Self-driving usually intersects with EV's simply for the fact that EV's are a newer idea, and self-driving can't (sadly) be really well currently added to existing cars. Comma.ai isn't really on the track to compete.
Those that are in market for an ICE car have an order of magnitude larger used car market available for them. You can get a car from the year 2000 and drive it no problem.
If you are looking at a car that's the best worth for your money, while "subsidizing" better future tech that will directly save lives, if the latter is an added negative for you, well that's just misanthropy.
In my admittedly little experience running a tiny LLM service on a single 5090 for friends, I would guess a fair bit. It depends on how many tok/sec you're targeting and how much memory you're allowing be used for context (and their context). Definitely more than 1, MAYBE an order of magnitude.
A dull knife can still cut. I’ve run across ones so dull they couldn’t cut clear tape on packages. Literally just grabs the tape and it rips off . Like I used my finger.
i find these discussions comparing the "vision language models" to the old computer vision tech pretty interesting
since there are still strengths the computer vision has, i wonder why someone hasn't made an "über vision language service" that just exposes the old CV APIs as MCP or something, and have both systems work in conjunction to increase accuracy and understanding
