Monthly Archives: August 2017

This sweet little song was written by AI

Like frogs placed in a pot of cool water set on a stovetop burner turned up high, we have little idea if the growing warmth of artificial intelligence experiments now around us will comfort us or lead to humanity’s devastation. Artificial intelligence, or “AI,” what was once the plot sidekick of science fiction movies such as “2001” and “Alien,” has been gaining speed, surrounding us with tactical adjustments at first, and now, startling advances. For example, the pretty little pop ballad in the video above was entirely scored by an AI algorithm.

If this AI debate sounds theoretical, watch the video, then ponder that a computer wrote that song. If you’re smart, that blows your mind. Then keep reading.

There are two schools of thought on where AI is going. The positive view is led by Kevin Kelly, the granola-y co-founder of Wired magazine who spent his youth backpacking China with a camera before becoming a technology savant. Kelly posits that AI is the electrical grid of the coming era, a growing intelligence service that future applications will simply plug into. We’re seeing evidence of this utility today. Cars have automatic brakes that pulse faster than your feet can reach the peddles. Most of your next airplane flight is navigated and managed smoothly by computer systems, not real pilots. Apple iPhone’s Siri (an offshoot of the government’s DARPA) and Amazon’s Alexa answer almost any question. Because most human requests or desires can be broken down to algebraic solutions, even songwriting can be managed by computers today.

The pop ballad above is sung by human Taryn Southern, but the entire musical score and orchestration were written by an AI algorithm named Amper. Amper was designed by a team of professional musicians, but only computer logic is at its core. Somehow it analyzed what makes pop songs “work,” then built an entirely new melody that you want to hum. The idea that something as nuanced as music that delights your soul can be written by automated code is a bit scary. If emotional storytelling can be modeled and then evolved by a computer, what’s next? Entire novels, written so eloquently that Will Shakespeare or F. Scott Fitzgerald or Cormac McCarthy couldn’t compete? Or perhaps deeper logic structures, such as future religions, with spiritual tales that resonate above and beyond what mere human prophets can conceptualize? God may exist, but if our understanding of God depends on logical interpretation, surely an advanced computer system could write the better belief code.

Which leads us to the negative forecast for AI as well, led by the Oxford philosopher Nick Bostrom. Bostrom is notable for two efforts: His challenging book “Superintelligence,” which predicts there are multiple research pathways to launching AI, so it will happen, and that when it happens, it will soon go beyond our control. And second, for a little paper “Are You Living in a Computer Simulation?,” which gave even Elon Musk pause. Bostrom sure sounds right: If AI can get smart, it will easily soon get smarter than humans; and if not tempered with the right controls, that AI could chase missions at odds with our survival quickly.

Paperclips everywhere

For example, Bostrom imagines an AI working within a factory that manufactures paperclips that escapes the building, with the simple mission to optimize paperclip output, and reconfigures the entire world into machines mining metal for bending wires while humans suffer, unable to stop it. Silly, perhaps, but the U.S. stock market “flash crash” of May 6, 2010, almost tanked the economy when two different computer algorithms started feeding off each other in a strange way that caused the market to tank a whopping 9 percent before circuit breakers kicked in.

AI is becoming a buzzword, but it really is an ecosystem of many moving parts. IBM has rebranded itself around its “Watson,” an AI that does everything from predicting weather patterns and healthcare outcomes to managing display advertising programmatic buys, in tests that outperformed human bidders by 2 to 1, according to IBM’s head of brand, Jon Iwata. In a speech to the U.S. Association of National Advertisers a few years ago, Iwata stood on stage with the image of hundreds of nested hexagons projected behind him, like the honeycomb of a beehive, each with a tiny label showing a different service that AI could complete. IBM had just acquired to pull feeds about atmospheric conditions into its AI honey trove. When Iwata told the gathered marketers there that his artificial intelligence behemoth had recently beat humans in planning digital advertising, the audience didn’t gasp, but simply nodded. Later, on the patio of the Florida hotel, real humans representing advertising technology vendors met with marketers over wine, trying to convince each other that their human teams would perform better for advertising results. Watson didn’t show up, but I could feel him lurking, observing somewhere behind the scenes.

Do you speak Chinese?

There is a philosophical puzzle behind AI, which of course is your question, is artificial intelligence really “intelligent”? Philosophers suggest this doesn’t really matter. If IBM’s Watson or Amazon’s Alexa are not really real, that won’t change the outcome of what those AIs do or how you experience them. The classic answer to this question is the Chinese room mind experiment by John Searle.

In essence, Searle figured out that AIs can be intelligent even if they don’t have the recursive, self-aware consciousness (i.e., “living souls”) that we humans believe we have. His example: Imagine there is a room with an English-speaking person inside it, who can’t speak Chinese. Before him is a door with a slot, and behind him is a massive library filled with drawers with the answer to any question posed in Chinese.

A Chinese-speaking person on the other side of the door writes down a question, and slips the paper inside the slot.

The English-speaking person picks up the paper. He has a coded instruction manual that leads him to the right drawer; picking up the answer, he pushes the Chinese solution back through the slot.

Here’s the rub: The Chinese speaker outside gets a perfect answer in Chinese. The English speaker inside the Chinese room didn’t understand what he did, but somehow he gave the perfect answer.

The English speaker was not “intelligent” or “self aware” in Chinese, but he got the job done.

Computers are doing that, already, today.

AI is here, and it’s getting warm

The point is, AI is more than a buzzword and much more than a digital tactic. It is no longer a movie fantasy, and it does not need a consciousness to complete the recursive loop of understanding and doing things, even better than humans do. The looming wave of coming intelligent algorithms will be the new electrical current of society, and its application will expedite everything from marketing predictions to media buys to the creative advertising images and stories and music that get people to respond.

Where AI goes is hard to predict, but its mission seems obvious. Computer intelligence will be an accelerant to human desires and objectives. It is more than databases and algebra. AI will do more than pump your brakes or direct your airplane. It will empower anyone to try to predict what anyone else wants.

Like a song structure that anticipates the next note to delight you, AI will build structures that align with your needs. We just hope you know what you really want.

(Thanks to Angela Natividad for the song tip.)