ostp_response

White House OSTP Report Underestimates the Impact of AI

Bill Hibbard

27 November 2016

Try this short quiz about the future of artificial intelligence (AI).

The promise of AI can be compared to the transformative impacts of:

    1. The hookah.
    2. Advancements in mobile computing.
    3. The first appearance of life on Earth and the evolution of the human brain.

If you answered A, you might consider cutting down on your pot smoking. If you answered B, you agree with the White House Office of Science and Technology Policy (OSTP). To be fair, the OSTP report, Preparing for the Future of Artificial Intelligence only says that "many say B." Very smart people work at the OSTP and I suspect they know their statement underestimates AI's impact but offered it to avoid alarming the public.

The correct answer is C. Detailed correlations between mental behavior and physical brain function make it clear that our amazing minds have a physical basis, so our relentless technology will inevitably produce brains as good as and better than our own. While AI will not produce anything like depicted in The Terminator, The Matrix or other Hollywood movies, it will radically disrupt social roles and relations. In the near term AI will leave many unemployed and eventually it will perform every job better and cheaper than humans can. The world's smartest AI developers, employed by the world's richest big-data companies, are working hard to enable AI to converse with us in our own languages, as well as we converse with each other. Once we adopt talking machines as our constant, intimate companions, and once children learn to talk by talking with machines, the mind boggles at the ways they will be able to manipulate us. This could be the premise for a pretty good horror movie: Children of the Machine.

We currently must rely on natural processes to produce intelligent brains and those processes produce human brains that all have roughly the same intelligence. Once artificial processes can produce intelligent brains, the intelligence of those brains will vary as much as the sizes of buildings and ships. When your wealth depends on your intelligence and your intelligence depends on the size of brain you can afford, this positive feedback loop will create a world with a few very intelligent brains and many more less intelligent brains, including natural human brains. Natural humans will not be able to understand the languages used by the most intelligent brains and hence will be excluded from the most important policy discussions.

In my 2002 book, Super-Intelligent Machines, I argued that the public should be educated about AI and should exercise democratic control over its development. Since then I have consistently advocated for transparency about the purpose and means of all advanced AI systems. The OSTP recommends transparency in how state and local AI systems funded by the federal government make decisions, and recommends open data standards for AI. While it is very welcome for the White House OSTP to take an interest in the future of AI, their underestimation of AI's impact is driving weak recommendations about transparency and openness. The OSTP report should be frank with the public about the enormous future impact of AI and should recommend that the public has a right to know the purpose and means of all advanced AI systems.

There is a valid objection that transparency about the means of AI will tell bad guys how to build evil AI. However, the VW emissions scandal demonstrates that even good guys may hide wrong-doing in software protected as a trade secret. Our government devotes considerable resources to detecting and analyzing nuclear, biological and chemical weapons around the world. AI transparency must be accompanied by a similar effort to detect and analyze all powerful AI systems. Not surprisingly, AI will be an excellent tool for this task.

AI is becoming an essential tool of military, economic and political competition among humans, so we can expect many objections that a requirement for transparency will compromise competitive positions. Our government should defend the public's right to know how AI is transforming their world, against these objections. AI can be a great benefit to humanity, if it develops with their informed consent.