Emerging Technology

If you happen to be one of the 28 million people who purchased an iPod in the past few years, chances are you've spent some time listening to your music in shuffle-play mode. I am a fan of shuffle play in theory. Increase the ease of assembling music collections that number in the thousands of albums, build a system for randomly accessing any song in that library at the touch of a button, and you have an engine of surprise and serendipity. Certainly, it's a terrific antidote to the plodding predictability of radio stations. But shuffle play in practice is not as joyful. Granted, a randomized playlist can sometimes unearth songs buried on your iPod that you haven't listened to in years, and the surprise of hearing those songs afresh might reveal nuances you missed the first time around. But some songs should stay buried. I don't know how Duran Duran's "Girls on Film" got in my iPod collection, but believe me, there's no unexplored subtlety to that song that is going to be revealed on future listenings. I would listen in shuffle-play mode nonstop if my iPod would give me a one-click mechanism for effectively voting a song off the shuffle island. As it is, I spend too much time on the New York subway pulling the iPod out of my coat pocket—thereby attracting an entire carload of potential thieves—to instruct the machine for the umpteenth time that I do not want to hear Rush's "Tom Sawyer." That wouldn't happen if the underlying software were programmed to grant me a single wish: Watch each time I fast-forward past a song, and if I do it more than three times to the same song, drop that song from the shuffle rotation. Don't delete it, don't rub it from my consciousness altogether—just stop recommending it to me. This may seem like a small complaint, but it points to a larger issue. Think of all the decisions we now regularly off-load to various machines: We get product recommendations from Amazon; our TiVos record programs for us based on their knowledge of our general tastes; even dating services are starting to use software algorithms to suggest matches. So if we're going to ask machines for help, we need tools for training them. We need, in effect, a few new verbs. What do I mean by verbs? Think about these familiar symbols: Would they have meant anything to 99 percent of Americans 50 years ago? Yet now they are as recognizable and intuitive as a red light. We needed these symbols to help us navigate the linear, largely tape-driven technologies—audiocassettes and VCRs—that emerged more than three decades ago. Each new consumer-technology platform creates the need for new verbs. The first generation of television and radio gave us knobs that meant "change the channel" or "adjust the volume." Thanks to CDs, "fast-forward" and "rewind to the next chapter/song" have entered the pantheon. After a few years of seeing these icons on multiple appliances, they become second nature to us. Verbs carry over from platform to platform, but new platforms also create new verbs. "Move to the next chapter/page" is largely a digital verb, the kind of basic command you need when dealing with a medium that knows something about the structure of the information it's presenting. You don't skip ahead automatically to the next chapter on traditional video, film, or audiotape unless it has been augmented with digital information that tags chapter or song divisions. Old cassette-tape players used to get around that limitation by looking for patches of silence in the audio signal and interpreting those patches as song breaks, which worked fine if you weren't listening to John Cage. In the age of iGadgets, it is the loss of control that requires new verbs: We're handing decision making off to software, letting it set the mood at the party or suggest books to us. For the most part, I think this is a good thing, a technological trend likely to produce more diverse media consumption in the coming years as we shuffle through ever larger libraries. But smart algorithms, like smart pets, need to be trained. You have to give the software good feedback about its recommendations. That feedback needn't be nuanced; in fact, if the ultimate aim is to create new verbs that are universally recognized, the simpler the better. So here's my proposal. Any media platform that relies extensively on recommendations needs two universally recognized verbs, as fundamental as stop, play, cut, paste, open, and close. They would embody the commands "don't ever suggest this again" and "pay no attention to what I'm doing"—or in shorthand, "remove" and "ignore." Remove is simple enough: When you're in the subway and Jethro Tull's "Aqualung" comes on, you click remove once and that hideous flute solo is forever banned. Do that for a few weeks, and you'll have a shuffle rotation that you can listen to without frantic adjustments. When your TiVo decides that you might like Fear Factor and starts recording episodes for you, you can tell it with one stern admonition: never again. Ignore is a command directed at algorithms that learn by watching your behavior, like Amazon's recommendation system. Anyone who has bought more than a few books from Amazon knows how one quirky purchase can suddenly throw off the recommendation engine. You order a copy of Curious George Flies a Kite for your nephew, and for the next three weeks Amazon seems convinced that you've regressed to a kindergarten reading level. With an ignore button, however, you can simply say, "Don't pay any attention to this particular purchase; it's an anomaly." You can find examples of these verbs already implemented throughout the digital worlds. TiVos include the wonderfully intuitive "thumbs up/thumbs down" buttons built right into the remote control. A number of browsers let you adopt a "private browsing" mode that doesn't record your surfing itinerary for as long as the option is selected. (Cynics sometimes call this porn mode.) The major search engines recently adopted a new hypertext standard called "no follow" that allows you to link to a page online without the search algorithms interpreting your link as an endorsement of the site. Amazon offers tools to modify its recommendation engine and does a great job of explaining the logic behind its picks. And you can tell Apple's iTunes software to take a song out of shuffle rotation, but you have to click through three different windows to do it. The World Wide Web has its own version of shuffle play: a site established a decade ago called URouLette (www.uroulette.com). If you click the image of a roulette wheel on the front door, it takes you to a random page somewhere on the Web. Although not as useful as Google, the site provides a good introduction to the Internet's vast database. The diversity of these solutions suggests just how serious the need for these new verbs is. But diversity is also part of the problem. For example, imagine living in a world in which every CD player had a different symbol for stop and play. Having too many solutions means that the people who end up using them are either the people who bother to read the instruction manual or the people who don't need to read the manual. Most consumers don't realize that these training tools exist, because they haven't been standardized into a simple, consistent vocabulary. Designers, take heed. We need simple symbols that will help our smart software get smarter. The trend here is a dramatic one. Software recommendation engines were unheard of 15 years ago and have only been adopted in any mainstream sense for 5 years. Imagine how many cultural and social decisions will be made through them 20 years from now. If those decisions are going to be smart ones, the software algorithms will need to be trained by their masters.

Emerging Technology

Hey, Steve Jobs: How about a machine that knows some verbs?

Newsletter

Steven Berlin Johnson

Published in:

The New Planet Machine

AI Supports Dishonesty in Humans, Making It Easier for Users to Cheat With an Accomplice

It's Possible to Manipulate This New Interactive Hologram, Which Could Improve Medical Technology

Self-Driving Cars Are Communicating Better – What Does This Mean for Our Safety?

Quantum Computers Are Like Kaleidoscopes, Helping Illustrate Science and Technology

Synthetic Reality Can Generate Worlds and Drive Innovative Solutions

The Year of the AI Conversation

What Is the History of the GIF?

TikTok's Algorithm and How It Affects Your Viewing Experience

Was the Y2K Bug Real ... or a Hoax?

Do Smartphones Need to Get Smarter?

AI Voice Replication Is Transforming Star Wars And Other Media

The Future of Baseball Might Include Robot Umpires

What a Digital Worker Could Mean for the Human Workforce

How Hackers Take Down Websites

3 Ways You Use Quantum Physics Every Day

Stay Curious

JoinOur List

SubscribeTo The Magazine