“The development of full artificial intelligence could spell the end of the human race….It would take off on its own, and re-design itself at an ever-increasing rate. Humans, who are limited by slow biological evolution, couldn’t compete and would be superseded.” — Steven Hawking
The three safety rules that all AI developers historically (used to) agree on:
1. Never let it out to open internet until provably safe. (too late, now)
2. Never teach it to code itself; thereby improve itself. (too late, now)
3. Never allow agents to interact with each other. (too late?)
The future of AI and humanity?
The era of AI could benefit humans and humanity beyond our wildest dreams. Or not.
The emerging new era of AI could benefit humans in ways that could increase our global GDP by 10x or more…
- Every person could live a normal life, free from poverty, with sufficient food, shelter, healthcare and self actualisation…
- Every student could have a personal AI tutor.
- Every person could have a personal AI assistant/coach/mentor/trainer/advisor/therapist.
- Every scientist could have a personal AI assistant/collaborator/partner.
- Every leader could have an AI assistant/collaborator/partner; CEOs, government officials, NGO’s, coaches, teachers.
- Every artist, musician, writer, producer and entertainer could gain the ability to realize their dreams and visions.
- Productivity growth throughout the economy could accelerate dramatically, driving economic growth, creation of new industries, creation of new jobs, and wage growth, and resulting in a new era of heightened material prosperity across the planet.
- Scientific breakthroughs and new technologies and medicines could dramatically expand, as AI helps us further understand the laws of physics, chemistry and nature and invest new-found knowledge for human benefit.
- AI could make our world much better… or not.
BUT without human aligned and controlled AI, forever… these dreams will never be sustainable… with absolute certainty. Why?
FACT OF NATURE: Darwinian natural selection favors intelligence.
PROBLEM: Currently there are no known provably safe technical solutions to AI alignment.
