Singularity is near — and it’s in the form of a dachshund named Elly / by Karen Frances Eng

Meet Elly—the world’s first artificially intelligent dog.

Restless tech innovator Yale Fox gave voice to his dog Elly by mounting Amazon Alexa software to a Raspberry Pi and hooking it up to a mic and speakers. Now Elly can talk to people she meets on the streets, picking up keywords that prompt responses.

On one level, it’s just a fun, simple prank, but Fox has bigger plans. We asked the open data advocate, entrepreneur, and brains behind building-quality standards organization Rentlogic to tell us why he created this device — which he calls EllyAI — where he’s going with it, and how this is just one example of the potential for human-to-machine interaction.

Meet Elly—the world’s first artificially intelligent dog. Image: EllyAI

Meet Elly—the world’s first artificially intelligent dog. Image: EllyAI

How does EllyAI work?

Similar to the way that you train a dog, you can train an artificial neural network. EllyAI version 1.0 listens for speech/keywords in the environment, then responds. When EllyAI hears the word “Elly,” it wakes up and listens for keywords. Elly herself also perks up when she hears her name. If EllyAI recognizes the keywords, it tries to guess what someone is saying to her — and then she responds with something sassy. (See videos.)

Ironically, the hardest part of this project was getting her to wear it. She has a very short attention span, so it took a lot of cookies to get her to cooperate.

What next?

Version 2.0 (which will come out early next year) will start to incorporate additional data, like location (via GPS) and be able to monitor whether she’s walking or chasing a squirrel (via an accelerometer). This is for entertainment purposes, but also to make the effect more real, which will hopefully get people to think. Also it’s an excuse for me to buy the hardware accelerometer add-on for the Raspberry Pi and start tinkering with it.

In a few years, I should be able to sync my Gmail account to EllyAI, train it to learn how I communicate with people, and then have Elly respond in a way that sort of sounds like me. That’s the ultimate goal — but the technology out there is still a bit too nascent. For Version 4.0, Elly will be able to give a TED Talk!

Elly with her gear on, in the park. Photo: Jane Sosi

Elly with her gear on, in the park. Photo: Jane Sosi

What possessed you?

For the first few years of my startup, I was working from home, and having a dog made a world of a difference. I’d often talk to her, and I always thought she was trying to speak back but didn’t have the ability to do so. This led me to wondering, “What would Elly say if she could actually talk?” Years later, that spiraled into this project. It also gave me a good excuse to start tinkering with Amazon’s new machine learning and AI-related services. Building this was just an itch that I needed to scratch. But if I had to categorize this, I would call it an art project that’s meant to get people to laugh — and to think about the possibilities of AI in a lighthearted way. After all, we do things with software/hardware bots all the time. What if we incorporate something that’s actually alive? How does that make you feel, when we attribute a personality to a living creature compared to a machine?

I can’t help but be reminded of the virtual girlfriend in the recent Blade Runner sequel. Is that where this is headed? Do you think artificially intelligent virtual companionship could become a reality?

Absolutely. I think if you can speak to a truly artificially intelligent creation, then it would naturally have its own unique personality. In turn they’ll learn from their interactions with individual humans — which will make each bot even more unique. Like human beings, bots will be shaped by their relationships. (I also wanted to get two dogs to speak to each other and learn off each other. It would create a feedback loop that would lead to unexpected — and in this case, probably funny — conversations.) Really the question is, what happens after you you can have a conversation with a bot that’s indistinguishable from a human? We’ve focused on passing the Turing test for decades, but I think that’s only half the question.

One way or another, I believe humans will form relationships with bots because that’s how we are biologically designed. This concept tends to be polarizing. Some people think AI is the end of the world, others think it’s just the beginning. Both the positive or negative implications are tremendous, and I think as humans/explorers, we should work towards building a true AI, safely and ethically. That means, if we were to give it life — it would have to have rights just like any other living organism. I think that’s a hard concept for people to understand, to give a computer rights. I think the jury’s still out.

Follow EllyAI via Instagram @itselly.bitch or her website

This is an excerpt. To read the full article, visit the TED Fellows blog >>>