Also it will be used by thousands of people consequently for 8 hours straight so we can speedrun climate change I guess.
I’m going to assume this thing would run a model locally because the alternative would be insane, especially if you’re doing it to reduce the costs relative to hiring a teacher. And if it’s running a local model, then it’s at worst speed running climate change as much as a similarly-long gaming session.
Which is probably the future of embodied AI in practical situations - running a local model in an appropriately designed robot body to the task domain at hand. If for no other reason than because in real applications you’re not going to want response time relying on network turnaround for any important task, or most tasks that involve interacting with the physical world in general.
It will specifically talk in a context where the human cannot tell if it’s hallucinating or not.
LLMs are always hallucinating. That’s how they function. It pulls something that’s probably coherent from literal random noise and techbros pretend it was “thinking”, then when the result is something undesirable, they act like it made a “mistake”. Don’t buy into their bullshit.
Definetely worst.
It will specifically talk in a context where the human cannot tell if it’s hallucinating or not.
Also it will be used by thousands of people consequently for 8 hours straight so we can speedrun climate change I guess.
I’m going to assume this thing would run a model locally because the alternative would be insane, especially if you’re doing it to reduce the costs relative to hiring a teacher. And if it’s running a local model, then it’s at worst speed running climate change as much as a similarly-long gaming session.
Which is probably the future of embodied AI in practical situations - running a local model in an appropriately designed robot body to the task domain at hand. If for no other reason than because in real applications you’re not going to want response time relying on network turnaround for any important task, or most tasks that involve interacting with the physical world in general.
LLMs are always hallucinating. That’s how they function. It pulls something that’s probably coherent from literal random noise and techbros pretend it was “thinking”, then when the result is something undesirable, they act like it made a “mistake”. Don’t buy into their bullshit.