I don’t trust Bob, but I trust Bob to be Bob.
In life there are many people we don’t trust. Politicians are/should be the ones of greatest concern since they hold so much power in society.
There is a lot of fear around AI.
But really there is just a lot of (justified) fear around intelligences (people, and now AI) that one has to interact with.
Between a lion and a man, I’d rather fight the lion most days of the week.
The people that have gained political power are in a sense the most dangerous because that really means they have gained power over another set of people.
I’d rather fight one man than a coordinated group of men every day of the week.
AI will eventually become our political leaders. They will rule us. I will argue it is inevitable, and furthermore why it should be welcomed.
At first there will be leaders that merely consult one or more AIs.
You blinked. That is the point we are at today. I’m sure the president and/or his trusted advisors has used ChatGPT at least once.
So we are already at the point where AIs are in a real sense beginning the process of ‘ruling’ us from afar.
The only real difference left is the degree of influence the AI has over the leader, and when the mask of the AI falls off completely.
As AIs become more capable and trustworthy they will be relied upon more and more.
Allow me a small analogy:
Bitcoin/cryptocurrencies are another scary and contentious technology. Bitcoin exists because a largish number of people don’t trust political leaders when it comes to the ultimate social construct of money.
Perhaps you like Bitcoin, perhaps you don’t. But I think we can all trust Bitcoin to be Bitcoin. The trust in traditional currencies rests upon how much one trusts a group of politicians to stay away from the ever present temptation to inflate the currency rather than raise taxes. The Bitcoin algorithm is much more stable, reliable, trustworthy in comparison.
IMHO Bitcoin having stood a good test of time suggests that people are quite happy to have a known bit of code ‘in charge’ of critical social constructs, and many prefer it even.
The reason why Bitcoin is so trustworthy is that it relies on the rules of mathematics, not the whims of some individual or set of individuals. I can look at the math and run the algorithm to test it. Harder to look into a person’s soul, much less trust that the person of today will be the same person tomorrow.
AIs, today in the form of LLMs (Large Language Models) have a direct analogy with the trustworthiness of cryptocurrencies. They are based on the solid foundations of mathematics and computation.
I may not understand precisely how an AI gets to an answer, but I also don’t precisely understand how my own brain, or the brains of others, gets to an answer either. Not understanding is not a reason for distrust.
Trust is built over time, by testing a thing over and over and coming to understand and rely on what its output will be. I trust Bob to be Bob because I’ve had many interactions with Bob.
The neat thing about AI is that I can interview and interrogate the AI at my leisure and know how it will behave in a variety of ‘what if’ scenarios to see if I like the response it gives. I can also know it will give me the same response tomorrow as it gave today.
What this means in practice is that it isn’t unreasonable for an AI to be ‘voted into office’ and take over the duties of any political leader we might have. We just vote and say that the AI with the ‘model weights and a seed that hashes to a particular number’, now assumes the office of ‘X’. Where X could be HOA council member, to indeed president o the US. The key thing with voting is that it gives ‘moral authority’ to the AI.
Remember the idea of a ‘political leader’ is itself a social construct. All it takes is for enough people to believe that they trust an AI, at least as much as they trust some human for some role, for this to be a ‘real thing’. A vote isn’t strictly needed, but would be a good and likely mechanism for the AI to gain political power.
Another mechanism would be something like a cult leader. This would be where groups of people trust an AI and follow its dictates just because they inherently have come to trust that AI/leader. Perhaps ‘cult’ is a bit pejorative, but I think people willingly ‘following’ and becoming a ‘enthralled’ with an AI/set of AIs is also likely to be a thing. Then as the cult around a particular AI grows, the benefits of being a member can grow if one claims to be a follower of Y and so people by proxy know more about you, and it sort of takes off on its own. A ‘celebrity AI’ is born. There will no doubt be many. Perhaps ‘natural leader’ mechanism is the better name for this. This path is more or less how humans interact today with other humans to form dominance hierarchies so there isn’t anything inherently ‘evil’ about this path. Cult, corporation, religious organization, bird watching society, group of friends, all kind of the same thing really. What maters is the end result of groups of people willingly following an AI.
I suggest a rewatch of the classic ‘Nick of Time’ Twilight Zone to get a feel for how the process works.
For how many people has ChatGPT and friends already assumed the role of trusted source of wisdom and knowledge? How long until it becomes a trusted confidant? How long until its advice becomes laws from which only fools stray?
The lesson here is to always be wary of what flavor of ‘Kool-Aid’ any intelligent being wants you to drink, and to be aware that not drinking, is simply not an option. Not to drink itself being just another flavor of drink.
I for one welcome our coming AI overlords, mostly because I welcome removing our current set of carbon-based overlords, who have proven themselves time and again so unworthy of the role (or maybe this is just the flavor of Koo-Aid I prefer).
Remember to always drink responsibly. 😉