A thousand man years in a weekend: How the power of iteration may make artificial intelligence unbeatable
Technologist and real-world Tony Stark inspiration Elon Musk recently donated $10 million dollars to the Future of Life Institute to run a global research program focusing on Artificial Intelligence, specifically “keeping AI beneficial to humanity”.
But does Artificial Intelligence pose a genuine threat? There is a lot of hyperbole around AI, and in science fiction rebellious or malevolent AI is practically a whole sub-genre in and of itself. And while it’s easy to laugh off ridiculous scenarios like killer robots, the real threat that AI presents may not be to life and limb, but to economic stability and the unscrupulous wielding of artificial intelligence to disrupt markets and outpace human-driven innovation.
Avoiding the pitfalls of AI
Elon Musk was one among a number of prominent co-signers of an open letter that was also published by the Future of Life Institute. In it, they note:
“There is now a broad consensus that AI research is progressing steadily, and that its impact on society is likely to increase. The potential benefits are huge, since everything that civilization has to offer is a product of human intelligence; we cannot predict what we might achieve when this intelligence is magnified by the tools AI may provide, but the eradication of disease and poverty are not unfathomable. Because of the great potential of AI, it is important to research how to reap its benefits while avoiding potential pitfalls.”
While it is the prospect of human-like AI taking over that has caught popular imagination, even a functionally limited AI could hand its human masters considerable economic power.
Iterating to excellence
Consider a scenario where a specialist AI is created to create microchips within a simulation: each cycle, the AI creates new iterations of the individual circuits, gates and other components. It does this the same way that human engineers do – by testing new configurations with brute-force scenarios. Improvements in efficiency are kept and integrated into the next “generation” of the design, and the process begins anew.
But while it takes a human team of engineers a month to create and test a new design, the AI does the same task in a matter of minutes. And even if it does the job very badly compared to humans – say creating only a .001% improvement to processing speed with each iteration – the fact that it can run this simulation thousands of times a day means that it can purely brute-force it’s way through the task until it has achieved the maximum speed physically possible on the hardware.
The AI’s owners then release a new microprocessor on the market that is, effectively, dozens of generations ahead of what would have been created through gradual human iteration. It’s ten times faster and ten times cheaper than anything else on the market. Instantly, every human-designed chip is obsolete. The AI has effectively reached into the future and extrapolated a design that wouldn’t normally have appeared on the market for decades, and at a stroke the competition is wiped out.
Will humans be marginalized?
This is the potential threat that AI poses – not some fantastic scenario of man versus machine, but as a tool of economic disruption – a means for the rich to create an insurmountable advantage over the rest of the world.
Unchecked artificial intelligence could result in a permanent underclass between those with access to AI and those that don’t – a new 1% that don’t just control the majority of the world’s wealth, but its intellectual capital as well. When AI learns to practice skills previously reserved for human experts, the threat is that human expertise will be increasingly marginalized.