02.03.2022

Do machines see like humans?

They are getting closer, writes Robert Gheiros in our latest blog post.

Machines may drive you to work one day, but they currently still fail when faced with unusual situations or noisy data. That’s because machines see the world very differently from humans. This gap is beginning to close, however, and the image recognition capabilities of machine learning systems are increasingly catching up with those of humans.

What’s the key to making machines better at robust object recognition? Robert Geirhos’ latest results surprised him, as he writes on our blog: The key factor turned out to be not the type of model, but the amount of training data:

Read the full blog post here