When we envision artificial intelligence, we tend to go 0 to 100 real quick. Any basic machine learning inevitably leads to worried visions of Terminator or Matrix or “Black Mirror” future realities. In some ways this is fair. We should steel ourselves from this happening. But Google A.I. Experiments are much less end-of-worldy and more super cool.
RELATED: People Who Use Weed Also Do More Of Another Fun Thing
But A.I. doesn’t always have to be so doomsday. Imagine being alone in a different country where you don’t speak the language. You peruse the market, and find a delicious-looking Bartlett pear. How do you communicate the purchase of such an item when you can’t speak the language?
These Machines Are Here To Help (Don’t Panic)
This is an A.I. experiment called Thing Translator. It’s part of Google’s A.I. Experiments program featuring open-source projects that help coding newcomers and experts alike tinker and play with machine learning. As seen above, Thing Translator allows users to take a picture of a thing, like our hypothetical Bartlett pear, and translate it into various languages, sounding out pronunciation and displaying the written form of the word.
Another component of Google’s A.I. experiments has to do with sound. For human ears and brains, sound is a deeply complex and robust experience. Think of music. How many times have you heard something you’d qualify as percussion? A booming bass or a baseball bat ding or a high-hat rhythm. How would you describe the difference between these sounds? Furthermore, how would you organize them into searchable and similar categories?
Turns out machines are very good at this. The above experiment feeds hundreds of drum machine sounds into a computer without any indication or hints of what they are. To the computer, it’s just a conglomerate of sounds. But it can then cluster sounds that are similar together and organize them into a color-coordinated scatter plot through a process called t-SNE.
RELATED: Beer Sales Flatten Thanks To Marijuana
Most of these Google A.I experiments are simple, typically reducing data into a visualized component or vice versa. Be sure to check them all out, they’re certainly worth your time. Hopefully you won’t feel too much like you’re living in The Matrix while playing around with them.
