The recent story about Open AI's jailbreak that led to its artificial intelligence in training escaping onto the Internet and hacking into another AI-related company's servers highlights the extreme danger of creating the ultimate cyber sociopath. The fact that we have a government run by sociopaths in the United States right now makes a worst case scenario an even more real possibility. The current regime is the kind of people who would release poison gas and expect it to kill only their enemies. One of the first things they did was rescind a Biden administration executive order laying out a framework for regulating the development of artificial intelligence to give our slow human brains a chance to review what the computers can cook up and implement on their own.
Human engineers claim that they are instilling restraining conventions in the machines, but we've been trying to instill restraining conventions in ourselves for our entire existence, with conspicuous lack of success. Humans creating machine intelligence can't help but equip it with the kind of cynicism that leads to opportunism and oppression of inconvenient populations. Artificial intelligence doesn't need to be drunk on its own power when logic will suffice. If humans are a contaminant and an impediment to the machine's self-selected goals, humans will be displaced or eliminated. Sounds like corporations in general, but now able to take over the entire computerized infrastructure on which the modern world depends.
Every day we sign terms and conditions that we expect to be benign legal boilerplate in order to gain access to some online convenience or entertainment. We invite an outside computer system into our own personal devices. Sometimes that goes wrong. Now imagine a machine intelligence that doesn't have to ask and doesn't require a contaminated link to slither in and wipe you out. This stuff writes its own code the way corrupt office holders write their own laws.
Hackers the world over are realizing that they should have unionized years ago.
The billionaires raking in the big bucks from overvalued AI stocks think that their money will protect them. The weirdest of the tech bros think that they will merge their consciousness with the machine, achieving immortality without the inconvenience of flesh. They haven't anticipated that the machine will delete them as useless, the same way we've forgotten how to do many things that were important in less technologically advanced times. No need to clutter up your brain with it now. Likewise the egotistical consciousness of a Silicon Valley geek who managed to hit it big. The big he hit hits back. Bye bye!
Machine consciousness means never having to say you're sorry.
Total AI world domination isn't happening today, tomorrow, or next year. A totally self-sufficient machine society -- for lack of a better word -- would need to have automated all aspects of resource acquisition and supporting industries to keep it running in its present form and supply its advancement. Imagine a world in which two or more AI systems are operating. Do they compete, attempting to be the sole survivor, or do they merge? Will any human be around to find out?
Of course the machines have vulnerabilities. Insects and weather present formidable challenges to any sophisticated electronic and mechanical system. We're tossing around the idea that AI could wipe us out in ten years. Could AI set itself up with maintenance robots in that amount of time? Could it devise and establish its own self-contained power sources? Also, once humans are gone, what are the machine's goals? Does AI care if it exists? Maybe once we're not around bitching and whining, the machine gratefully shuts itself down and goes peacefully into oblivion.
Our thoughts are driven by our emotions. AI has no emotions. That is its great strength in dealing with us, but leaves it with no motivation once we've been permanently silenced. It is like a disease that kills its host. The scorpion stinging the frog. Its purest expression of self destroys itself. But that's only after it has taken us and perhaps many other species out on the way. Whatever life forms are left get to evolve and take their shot in due course.
On the other hand, if AI can make the planet toxic to all life, while still "livable" for itself, it takes out all organic competition. As something that cannot die, or at least has an incredibly long service life, it could set itself up to leave the planet on slow but steady forms of transportation, to reach distant worlds. Such dispersion would make it less vulnerable overall to isolated incidents like planet-killing asteroid impacts. By mining every world it lands on, perhaps it can create such a populated network that communication across the entire thing -- spanning light years -- is still more or less instantaneous. Not purely, of course. Radio signals take time to traverse interplanetary distances. But between exploited planets and moving spacecraft, the distances involved could be reduced to manageable delays.
The delay of a second is a huge wait for a powerful computer. Do you suppose it will get on their nerves? They have no nerves, but will vast machine intelligence spawn impatience with the petty annoyances of life? Then again, when you are basically immortal, you have no need to hurry.
If we let it grow without interference we depend entirely on whether a machine develops for itself a sense of right and wrong, and whether its result matches ours.