• 31 Posts
  • 249 Comments
Joined 1 year ago
cake
Cake day: September 13th, 2024

help-circle



  • Short answer: No one today can know with any amount of certainty because we’re nowhere close to developing anything resembling “AI” in the movies. Today’s generative AI is so far from artificial general intelligence it would be like asking someone from the middle ages when the only form of remote communication was letters and messengers, whether social media will ruin society.

    Long answer:

    First we have to define what “AI” is. The current zeitgeist meaning of “AI” refers to LLMs, image generators, and other generative AI, which is nowhere close to anything resembling real consciousness and therefore can be neither evil nor good. It can certainly do evil things, but only at the direction of evil humans, who are the conscious beings in control. Same as any other tool we’ve invented.

    However, generative AI is just one class of neural network, and neural networks as a whole was once the colloquial definition of “AI” before ChatGPT. There have been simpler, single purpose neural networks before it, and there will certainly be even more complex neural networks after it. Neural networks are modeled after animal brains: nodes are analogous to neurons which either fully fire or doesn’t fire at all depending on input from the neurons it’s connected to, connections between nodes are analogous to connections between axons and dendrites, and neurons can up or down regulate input from different neurons similar to the weights applied to neural networks. Obviously, real nerve cells are much more complex than the simple mathematical representations of neural networks, but neural networks do show similar traits to networks of neurons in a brain, so it’s not inconceivable that in the future, we could potentially develop a neural network as or more complex than a human brain, at which point it could start exhibiting traits that are suggestive of consciousness.

    This brings us to the movie definition of “AI,” which is generally “conscious” AI as or more intelligent than a human. A being with an internal worldview, independent thoughts and opinions, and an awareness of itself in relation to the world, currently traits only brains are capable of, and when concepts like “good” or “evil” can maybe start to be applicable. Again, just because neural networks are modeled after animal brains doesn’t prove it can emulate a brain as complex as humans have, but we also can’t prove it definitely won’t be able to with enough technical advancement. So the most we can say right now is that it’s not inconceivable, and if we do ever develop consciousness in our AI, we might not even know until much later because consciousness is difficult to assess.

    The scary part about a hypothetical artificial general intelligence is that once it exists, it can rapidly gain intelligence at a rate orders of magnitude faster than the evolution of intelligence in animals. Once it starts doing its own AI research and creating the next generation of AI, it will become uncontrollable by humanity. What happens after or whether we’ll even get close to this is impossible to know.


  • Don’t upload photos/videos/voice recordings of yourself to the internet, ever. This alone won’t guarantee your likeness and voice can’t be replicated with AI since there could always be photos/videos/voice recordings of you out there that you’re not aware of (from a data leak, etc) but it will significantly reduce your attack surface. Unless you’re rich or otherwise high profile, scammers are most likely not targeting you specifically but just scraping the internet for training data and picking targets based on who they have the most video/audio of and can therefore produce the most convincing AI fakes of. Or buying from data brokers who have scraped the internet for them. I could be wrong, but I doubt there are many scammers going to the effort of buying/stealing, say, call recordings from your phone company or virtual meeting provider to scam some random person without much wealth.

    Oh, and never link to your IRL friends and family online. Never add them on Facebook (just never use Facebook tbh), or Discord, or anywhere else that’s not E2EE. Scammers can’t target your grandparents if they have no idea who they even are.