Let’s build a talking dog.
First, collect data. You’ll have to collect two kinds of data. The first kind of data is the sensory input the dog absorbs from his environment (visuals, audio, etc.). A tiny camera and some clever computer vision hacks could do the trick for sight. Now handle the other senses.
The second kind of data is how the dog responds to the sensory input it receives. This can include EEG signals, heart rate, perspiration, hormone and neurotransmitter levels, just to name a few metrics. You can go deeper. More high-quality, diverse data will lead to more precise results.
Next, interpret the data. Run it through algorithms, in real time. Cross-reference the data against other data. Recognize patterns, and then go beyond that. Don’t just identify correlations: Yes, the dog experiences pleasure while he plays fetch. But does he prefer the front yard or the back yard? The lacrosse ball or the stick? Morning or mid-day? Isolate factors, go further, and find causation.
Finally, translate the interpretations of your data into speech. The intelligent agent(s) behind the data analysis must learn how to best translate its analysis into a useful message.
It’s mid-day and your dog just trotted outside. You’re about to begin a game of fetch. Suddenly, little Baxter hesitates. He growls. His stress response is rising. Baxter stares into the direction of the neighbor’s fence. This situation has happened before.
Now, this biofeedback is run through an algorithm and analyzed in light of previous experiences. Remember — the dog may not be consciously aware of what he’s experiencing. Baxter may not be able to articulate to himself why he’s having this adverse reaction. But the reactions and the memories and the preferences exist nevertheless.
Finally, Baxter — via a speaker on his collar — “says,”
“I do want to play fetch, but I hate the neighbor’s dog behind the fence and I just have a feeling that he’s nearby. Can we play somewhere else instead?”
We’ve successfully gone from bark to byte to brains. In other words, we ran biofeedback through an algorithm to translate the dog’s behavior into language. This communication tool reduces ambiguity in interactions between dog and man. It should not be adapted to enhance a dog’s obedience. This is a tool meant to augment the freedom — whatever that may be — of the dog.