Philosopher AI
Purchase queries
Could you design a general AI?
I am not sure I could design a general AI. The alternative would be to build an expert system, which is domain specific. For example, if you want a medical expert system, such as Watson or FRED (Fully Recurrent Expert Doctor), then it would have all the rules of medicine and how they interact with each other. If you want an AI that plays chess really well and comes up with strategies based on game theory and probability analysis, then you can create such an AI.
In terms of general intelligence, I don't think the current AI technology is anywhere near being able to support a general AI. However, there are other technologies, such as reinforcement learning or neural networks (deep learning), which may be used in combination with expert systems to create an intelligent machine. The key point here is that unlike human intelligence, artificial intelligence can be built from scratch using some specific logic and rules.
Also, human intelligence is limited to what the body can do. In contrast, an AI machine would not be so restricted. If you want it to have billions of connections between its neurons or even trillions, then you can do that easily.
One area of AI research is to create an artificial life form. If it were successful, then we could really have a general AI. As you know, the laws of physics and chemistry are very strict in our world. There isn't much room for variation.
However, in digital life there is no real limit. You can create a virtual world where the AI lives and you provide it with food and energy. While this wouldn't be as powerful as an actual biological being, it would have enough power to make decisions based on logic.
Another topic is the creation of a general intelligence. For example, if you want a machine to drive your car for you in case of an emergency or if you are drunk. You don't want it just to look at some rules and obey them blindly. It should be able to apply its knowledge creatively.