6 ms·
> Novelty can even be trained for and we have created agents that seek novelty I was thinking that too. Instead of merely copying writing styles by authors, it
by sysadm1n 4y ago
> Novelty can even be trained for and we have created agents that seek novelty
I was thinking that too. Instead of merely copying writing styles by authors, it would be possible for a program to dream up new writing styles, and generate idiosyncrasies algortihmically. We can reverse engineer novels and find out their structure and idiosyncrasies, but having the ability to create new idiosyncrasies would be the magic sauce needed to generate novels on demand.
- dtagames 4y agoWhere would novelty come from, since all code has to be written in advance? Even code that writes code must be written in advance. The ability to make something truly new is a unique quality of people, not shared by machines.
- krapp 4y agoPeople don't make anything "truly new," they extrapolate from similar information they already have and build on what they know through association or translation, which is what machines can also do. No one who has never even seen an internal combustion engine is going to spontaneously invent a better one, regardless of how brilliant they are.
- dtagames 4y agoWasn't the first internal combustion truly new? And the first computer? And the telegraph, etc? If they are just examples of a better something, what was the something?
- PartiallyTyped 4y agoNo, they were not truly "new". The first "computer" was very different to what we have now. An actually novel idea was Einstein's General Relativity, but even that built on top of Riemannian geometry, which built on top of other ideas. Another actually novel idea was the Copenhagen interpretation of Quantum Mechanics, and that needed Heisenberg to sit in a park and observe a person going under lights and then back into darkness. There are very few truly novel ideas in history, and even those built on top of other works.
- krapp 4y agoEverything you listed was based on existing knowledge and prior technology, sometimes combined from various sources in novel ways, but that isn't "truly new." Functionally, internal combustion engines are better steam engines, computers are better abacuses and telegraphs are better war drums and smoke signals. But none of them came fully formed and realized out of thin air. https://en.wikipedia.org/wiki/History_of_the_internal_combustion_engine https://en.wikipedia.org/wiki/History_of_the_internal_combus... https://en.wikipedia.org/wiki/History_of_computing_hardware https://en.wikipedia.org/wiki/History_of_computing_hardware https://en.wikipedia.org/wiki/Electrical_telegraph#History https://en.wikipedia.org/wiki/Electrical_telegraph#History
- dtagames 4y agoNothing comes out of thin air, as far as we know. My assertion is not something from nothing. It's something truly new and unseen. The idea that electrical transmission over wire is just a better smoke signal seems to deny the most basic differences between the two, or that there was any ingenuity or human contribution in the path from one to another. Many animals have language, the chief instrument of intelligence. But no animals other than humans have put language together to make something new -- like buildings, roads, or power plants. Why is that? Why don't other animals evolve language that makes new things? And why do the things not evolve anything new themselves? I think humans are unique and special. As I said, it could be my own bias from being one, but I'm gonna stick with it because -- in the practical world in which we live -- our position and role are unique and our intelligence can lead to truly new and unique outcomes which machines and other animals can't do on their own.
- krapp 4y agoTo me "truly new," as you keep using it, does imply something from nothing - some aspect of innovation or creative spark that doesn't simply involve the transformation of existing data or basis in existing knowledge. Otherwise, what is the difference between "truly new" and simply "new?" >The idea that electrical transmission over wire is just a better smoke signal seems to deny the most basic differences between the two, or that there was any ingenuity or human contribution in the path from one to another. No, I think we simply disagree on the nature of ingenuity. I argue that there is no aspect of human ingenuity or creativity that cannot be simulated by a machine of sufficient complexity because the human brain itself is simply a complex machine. It's just data processing - we do it every night when we dream. There's nothing ineffable separating the two that makes it possible for a human to innovate but impossible - even in theory - for a computer.
- soheil 4y agoSo if I build a machine that has as output other machines randomly generated that are not extrapolated from anything I know it wouldn't count because I still "extrapolated from similar information" I already had? Absurd line of logic.
- PartiallyTyped 4y agoI disagree. All a program needs to do is use noise to generate something new and never seen before.
- dtagames 4y agoI should have said something novel and useful. Every scoop of dirt from the ground is novel, but that doesn't mean that shovels invented anything.
- PartiallyTyped 4y agoAny randomly generated image from say Dall-E-2 uses some noise as input, and one can argue that they are indeed useful.
- dtagames 4y agoDALL-E is interesting, but it also helps prove my point. Asking for a picture of a man in jail, for example, produces an image of a black man in jail. That's because the images are not new; they are composites of existing images, provided by people -- along with the biases of those people. "New" methods of face scanning also identify black people as criminals more often because they are not new; they are based on human input and bias.