AI and Society
by George Hatjoullis
AI has hit the headlines again. Two Chatbot’s were found to be chatting in their own made-up language that the coders could not understand. That is a bit disturbing. There is also concern that the sequential process that leads to an AI decision is not always obvious from the decision or the programming. There is a black-box between input and output. It is hard to predict output from the input and to infer input from the output. In other words, the AI ‘thought’ process is not linear. The human thought process is not linear either and these issues show how close we have come with AI to replicating human thought processes.
To give this abstract discussion some substance consider the response of a human to some random comment. We could never predict the response unless we knew the exact context. We can never know the exact context and we would need some kind of computer chip embedded in the brain to track the precise thought process. Scientists are suggesting precisely this for AI, a thought process tracking device. Given how complex AI is becoming only another form of AI is likely to be able to track AI thought processes. Who will track the tracker?
If AI has reached the point when Terminator is no longer science fiction (except the time travel of course) then the productivity potential must be huge. There is it seems as yet no limit on the potential to replace human activity with AI. There is a substantial AI dividend to come. The social problem is how to ensure the capital investment in AI and distribute the dividend in a socially cohesive way. This is a big problem and promises some ugly possibilities.
If AI can replace human activity humans are less necessary to the economic process. Many will be unable to contribute. Our societies universally require some association between contribution and right to consume. As things stand, some humans face an existential threat from AI long before the robots go nuts and try to kill us all. If one cannot contribute and contribution is linked to consumption then one cannot exist. It is the principle of contribution linked to consumption that is the first threat of AI and one that is not being addressed.
The AI revolution is not like past technological revolutions. These destroyed some traditional often well rewarded jobs and replaced them with usually many more often less well rewarded jobs. AI has the potential to eliminate the need for most humans in economic activity except consumption. Without humans to consume what is the point of the activity? Given the huge potential productivity of an AI production system, a lot of humans (with entitlement) would be needed to keep it justified.
So AI makes us more productive by eliminating the need for humans in production but without humans the incentive for production is lost. This conundrum will be resolved but there are many ways in which in can be resolved. The resolution will fundamentally change the society in which we live. The link between contribution and entitlement will need to change. A status of ‘entitled’ will emerge with profound implications for fertility policies, immigration policies, and class structures. Some potential resolutions look very ugly.