People in security and computing have been saying for years - there's no cloud. There's just someone else's computer.
Right now, there's no AI. There's just someone else's work.
Stop calling generative text and image programs AI. It's inaccurate and insulting. They are just the evolution of corporate creative theft that's been going on as long as media corporations have existed.
I guess this is like computer generated random numbers, for a computer to generate a random number it needs instructions as to how to do that, so it is not true randomness, like it would be if pulling numbers out of a hat.
@dj3ei @zleap @ianrosewrites it's a matter of information theory: a computer cannot add information, just erode it through computing.
To get a good likeness of random we introduce entropy from the external world.
As of this day we don't know if quantic generator are randow or if they are just unpredictable with our current knowledge of science.
Humans are very bad at random, that's how were catch many killers trying to randomly pick victim, but subconscientiously making patterns.
@dj3ei @zleap @ianrosewrites The brain is much more complex and unpredictable (in the general case) than computer, but as of this day, nothing from the brain as been demonstrated impossible to model.
LLM is a surprisingly efficient way to extract knowledge from text, and as our (ie human) knowledges of the world is only partially constructed from text, we can easily conclude that we are not «the same».
But it doesn't mean LLM chat does not mimic right the computation of some part of our brain.
Completely analyzed (non-random) and unpredictable can occur together.
That's one of several highly interesting "can't be done" results of the 20th century. "Chaos theory" says certain systems amplify minimal changes over time (changes below whatever measurement precision you payed for), thus defying long-term predictability. Weather is an example.
Other such results: Mathematics will always remain incomplete. No algorithm can analyze all (other) algorithms.
@dj3ei @zleap @ianrosewrites I would not spontaneously group G. incompleteness with chaos theory.
Incompleteness imply that some problem cannot have their solution computed.
Chaotic system can be simulated (but the real world does not allow you to collect all the required initial informations), and that's how work pseudo-random generator. Cryptographic PRNG work on the hope that it will be hard to deduce the state of the system from the output.
I completely agree with your analysis.
I just take the liberty to throw assorted facts from various disciplines into the same drawer, labeled "stuff we found out cannot be done".
A bit deeper in that same drawer, I might dig up "perpetuum mobile", "squaring of the circle", "prove the existence of god with philosophical arguments", "root(2) as a fraction".
I find this drawer amusing, and seriously think those results each an important achievement. That's all.
@dj3ei @zleap @ianrosewrites I can understand that :)
@dj3ei ISTR the original ERNIE did something along these lines.
@only_ohm Interesting keyword to feed to the duck! https://www.i-programmer.info/history/machines/6317-ernie-a-random-number-generator.html says ERNIE used neon tubes to generate noise. Their ADC was essentially one bit, then counting state changes of thay one bit occurring within a set time window.
To generate true random numbers, you could integrate some radioactive probe and a Geiger detector tube into a computer. Alternatively (and cheaper), amplify the noise produced by a avalange diode (Zener diode with more than 5.6 V) and pipe it into some ADC.
@zleap @ianrosewrites