Almost the entire AI / ML landscape is still stuck on the pretense of: "lets feed it a trillion pieces of data, correct the most obvious defects, and call it a day", which leads to some terrific results, but mostly really poor and low quality results. It is fast becoming the buzzfeed of software. Generated images are full of obvious glitches. Chats with AI are full of logical mistakes and very weird changes of topic. Generated text is a lot of the time borderline unreadable. What exactly is the end goal here? Release early, release often, release all hopes of being taken seriously?
No comments yet.