[suyavuz]

I think this is basically right. You don’t hand out calculators before kids understand arithmetic. LLM version is sneakier because skipping the work still produces something that looks finished.

[aykutseker]

A calculator gives you an answer. An LLM gives you an answer that sounds like it already checked itself.

[cryo32]

Not quite. The LLM gives you a statistically probable sounding token stream. The calculator gives you a qualified answer within documented and deterministic limits of the device.

No one knows how to use either.

[charcircuit]

That is GPT-3. Modern models are rewarded based off the accuracy of their responses.

[Aerolfos]

By... another AI model. Which uses statistical generation to decide whether the answer is likely to be accurate or not.

[simianparrot]

No they are not

[p-e-w]

> You don’t hand out calculators before kids understand arithmetic.

You do hand out driver’s licenses before people learned to ride a horse though, and you do hand out matches before people learned to make fire by hand. So this isn’t a universal pattern, and you shouldn’t expect it to hold true in the future.