Vitalik Buterin, Ethereum co-founder, has released an article where he states his opinion on the direction we should take regarding the AI debate. In the article, the author presents two views currently dominating the future of AI: one view calls for the cessation or slowdown of AI developments (The dystopian view). The other view calls for continued acceleration of AI research and development (The Utopian view).
Vitalik presents a hybrid model that looks at how we can safeguard ourselves from AI turning rogue while tapping into the benefits that come with AI adoption. He calls this the defensive acceleration view.
AI is a new type of Mind
The author starts by stating that while we may have made rapid progress in the last 2000 years, there hasn’t been an innovation like AI. Granted, we may have experienced other innovations like the wheel, technological innovations, and even industrial revolutions. However, the advancement of AI is unlike anything we have ever seen.
The author argues that technology is inherently good and has led to an increase in human lifespan, facilitated a reduction in global poverty, enhanced interconnectedness, and generally increased our quality of life. At the current rate, the author argues that technology will make it possible to end aging and diseases.
However, even with all these benefits, the advent of AI cannot just be dismissed as another technological disruption. AI brings with it radical changes in the way that humanity will interact with technology. “AI is a new type of mind that is radically gaining intelligence, and it stands a good chance of overtaking human faculties,” he argues.
AI goes Rogue
Vitalik then goes on to paint a picture where AI has gone rogue. He states that if this happens, then humans will not be safe, even if we went to Mars. He states that a super intelligent AI will definitely end humanity if it decides to turn against us. To back his proof, he states that a recent survey on ML researchers shows that 5-10% of those surveyed believe that AI has the potential to end humanity as we know it. There is also a possibility that we may get into a future where AIs view us as pets. This is because what a human can do, an AI will probably do it better. He further states that we may get into a world where totalitarian regimes leverage AI and use it to advance their rules, effectively killing any form of resistance from the masses.
Defensive Acceleration
Vitalik proposes an AI acceleration model where humans are at the centre of all decisions. Though he is not particularly optimistic about this, he states that this model is, at the moment, the best way we can tame AI. This model includes decentralizing the decision-making process of AI advancement so that we do not have a few entities retaining too much control over AI advancement. He also suggests a world that involves a human-computer brain interface. This basically entails merging human brains with AI to take advantage of the benefits that AI brings. While he accepts that this path is worrisome, he insists that governance on such a path should be decentralized and led by a security -focused open-source movement.
He concludes by saying that if humanity wants a future where we will not be viewed as pets, then merging AI with human brains is the least undesirable path to take.