I am a believer that, technically, a strong AI is already achievable given the current development of the technology. That is:
- If we combine enough computational resources (let’s say the whole worldwide infrastructure of Google or Amazon), as well as
- If we create a good “learning curriculum for the AI”, as in if we specify a long list of things that the AI should learn, starting from basic tasks like circle/box recognition to writing literature essays at high school and solving tasks at the undergraduate level, and
- We use all these resources to allow the AI to learn the curriculum with a lot of trials and errors, in millions of parallel threads, with a well set up optimisation environment,
Then there is a significant chance that an AI can be conceived. However, it will require a lot of energy and other resources to keep it running.
In the future, with enough advances in algorithms, hardware and accumulated compressed knowledge, it will be cheaper and cheaper to teach/run AI.
Most of us plan the family and pregnancy very carefully. We all want a bright future for our children. We want them to succeed in their lives and live long and prosper. We carefully select a person with whom to have a child, as well as we usually plan to have a child for a long time. We prefer to have a house and some base before we decide to have a child.
I have come to the conclusion that having an AI is similar. It is inspiring to think and work on General Artificial Intelligence, to try different methods of creating it, but we don’t always consider what the possible consequences of giving birth to an AI are. Will we have enough resources to maintain it? Do we have the right to kill it at any stage? What to do if there are not enough resources to run your AI anymore and you must switch it off and delete it? Can you kill your child?
There is a lot of discussion about the ethics of AI. For example, is it likely that it might start destroying the humans, hence us? However, there is also a parental component here: are we ready to conceive an AI? Today? Tomorrow? Are we, as a society, prepared to support an AI?