LONG KNOWN

The last few weeks folks have eagerly asserted that we are in unanticipated and unprecedented times. I couldn't disagree more. Western civilization has been pondering and worrying about intelligent machines for all time and with evermore detail and precision.

Homer and Aristotle talked about machines with intelligence and reasoning and considered those replacing human labour (in that time largely slave labour). Descartes grappled with the idea that machines could use words and the inherent trouble then with distinguishing a talking machine from a human and of course his philosophical troubles with matter, mind, and creator. Samuel Butler’s 1863 essay “Darwin among the Machines” talks about mechanical devices being a new kingdom of life in which development becomes vastly faster than evolution and as such machines being “man’s next successor in the supremacy of the earth.” Butler contemplates humans having raised a race they cannot destroy and humans becoming enslaved to them.

In the 20th century, many considered the problems posed by thinking machines. Turing published a paper in 1950 titled "Computing Machinery and Intelligence" in the journal Mind. He talks about whether machines could produce original material or become conscious and if they must only respond to human input and do what we tell them. He contends that machines would not die, could talk to one another and self-improve in this way, sharpen their wits until they are more intelligent than us. From there, he warns, we could expect the machines would then take control...

In "Some Moral and Technical Consequences of Automation" (Science, 1960), Wiener argued that learning machines can invent strategies their programmers did not foresee and could fail to do what humans intend. He noticed that just the rapidity and precision of machine computation threatens human comprehension and response, realizes that human action is merely reactive, and foresees disastrous consequences as inevitable “whenever two agencies essentially foreign to each other are coupled in the attempt to achieve a common purpose.”

Good anticipates our own time in his "Speculations Concerning the First Ultraintelligent Machine" (1965). He speculates that “The machine will have a multimillion dollar computer and information-retrieval system under its direct control. The design of the machine will be partly suggested by analogy with several aspects of the human brain and intellect. In particular, the machine will have high linguistic ability and will be able to operate with the meanings of propositions, because to do so will lead to a necessary economy, just as it does in man.” He doesn’t just imagine the neural net origins of AI and Large Language Models but also the recently added layer of reinforcement learning. He tells us “The first ultraintelligent machine will be educated partly by means of positive and negative reinforcement.” Good writes about the ultraintelligent machine that surpasses every human intellectual activity that can design a better machine, resulting in an intelligence explosion. The first such machine is, he says, “the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.”

Of course, in the last quarter century these same thoughts have distilled and crystalized into proposals and problems incorporating the specifics of programming languages, the internet, vast hoards of interconnected graphics processing units, and the like.