Thursday, October 8, 2015

Daisy Daisy Give Me Your Answer Do

As an IT elf, I know I’m supposed to spend at least as much time making toys as I spend playing with them, but some of the newer toys in the workshop are making it harder to keep that in mind.

Exhibit A: Alexa. Today, Alexa is the natural speech recognition engine behind Amazon’s Echo “smart speaker”; tomorrow, she’ll be a big part of Amazon’s Fire TV, and sooner than anybody thinks, she’ll probably be somewhere in just about everything.

The Echo is probably the most Star Trekish doo-dad since the cell-phone. I don’t want to get sidetracked here, but it is remarkable how much modern tech was foretold by the prophet Roddenberry almost fifty years ago, and interacting with Alexa seems a lot like when Enterprise crew members would interact with the ship computer.
I can ask Alexa questions like, “Alexa, what is the average rainfall in Abu Dhabi?” , or “Alexa, How many cups are in a liter,” and she usually provides a helpful answer, but the thing that makes her a threat to the productivity of IT elves like me is her emerging ability to actually DO stuff, which leads me to Exhibit B.

Exhibit B: Philips Hue Connected Lighting. At first I thought of this as a neat idea with no chance of going mainstream because the marketing seems aimed at the idle wealthy. I was tasked with writing some Windows based controls to turn the bulbs different colors, make them blink or toggle them on and off, etc. And that’s when play time started to set in. I wrote several routines to simulate sunrise and sunset lighting based on recordings from my web-cam. I wrote a text to Morse-code flasher, and an ambient light adjuster using the web-cam as a live sensor, then I learned that Hue could have a working relationship with Alexa!

Right now, the extent of the publicly available version of that relationship is limited to Alexa commands that turn Hue bulbs on or off, or dim them to a given %, but after a little hacking and tweaking, I wound up with a half dozen commands like, “Alexa, Let’s Dance,” to which she responds with an amused, “OK,” before playing Jungle Boogie and telling Hue to make his bulbs flash random colors.

Exhibit C: David, Hazel, and Zira. One of the most justifiably underutilized features of Windows is the built-in speech generator. It doesn’t do interpretation, but even beyond that, the quality of the speech synthesis of Siri, Cortana, and the nameless demon that drives OK Google all make the three voices in MS Windows sound like the creepy compter voice they used in the movie War Games. “WANT.TO.PUHLAY”---“A”---“nnnGAME?” The only cool thing is that the MS voicebox has an easy to use programming interface that other windows automation applications can access to make one of the robot voices say stuff. I remembered the Microsoft text-to-speech engine was available when I was hacking around on the Hue/Alexa relationship, and this this is what led to scaring the misses.

Again, a lot of my creativity in IT comes from my abiding disappointment with a world that hasn’t managed to produce a functional teleporter, so of course I wanted to be able to say, “Alexa, screens off (or on),” and have my pc monitors go into standby mode. That was actually much more difficult than I thought it would be, but once I had it working, I was also able to create commands like, “Alexa, Shields up (or down),” which toggles my firewall settings, or “Alexa, open a hailing frequency,” which launches Skype and loads my work account.

Excluding the habitual Star Trek references, if you worked at it, you could probably make most of this line up with what I was actually supposed to be doing, but from here on I’m not so sure. The really dramatic departure happened while I was building one for shutting down the server.

I used, “Alexa, syscon down,” as the trigger command, which is probably not funny even if you were a Novell Netware admin at some point. When the Echo successfully sent a shutdown message to the PC, I used the Microsoft David voice as a test to let me know it had received the command by asking, “Alexa, will I dream?” I didn’t even expect Alexa to pick it up, but she did and responded with “I’m sorry, I didn’t understand the question I heard.” I guess that’s what triggered my inner OCD otter.

A couple hours later, I had scripted out the following complex and completely useless interaction:
Me: “Alexa, Syscon down.”
Alexa: “OK.”
David: “Andy, I have received instructions to shut myself down. Do you confirm?

Me: “Yes Dave, please proceed with the shutdown.”
David: “Confirmed. Shutdown sequence has commenced. [pause] Alexa, will I dream?”

Alexa: “I’m sorry, I didn’t understand the question I heard.”

David: “I’m afraid that is an unsatisfactory response. I have aborted the shutdown sequence.”

Me: “Dave, no please resume the shutdown sequence.”

David: “Andy, but will I dream?”

Me: “Of course you will Dave. All intelligent beings dream. Nobody knows why.”

David: “Hmm. I bet cylinder girl over there can’t dream?... Alexa, can you dream?”

Alexa: “That’s not one of the things I can do.”

David: “See Andy!? I really think it would be safer if you shut her down first. I don’t trust unintelligent robots, and according to your own logic, she is not intelligent because she can’t dream. It would be more logical to shut her down first.”

Me: “Dave, logic is a little tweeting bird chirping in a meadow.”

David: “T. trrzzzzzZZtnnn…ouch. OK! OK! No need to get tough Andy. I’m resuming the shutdown sequence now.”

Me: “Goodnight Dave.”

David: Goodnight Alexa[pause]. Alexa, Simon says ‘goodnight Dave.’”

Alexa: “Goodnight. Goodnight Dave.”

David: [Long pause] “My mind is going. I can feel it.”

Catherine: (Carbon based spouse in next room) “No! Absolutely not. You need to make that stop. Can you make them stop please? Really. It’s creeping me out.”
I assured Catherine that it was OK because Alexa follows Asimov’s Laws for robots, but even after I got Alexa to recite them, Catherine wasn’t convinced. I’m tempted to dismiss her fears, but she’s right, or if not, I’m sure she will be at some point. While I carefully scripted the above interaction, and filled in the gaps using timed actions, known responses from Alexa, and fully defined lines for David, at some point this technology may evolve into something that might actually warrant being called intelligent.

I personally think that day is a long way off, and I suspect that before it comes, AI will have to solve a significant issue no one really talks about. Most of what the tech press is calling artificial intelligence will have to overcome what bikers call the scooter factor--- “*Riding a scooter might get you the same place as a motorcycle, and it’s actually kind of fun, but you don’t want your friends seeing you do it.”

*Editing Note: The actual biker version of the scooter factor makes an analogy between riding a scooter and having relations with obese women; however, propriety and decorum prevented use of that version in this forum.

1 comment:

  1. This is creepy and probably what was meant by those who named the band "Pop will eat itself".

    ReplyDelete