It’s not going to deliver.
The problem as I see is defined by the limitation inherent to all forms of mathematical modelling. Every mathematical system relies on things outside of itself in order to be a logically consistent system, and it cannot demonstrate it’s own efficacy without reference to something outside the system.
To put it in very superficial terms take the following equation:
1 + 1 = 2
The equation cannot possibly be used to demonstrate the the numbers 1 and 2 really exist or that they obey a logical relationship, it is assumed before the equation is written and are outside of the tiny mathematical model presented here.
So what does this have to do with AI?
Let’s look for the things that lie beyond the scope of the model on which AI depends to even be an AI:
- The data on which it is trained
- Prompting given by human users
- Verification of results
All of these things obviously lie beyond the AI model and yet the model is totally dependent on them to usefully exist. It follows therefore that no AI can ever be truly autonomous and therefore intelligent as we would understand it. I would go even further and say that any AI system used in the real world will require constant maintenance and training and can therefore never replace human workers in any sustainable manner. The AI will always be intrinsically backward looking, it will at best be only able to respond to yesterday’s crisis. In areas where AI is adopted in such a way as to reduce human input and autonomy there will only be stagnation.
There will neither be mass-unemployment nor luxury space communism.
What role do I see for AI?
So far as I can see the true power of AI is in machine learning, it can be used to program computers to operate within complex but closed systems.
My personal favourite example is machine learning and animation:
https://medium.com/embarkstudios/transforming-animation-with-machine-learning-27ac694590c