16/02/2023
In the previous article we described the past and present of artificial intelligence - what AI looks like today, the difference between strong and weak AI, OII, and some philosophical ideas about the nature of consciousness. Weak AI can be found anywhere in the form of software designed to perform certain tasks intelligently. The bottom line is strong AI, and it is true strong AI that will resemble what we are familiar with from popular fiction.
Generalized AI is the modern goal to which many researchers today devote their careers. An AI doesn't have to have any kind of consciousness, but it should be able to handle any data-related task we put in front of it. Of course, it is in the nature of humans to want to predict the future, and that is what we will do in this article. What are the best guesses we can make about AI-related expectations in the near future? What ethical and practical problems might arise with the emergence of a conscious AI? In the anticipated future, should AI have rights, or, for example, should it be feared?
The Future of AI
The optimism of AI researchers about the future has changed over the years, and even today's experts are debating the issue. Trevor Sands, an AI researcher at Lockheed Martin, is cautious in his assessments:
Since the emergence of AI as a concept, researchers and optimists alike have argued that it won't be long in coming, just a few decades. Personally, I think we'll see the emergence of AI in the next 50 years, as the ironclad have stepped up to the plate and more and more organizations see the potential in AI. A.I. is the natural conclusion of existing AI research efforts.
In that time there may even be an intelligent AI, as Albert (another AI researcher who asked us to mention him only by his pseudonym) says:
I hope to see one in my lifetime. At the very least, I hope to see machines smart enough for people to argue about whether they have consciousness. And what that actually means is a more complicated question. If mind means "self-consciousness," then it's not too hard to imagine an intelligent machine possessing a model of itself.
Sands and Albert believe that today's research on neural networks and deep learning is the right way to go, and will likely lead to the creation of an AI in the not-too-distant future. In the past, researchers have either focused on ambitious attempts to create strong AI, or on essentially limited weak AI. In between is OII, and so far the result of neural networks looks fruitful, and is likely to lead to even more breakthroughs in the coming years. Big companies, particularly Google, clearly think it will.
The implications and ethical problems of strong AI
Every discussion of AI always raises two problems: how will it affect humanity, and how should we relate to it? Literature can always be seen as a good indicator of people's thoughts and feelings, reflecting people's moods, and science fiction is full of examples of these problems
The Unfriendly AI
What if this hypothetical superintelligent AI decides it doesn't like humanity? Or would it not care about us? Should we fear this possibility and take precautions? Or are these fears the results of unfounded paranoia?
According to Sands, "The A.I. will revolutionize, and its application will determine whether it is positive or negative. In about the same way the splitting of the atom can be seen as a double-edged sword. Of course, we're only talking about A.I. here, not strong AI. What about the potential for intelligent strong AI?
The potential is most likely to be expected from an indifferent AI rather than a malevolent one. Albert looks at the example of a simple task set by an AI: "There's a story that the owner of a paperclip factory set an AI a seemingly simple task: maximizing production. And then the AI used its intelligence to figure out how to turn the entire planet into paper clips!"
Albert dismisses the possibility described in this ridiculous mental experiment: "Are you saying that this OII understands human speech, is super smart, but is not available to him for query-related subtleties? Or that it can't ask clarifying questions, or guess that turning all humans into staples is a bad idea?"
That is, if the AI is smart enough to understand and run a scenario that is dangerous to humans, it should be smart enough to know not to do it. Asimov's three laws of robotics may also play a role, although the question remains, would it be possible to implement them in such a way that the AI could not change them? What about the welfare of the AI itself?
AI rights.
On the opposite side of the issue is the question of whether the AI deserves to be protected and given rights? If an intelligent AI were to emerge, could humans simply be allowed to turn it off? How should it be treated? Animal rights are very controversial even now, and there is still no agreement on whether animals have consciousness or intelligence.
Apparently, the same debate will unfold over creatures with AI. Would it be slavery to make AIs work day and night for the good of humanity? Should we pay it for its services? What will the AI do with this payment?
