gimlichgu

because selling inference is not their end-goal product, its just a milestone to continue to maintain positive public sentiment and attract talent

article is complete unreadable slop.

But to reply to the headline, becuase it's impossible to predict reliability of something new, and it's not possible to evaluate reliability based on a single example, you need lots of data collected over long periods of time.

The only thing most reviewers can do is speculate based on the reputation of the designer/manufacturer, and an educated guess based on the object's architecture and materials.

maybe, or maybe some problems turn out to be reasonably defeatable by educated guess brute forcing with lots of compute and big models pointing the search in reasonable directions.

If I'm not mistaken, a lot of math researchers work at openAI full time on using LLM's to tackle open research problems

this must be nailed to the wall somewhere at autodesk

dont question computer god!!

You're a genius and the machine god just confirmed it.

its helpful so i dont click on it

i'm telling ya clip the fire department, if a car wreck happens you can just have a doorfash guy come put cones around and light the accident on fire, itll clear up in a couple hours boom i just saved taxpayers trillions of dollars in waste

we need approximately 0 of you, want much less though

No! the only people that deserve to have enough money to be comfortable are people writing react frontends for surveillance tech