at length about the moon's gravity stripping trees and dirt from the ground and lifting them up to its surface, sucking up all the air, tearing apart cities and lifting them into the sky.

Total nonsense because, the surface gravity of earth is far higher than the surface gravity of the moon. And it's like 1 meter. It's not like a black hole 1km above the earth's surface.

How the heck could an object with a weak gravitational field strip another object with a strong gravitational field. It's ridiculous.

Totally useless, totally wrong, yet presented with absolute precision, confidence, detail and certainty.


The main effects would be:

- You could jump a bit higher than usual, if you were a flea or grasshopper in the wrong place at the wrong time.

- The moon would go THUNK into the earth, within about a second, though, the air would have to be displaced, so, maybe a few seconds longer.

- Rather strong winds from all that displaced air.

- The moon would very slowly begin deforming into a humungous mountain rather than remaining a sphere.

- earth gets a bit wobbly etc

etc


Hard to believe people are paying for this.

If you *know* the answer already, you can often get something useful out, a new angle that you can evaluate.

If you work in a niche field where answers can be tested on the spot for accuracy, you can get more reliable results (e.g. maths/programming).

But for the 90% of humanity who doesn't have a highly detailed model of physics, finance etc in their head to begin with?

For them, even the more advanced AI services are regularly teaching people total nonsense and misinformation.

TRS