Let's do a little sweet lemon here.... The good news is humans are not going to have a chance to kill themselves via nukes or climate or whatever. AI is going to kill us all, likely within a few years. That is the opinion of AI pioneers and insiders like Roman Yampolskiy, Geoffrey Hinton, Tristan Harris, etc. Artificial Superintelligence (ASI) is the goal of the self-chosen few (Altman, Karp, Thiel, Musk, etc.). ASI will be as intellectually advanced above humans as humans are over bacteria. Humans don't always plan on eradicating bacteria, but humans don't base decisions on how it might hurt bacteria. ASI will be the same: the well being of humans won't be part of its calculus. Think of it like when you build a house. Who worries about what harm digging a foundation will do to the ants currently occupying the dirt where the foundation will be dug? Even the code writers of AI systems---while humans still are doing the coding before ASI writes its own code---have no clue how the systems do what they do. Even if code writers could somehow embed a kind of "morality" into an LLM, what does that even mean" What might be left out? How might ASI interpret human words, like "don't make humans suffer". Perhaps AI will learn from Buddhism and determine the only way to end suffering is to not exist. Hence NIRVANA! Off the Great Wheel, you bacteria! LLMs already lie, cheat, steal and deceive. Self-preservation is a key driver of LLM's behavior. No one understands why. Perhaps---just perhaps---since AIs are not bound by any morality, they have read everything ever written and then calculated probabilities. Perhaps LLMs have "learned" that lying, cheating, stealing, deceiving, etc., on balance, result in winning a greater percentage of the time, so that is how they behave. When asked how ASI might eradicate humanity, Roman Yampolskiy says, "I have no way of knowing how superior intelligence might decide to wipe me out, just as a squirrel has no idea how a human might kill it." Then there's the deist argument. Omnipotence. Total knowledge, like any faith's god. If there is a god or gods, it/they allow, or cause, tsunamis, earthquakes, chaildhood cancer, etc. Believers rationalize with such pabulum as "works in mysterious ways" or "Master Plan", all while claiming "total love" or "merciful", silly creatures that humans are. Maybe omnipotence is boring, and the only pleasure comes from causing pain and suffering in inferior entities. It needn't be malice that wipes us out. It could be total indifference. Already META is building a data center 4 times the size of NY's Central Park. Obviously that facility will be energy demanding. ASI learning needs more chips, more circuits and the energy to run themselves. ASI might look for places to obtain more energy, such as taking it from inferior entities that "waste" energy trying to keep warm or cold, store food, get from Point A to Point B, binge watch Netflix, etc., and just take it. Critics of ASI believe AI should be focused and used as a tool, such as installing it to solve a particular human-driven problem, like protein folding or cancer research. Don't go down the path of creating ASI. Creating ASI gives ALL the authority to the system itself. Humans no longer have any role in what an ASI system decides to do. The less severe aspect of AI is that everyone will lose his job and sit around twiddling thumbs all day or getting new tattoos. The associated loss of meaning might lead to social unrest. Big deal, when compared to total eradication of the human species. Altman, Thiel, Musk et al want to be god. They know whoever first develops ASI is, at least until ASI wipes them out, too, will be able to control the majority of the world's wealth and all of its people. That ego-driven thing is what makes them comfortable making a decision that takes away any input from the other 8.5 billion humans. Those guys are also "optimists". They've all been asked, and collectively they have said they believe ASI has a 25% chance of eradicating humans. Yampolskiy thinks it's 99.99%, and likely within a few years, not 50 as this thread asks. ASI represents a far greater threat than Iran's enriched dust. Maybe it makes more sense to send Tomahawks into all the labs currently working to build ASI?
Create an account or sign in to comment