Google allows us to train AI with reCAPTCHA
Google allows us to train AI with reCAPTCHA

Google offers dozens of services for different types of businesses and users. Its services include reCAPTCHA, which is used to protect websites from bots.

All users have agreed to reCAPTCHA while using the Internet. These tests appear here under the heading "Make sure you're not a robot." This is because websites that use this service are trying to prevent bots from entering.

reCAPTCHA contains the images that are displayed to the user. You have been asked to select certain items from these images. Examples include identifying traffic lights, cars, pedestrian paths, etc. in these images.

We may not know that these tests contribute directly to Google's AI model. Google officially announced it. Identifying elements in the image helps prepare data for AI models and improve the company's machine learning technology.

Google's reCAPTCHA service

Like humans, AI requires lifelong learning. It does this through a well-known process called machine learning.

For a machine to learn, it must see clear models, just like humans. For example, in order for a child to see the shape of an airplane, he must see a picture of the airplane. If he saw her more than once, he could easily recognize her in the future.

The device should provide different images so that you can quickly and efficiently identify them in the future. This is the user's role in reCAPTCHA. We show the AI ​​the image of the car, the image of the sidewalk or the image of the motorcycle and all the images that appeared in the test.

When the user recognizes the images in the test, they are called "tags". So he passed the test and went to the website while learning about artificial intelligence from Google.

artificial intelligence training

But the most important question is: How does Google know that the user has chosen the correct answer?

The importance of this question stems from the fact that users' efforts are aimed at defining the undiscovered picture of artificial intelligence and machine learning systems. Thus, there is no guarantee that the user's choice will be correct.

Here for example, Google is using a clever method by asking users to choose five photos out of ten, all of which must contain a motorcycle.

In this case, Google places four known photos of the motorcycle and six unknown photos. If the user succeeds in recognizing the first four images, then the fifth image is actually a motorcycle image. So it is authenticated and sent to the AI.

All of this data is used to develop Google's artificial intelligence and machine learning models. In collaboration with Waymo, his autonomous car project. On the other hand, website owners continue to benefit from the protection of their websites.

Previous Post Next Post