Will Strong AI Development Weaken or Bolster Humanity?
Daniel Faggella
2016-03-14 00:00:00

Today, as developments in AI progress at lightning speed, there are many scientists expressing the opinion that a “real AI” is within reach (though as to when that might occur is anyone’s guess). In a recent TechEmergence interview, Cognitive Scientist Joscha Bach noted that the artificial intelligence applications we have today were almost unthinkable half a decade ago, and that the trajectory toward development of strong AI is likely to continue forward, for better or worse.



Better hardware, software, computing power, and the consistent work of many researchers have made this growth possible. In addition, AI research has seen more overall growth - in funding, development, and hiring - in the last five years than in the entirety of its previous history.



“Most of this funding doesn't come from academic sources, but it comes from companies like Google. This makes it possible to make tremendous progress that wasn't possible before,” Bach said. “Some people have gotten extremely wealthy building AI companies that leverage AI technologies. In some sense, the biggest change that we've had in the last five years is cultural, and it’s an important one.”



As all that development accelerates the growth of strong AI, Bach sees an eventual slowdown on the horizon, as researchers continue to work through through the concepts and details of the human mind as an information processing system and how those principles could be translated into an AI.





Only a small subset of our genome encodes for our nervous systems, and the genome itself could technically fit on a CD-ROM. Joscha believes that developing an understanding of the principles behind the complexity of getting the mind to work is not so big, relatively speaking; however, reverse engineering this functionality is extraordinarily difficult.



“We don't know how unknown the rest of the way will be, how far the current insights will push us and how easy it will be to get to the next wave of insight. It’s very hard to say how long it will take,” Bach said. “As a software developer, I know that this is all a pipe dream. If I don't have the specification, I don't know when it’s done. It can be any order of magnitude.”



Bach also doesn’t rule out the possibility of a so-called “silver bullet” discovery of unified principles that could speed the process of building a human-like mind. While he’s optimistic about the development of strong AI in many of our lifetimes, the gap between where we are now and the end-point is still vast.



Looking at the current state of strong AI development, Bach recognizes that researchers are still working through a number of issues in reinforcement learning. While we can model the repetition system to emulate how the mind recognizes images, sounds, colors and more with respect to actions and events, we’re still in the early stages of exploring the feature detection process, for example, in the brain.



“One of the questions is, how much do we have to put into the machine in the first place? How much do we have to hard-wire to make the machine do all these things? How can we understand what it is and how can we get it to work? It’s related to the question (of) how much is hard-wired in our brains,” says Bach.



Though strong AI development may still be in its early stages, Bach shares the concerns of Elon Musk and others about the potential perils that may lie ahead; however, he also believes that warning people about the dangers of artificial intelligence won’t stop its development. Instead, the development and funding of AI research will only cease when people and corporations think it is no longer possible, and the clear opposite has already taken root and sprouted; there are more and more AI startups in the financial and other industries growing into reality every day.



“AI is going to come from the top as an extension of business intelligence. It’s dangerous if we make corporations more powerful, because sometimes corporations do things that are not in the best interest of humans,” Bach said. “Right now (corporations) don't care about externalities. They do care about the parts that make it more efficient and help it survive. The difficulty in making a world that is safe for AI is very much aligned with a world in which the different components of our society and economy (are) aligned with the needs of humans.”