3 ms·
I'm not convinced LLM's are capable of any truly original thought, which is necessary for good compelling writing. LLM's spit out derivative nonsense. Highly c
by babypuncher 3y ago
I'm not convinced LLM's are capable of any truly original thought, which is necessary for good compelling writing.
LLM's spit out derivative nonsense. Highly convincing nonsense, but derivative nonetheless. Any true originality in the output still comes from the human engineering the prompt.
Basically, LLM's know how to follow the rules, color within the lines. A real artist understands those rules well enough to break them and create something novel.
- gojomo 3y agoThe latest models generate original combinations that can surprise the prompter with their inventiveness, so your confident assertions about what LLMs can't do seems calibrated to the models of a year or more ago – and not at all an enduring human advantage that one could bank a career (or society) on. Attempts to rigorously study the relative creativity of LLMs have already found them more creative than average humans: https://www.nature.com/articles/s41598-023-40858-3 https://www.nature.com/articles/s41598-023-40858-3 They're only getting better, fast. If you're not convinced, what GPT4-level (text or multimodal) or Midjourney-level (imagery) models have you tried, and how much?
- randomcarbloke 3y agoSo what you're really saying is there are no true artists left in Hollywood.
- babypuncher 3y agoOh they exist, they just have a much harder time finding work, and when they do get to make something it rarely gets the same marketing push as comic book shared universe project #87. Some great movies still come out, but you probably never hear about them if you aren't actively plugged in to the scene. I think filmmakers like Daniel Kwan and Daniel Scheinert (Swiss Army Man, Everything Everywhere All at Once) are where Hollywood should be focusing their attention, and their sweep at the Oscars this year has me cautiously hopeful that at least a few people in power are paying attention.
- zeroonetwothree 3y agoThis hasn’t been my experience at all. It’s just as original as humans, especially at the “derivative” style of most popular media. It’s probably true that humans are still better at extremely creative thought but that’s not what most commercial writing is.
- babypuncher 3y ago> especially at the “derivative” style of most popular media This is what I see as the problem. Hollywood needs to stop relying so heavily on derivative goop and give actual creatives more leeway in creating something original. You can only ride old IPs for so long before people lose interest and stop buying tickets, as Hollywood has been learning recently. You need people coming up with fresh novel ideas if you ever want to come up with the next big thing. Some executive asking Chat-GPT to spit out a new Marvel movie script isn't going to do that.
- dleslie 3y agoI'm not convinced that most, or any, humans are ever engaging in truly original thought. This idea that AI cannot be creative but humans can is something that begs for a test; and that test needs to be applied against humans, as well. If AI can pass it at least as often as humans can, then AI is at least as creative as humans; within the bounds of that test. As it happens, an AI has recently passed such a test[0]. 0: https://www.technologyreview.com/2023/09/14/1079465/ai-just-beat-a-human-test-for-creativity-what-does-that-even-mean/ https://www.technologyreview.com/2023/09/14/1079465/ai-just-...
- babypuncher 3y agoone possible test would be if a model trained solely on naturalist and earlier paintings could eventually start producing expressionist and abstract works, without carefully engineered prompts.
- dleslie 3y agoEven for humans there was a transition period, and not a sudden explosion of new expression. Naturalist gave way to impressionism, which gave rise to expressionism, and eventually then came abstract works. There's a clear, gradual change in style between them all. I don't see why you couldn't have an AI do the same, by training it towards a particular aberration in output until that became something it understood well.