Raw LLM Responses

Excellent video! I'll definately be directing people who have questions about generative AI towards this. Two other points i feel are worth bringing up: AI inbreeding and the Environmental Impact 1: The sheer volume of generated images flooding the internet and art centric platforms, the next generations of image generation are coming across the problem that their newer datasets are going to have these AI generated images, from previous, worse generations of AI, which will teach them bad data. The potential solutions are to either curate these absolutely massive datasets manually, find a way to detect any and all forms of ai generated imagery to filter them, or completely start from scratch or mothball AI. Only the second option seems likely to happen, and even then, AI detection isnt foolproof. This might lead to a sort of "snake eating its own tail" problem, where even if the technology to learn gets better, the AI might be getting worse material to learn off of. 2: Currently, the processing power to power generative models of AI take an absolutely huge amount of processing power, and insane amounts of water to cool it off, to make work properly. Most generative AI doesn't just happen on a single phone or computer, most of the commercially available ones request that a massive data center fufill the request in the short timeframes needed to be appealing to consumers. As various forms of LLM based technology becomes more widely used these data centers are going to become bigger, hungrier, and more expensive. Unless quantum computing and fusion technology suddenly become as commonplace tommrow i dont see this issue being solved by just making it more advanced.