5 ms·
I think where people misunderstand these LLMs is they think they have creativity. But they don't. They cannot think of something new. They can simply rehash. So
by ilrwbwrkhv 4y ago
I think where people misunderstand these LLMs is they think they have creativity. But they don't. They cannot think of something new. They can simply rehash. So it is useless for any actual creative work. Once you have the creative idea, it can generate a few paragraphs which sound OK but you need to carefully double check.
- username_my1 4y agoIs it really a big problem though? Isn’t most creativity mashing things together that no one before thought of putting together?
- elif 4y agoI spent about 10 hours one day trying to convince my professional artist friend that AI could produce legitimate works of art, against his insistence that they could at best produce "kitsch" collections of art-like patterns, the kind of stuff that belongs on the wall at malls and not being discussed by academics in galleries. I have to say that, although I still disagree with him, I was unable to get GPT to produce midjourney prompts which I would consider creative. Even when I prompted GPT to take the perspective of a known artist, it produced an endless series of banal landscapes, cliche compositions, overplayed metaphors, and insipid imagery. Midjourney only really accels when a human is involved with the prompting. I had to concede to him at the end of the discussion that /today's/ AI are not creative. I firmly believe though that the inclusion of all mankind's perspectives possesses the raw material to generate creative ideas. My understanding is that current AI lack the ability to "turn off" the vanilla oversaturation of averages that pervade when you are exposed to all stimuli. One day, perhaps soon, there will be AI able to combine ideas they are exposed to without trying to combine ALL of the ideas at once.
- Galaxeblaffer 4y agohow is that different from humans ? how common is truly original thought in humans ? to my knowledge humans work exactly the same, we "simply" rehash and mix existing ideas and repeat ending up with "novel" concepts. To common people these ideas may seem original, while to an expert it's just rehashed information.
- peoplefromibiza 4y agooriginal work often does not mean doing something already done in a slightly different way (paint a different set of sunflowers than Van Gogh) What these models do is more similar to pastiche, if we really want to compare it to some technique. But the value of the pastiche depends on the intent and its perception by the viewer(s), because there's nothing inherently original in it. So basically what models produce has no value, unless we are able to attach some to it. [1] If we asked chat-gpt to analyze something it has produced, it will probably say that it is "similar to" or "in resemblance of" but it's unlikely that it would say "this is the work of genius, how original! lovely!" [1] edit: you don't read "pizza maker creates fabulous art on AI" but <person who's already in the business> won a contest of <some art form> submitting something created by <AI of your choosing>. Why? Because they know how to market it and can rely on other people believing that's their creation. Nobody says it upfront "I will submit an AI generated work", because they know it won't be judged the same way. The pizza maker was probably trying it for fun or to make a new logo for the pizza place or simply has no instrument to assign it a value and convince other people that it is true (including using their professional card).
- hawski 4y agoIt can't reason at all. When it is wrong it may correct itself, it may double down on being wrong and often just make something up again.
- Zinu 4y agoDoes "it" refer to GPT, or are you talking about humans in a derogatory way?
- kybernetikos 4y ago> It can't reason at all. This isn't true. You can ask it basic logic problems that it's never seen before, and it will apply the rules of logic to them. It can also identify correctly which rules of logic would make sense to apply in more complex situations, even when it doesn't get the answer right straight away. At doing this stuff GPT4 is better than GPT 3.5 which is better than previous GPTs. I fully expect that future models will be able to tackle more complex applications of logic to new domains successfully. If you use only examples that weren't in its training set, you'll get to its limits quickly, but basic level first order logic is definitely within its ability.
- bolyarche 4y ago[dead]
- flangola7 4y agoI have positively had GPT-4 create genuinely creative answers. "More creative" is even listed as one of its traits in the ChatGPT model drop-down. Ironically I think anyone who believes LLMs cannot be creative are themselves uncreative in their prompt crafting and use of the model, I've even seen this from senior developers and experienced writers. It's painful, like watching your mom try to find the downloads folder or enter terrible search engine keywords.
- psychoslave 4y agoThe technology provides very interesting and impressive results, but to me it seems people are interpreting very wrongly its underlying process, trying hard to see emerging consciousness and whatever any time the result doesn’t manifestly show that it’s not having such a characteristic. I tested GPT on its ability to handle word creation with suffixes. Some results are really interesting, but sometime it pretends a word is a sample of some suffix use when the word doesn’t encompass it. I’m rather confident that this kind of artifact can be overcome, I won’t slander the technology because it still has some obvious weakness. But I’m afraid that once overcome, the lake of snake will convince most people all the more that this impressive technological achievement is more than what it is actually.
- tomtomtom777 4y agoIt is only the unimaginative who ever invents. The true artist is known by the use he makes of what he annexes. - Oscar Wilde
- pmontra 4y agoI think that better philosophers than us debated this question for thousands or years but I'll give my take. Suppose that creativity means to find a particular place in the n-dimensional space of all possible ideas. Artists find some interesting places in that space in part because they started from other interesting places nearby. They know those nearby places and we don't, because we trained to do a different job. However we have our share of creativity. Algorithms like GPT or Stable Diffusion explore that n-dimensional space starting from possibly more places than a single human has experienced. They will miss many of the starting points that any human experiences, because of limits of those algorithms' interfaces with the world. However they can still find interesting places. Maybe it's not creativity as we mean it, but they can show us interesting places that no artist has found before.
- mhh__ 4y agoAre most things you do truly new? Obviously we can say things like "What about a car that ran on pencils" but real, useful, entropy comes from inferences and abduction rather than the new.
- dagss 4y agoWhat do humans who exhibit "creativity" typically do but rehash themes and ideas from others? Trends rather slowly evolve as they reverberate back and forth between thousands of humans -- whether in music, painting, clothing, books, memes, ... If an artist just thinks of something completely new not inspired by others it only appeals to a small minority. It is a minor portion of human creativity, and avoided by perhaps most humans.