by Laurie Sullivan @lauriesullivan, October 9, 2017

Google’s latest browser experiment aims to teach people who use the Internet about artificial intelligence (AI).

Teachable Machine provides a short summary of what AI can and cannot do. The technology learns from examples. The tutorial, which demonstrates some of the fundamental aspects of AI, requires the use of the student’s webcam to train a basic AI program.

The experiment allows people to train a neural network locally on a desktop, for example, without sending any images to a server, enabling it to respond quickly.

The post also includes links to some of the examples that people have done so far and posted to YouTube. To participate in the learning, Google says, users should capture at least 30 images per class by pressing and releasing the button, which starts and stops the ability to capture images.

The machine will whatever it sees through the webcam, and once it has learned enough, will output the content as a GIF or a sound effect or speech when it sees the object or activity the person used to train it, according to Google.

None of the images are stored on Google’s servers — they are all stored on the person’s device.

Teachable Machine was made in collaboration with Støj, Use All Five and Creative Lab and PAIR teams at Google.

Researchers have begun testing the smarts of AI. At Cornell University, researchers analyzed AI assistants to determine the ones with the highest intelligence. Google’s AI has an IQ of about 47.28, putting it slightly below the IQ of an average six-year old. Siri, Apple’s AI, registered an IQ of 23.9. Researchers also tested Microsoft Bing.

This is not the first IQ test from researchers run on virtual assistants. Stone Temple Consulting released a report in April titled “Rating the Smarts of the Personal Digital Assistants,” which analyses Google Assistant, Cortana, Siri, and Alexa, including their capabilities to take an action on the user’s behalf.

MediaPost.com: Search Marketing Daily

(61)