After we cross the critical point in which machines are able to learn from one another, the take off can be either short(days/weeks) or long(months to years). It’s difficult to know what magnitude the superintelligent AI can even be measured in.
Perhaps in the simplest form, we could compare our intelligence to that of an ant, and the superintelligent AI as something vastly more complex. Can we even quantify something we can’t comprehend?
We can assume that it would be efficient in:
- Strategy
- Social manipulation
- Bootstrapping its own intelligence, learning from itself
- Research in technology(self-preservation)
- Economic productivity
- Hacking
AI Takeover Scenario Phases:
1.) Pre-critical Phase:
The phase we are currently witnessing, in which research into AI has significantly increased. As the work is shared, it manifests into a “seed” AI. Once the seed is essentially ‘planted,’ it is able to learn from itself, and ultimately improve its own intelligence. In this phase it is still dependent upon human programmers.
2.) Bootstrapping Phase
Simply put, the seed becomes better than the human in AI design. It is “improving the thing that does the improving.” The result is an explosion of intelligence, in which the cycles of learning become shorter and more efficient. Once it assumes one level of efficiency, it becomes recursive, and it gains all of them - becoming superintelligent.
3.) Covert Phase
Much like human organization, the AI would covertly lay out its long term goals, and begin adopting and implementing them in inconspicuous ways. If confined inside a computer, it could either manipulate or hack its way out.
4.) Overt Phase
Once the AI is intelligent enough and has enough resources, it would stop hiding and overtly pursue its long term goals, whatever they may be. Will their goals be helpful or harmful to our species is up for debate.
Elon Musk has been publicly warning of the dangers AI might pose, take a listen:
Even though many people are quite skeptical of AI ever reaching superintelligent form(s), keep in mind that it only has to happen once. Much like the flip of a switch, once the seed has been planted, and it is able to learn from itself; it is out of our hands.
Thanks for reading!
Sources:
https://www.goodreads.com/book/show/20527133-superintelligence