Chris weighs up the ethical implications of Google Duplex:
The social hacking that could be accomplished is mind-boggling. For this reason, I expect that having human-sounding narrow AI will be illegal someday. The Duplex demo is a moment of cultural clarity, where it first dawned on us that we can do it, but with only a few exceptions, we shouldn’t.
But he also offers alternatives for designing systems like this:
Provide disclosure, and
Design a hot signal:
…design the interface so that it is unmistakeable that it is synthetic. This way, even if the listener missed or misunderstood the disclosure, there is an ongoing signal that reinforces the idea. As designer Ben Sauer puts it, make it “Humane, not human.”
I can’t recall the last time I was so creeped out by a technology as I am by Google Duplex—the AI that can make reservations over the phone by pretending to be a human.
I’m not sure what’s disturbing me more: the technology itself, or the excited reaction of tech bros who can’t wait to try it.
Thing is …when these people talk about being excited to try it, I’m pretty sure they are only thinking of trying it as a caller, not a callee. They aren’t imagining that they could possibly be one of the people on the other end of one of those calls.
The visionaries of technology—Douglas Engelbart, J.C.R Licklider—have always recognised the potential for computers to augment humanity, to be bicycles for the mind. I think they would be horrified to see the increasing trend of using humans to augment computers.
Merlin mentioned this service on a recent podcast episode. If you have an Amazon Echo, you can authenticate with this service and then point it at an RSS feed …like your Huffduffer feed, for example. From then on, Alexa becomes a Huffduffer player.
This 1993 article by Mark Weiser is relevant to our world today.
Take intelligent agents. The idea, as near as I can tell, is that the ideal computer should be like a human being, only more obedient. Anything so insidiously appealing should immediately give pause. Why should a computer be anything like a human being? Are airplanes like birds, typewriters like pens, alphabets like mouths, cars like horses? Are human interactions so free of trouble, misunderstanding, and ambiguity that they represent a desirable computer interface goal? Further, it takes a lot of time and attention to build and maintain a smoothly running team of people, even a pair of people. A computer I need to talk to, give commands to, or have a relationship with (much less be intimate with), is a computer that is too much the center of attention.
Our web browsers easily and invisibly collect cookies, allowing marketers to follow us across the web. Our phones back up our photos and contacts to the cloud, tempting any focused hacker with a complete repository of our private lives. It’s as if every tacit deal we’ve made with easy-to-use technology has come with a hidden cost: our own personal vulnerability. This new voice command exploit is just the latest in a growing list of security holes caused by design, but it is, perhaps, the best example of Silicon Valley’s widespread disregard for security in the face of the new and shiny.
The transcript of Josh’s fantastic talk on machine learning, voice, data, APIs, and all the other tools of algorithmic design:
The design and presentation of data is just as important as the underlying algorithm. Algorithmic interfaces are a huge part of our future, and getting their design right is critical—and very, very hard to do.
Jason breaks down the myths of inputs being tied to device form factors. Instead, given the inherent uncertainty around input, the only sensible approach is progressive enhancement.
Now is the time to experiment with new forms of web input. The key is to build a baseline input experience that works everywhere and then progressively enhance to take advantage of new capabilities of devices if they are available.
Some interesting outcomes from testing gov.uk with blind users of touchscreen devices:
Rather than reading out the hierarchy of the page, some of the users navigated by moving their finger around to ‘discover’ content.
This was really interesting - traditionally good structure for screen readers is about order and hierarchy. But for these users, the physical placement on the screen was also really important (just as it is for sighted users).
There are no videos from this year’s dConstruct—you kind of had to be there—but Mandy’s talk works astoundingly well as a purely audio experience. In fact, it’s remarkable how powerful many of this year’s talks are as audio pieces. From Warren’s thoughtful opening words to Cory’s fiery closing salvo, these are talks packed so full of ideas that revisiting them really pays off.
Then again, I’m something of a sucker for the spoken word. There’s something about having to use the input from one sensory channel—my ears—to create moving images in my mind, that often results in a more powerful experience than audio and video together.
We often talk about the internet as a revolutionary new medium, and it is. But it is revolutionary in the way that it collapses geographic and temporal distance; we can have instant access to almost any information from almost anywhere in the world. That’s great, but it doesn’t introduce anything fundamentally new to our perception of the world. Instead, the internet accelerates what was already possible.
Even that acceleration is itself part of a longer technological evolution that began with the telegraph—something that Brian drove home in in his talk when he referred to Tom Standage’s excellent book, The Victorian Internet. It’s probably true to say that the telegraph was a more revolutionary technology than the internet.
To find the last technology that may have fundamentally altered how we perceive the world and our place in it, I propose the humble gramophone.
On the face of it, the ability to play back recorded audio doesn’t sound like a particularly startling or world-changing shift in perspective. But as Sarah pointed out in her talk at last year’s dConstruct, the gramophone allowed people to hear, for the first time, the voices of people who aren’t here …including the voices of the dead.
Today we listen to the voices of the dead all the time. We listen to songs being sung by singers long gone. But can you imagine what it must have been like the first time that human beings heard the voices of people who were no longer alive?
There’s something about the power of the human voice—divorced from the moving image—that still gets to me. It’s like slow glass for the soul.
Creating telephone answering systems can be fun as I discovered at History Hack Day when I put together the Huffduffer hotline using the Tropo API. There’s something thrilling about using the human voice as an interface on your loosely joined small pieces. Navigating by literally talking to a machine feels simultaneously retro and sci-fi.
I think there’s a lot of potential for some fun services in this area. What a shame then that the technology has mostly been used for dreary customer service narratives:
Horrific glimpse of a broken future. I sniffed while a voice activated phone menu was being read out and it started from the beginning again.
There’s been a lot of talk lately about injecting personality into web design, often through the tone of voice in the microcopy. When personality is conveyed in the spoken as well as the written word, the effect is even more striking.
What happens when Customer Service bots start getting too smart? What if they start needing help too? How would they use the tools at their disposal to reach out to those they care about? What if they start caring about us a little too much?
It’s using the Voxeo service, which looks similar to Tropo.
The end result is amusing …but also slightly disconcerting. You may find yourself chuckling, but your laughter will be tinged with nervousness.