Hey Siri!
We met some time ago. I had to train myself not to say please and thank you, even though you were trained to say, you're welcome.
Last week I was at your birthplace, SRI, in Menlo Park California. Originally the Stanford Research Institute, One of those innovations was you, Siri! You were birthed as an app in 2010. You're a teenager.
Last week, SRI provided the setting for the Post-industrial Institute's Summit on Autonomous Organizations. These are organizations run by autonomous AI agents that can reason, learn, and act independently. The Post-industrial Institute, founded by Frode Odegard, focuses on research and leadership education for a society where mass production is being replaced by decentralized services powered by AI.
This could predict a society where knowledge is a form of capital and producing knowledge rather than cars or corn is the way to grow the economy.
Hey Siri!
I can't eat knowledge. Isn't corn vital?
In this understanding, corn, apple, meat, are ideas. Those ideas become products that provide nutrition to humans. Planting, growing, harvesting, extracting food products, delivering those to consumers, aren't done by humans. Those processes can be performed by various types of agents, some of which drive plows and trucks.
Understanding how to create the most value from that, is how knowledge turns into food.
Planting, growing, harvesting, extracting food products, delivering those to consumers increasingly appear to be things that autonomous agents can manage.
Two distinct differences emerge between AI agents and humans. Agents are extremely good at probability. It's why machines started beating humans in chess decades ago. Machine agents can process massive amounts of data rapidly and find the patterns of probability. They learn by engaging in other probabilistic situations, and in probability, the more samples, the more accurate the estimation of likelihood becomes.
Hey Siri!
When did Deep Blue beat Garry Kasparov?
Humans are bad at that. Really bad. We can't process enough information fast enough to calculate probability accurately. We process at an average of 10 bits/second of information. By contrast the fiber optic cable leading to my laptop can process a billion bits per second.
We make decisions, not by accurately processing the data of reality, but by making rapid judgments based on rough approximations of what reality is. We use biases, reflexes, habits, prejudices, propensities, appetites.
Nonetheless, we have some distinct advantages alongside the sequential decision-making algorithms generated by agents.
I asked the group if they have any evidence or awareness that machine agents have fear or greed? Do the machine agents have any stake in their own existence?
No answer yet.
That no answer points, however, to why humans still have advantages: we have motivations. Not just we humans, but even single-celled life forms move toward favorable conditions and away from unfavorable ones.
Motivation: fear, greed, a stake in our own existence. The ability to keep ourselves alive turns out to be a prodigious aptitude.
More on humans and machines next week: the switch in our brains that causes us to recognize a human.
Warm regards,
Francis Sopper