There are many attempts to train machines how to mimic us humans by showing them a number of labeled examples. At Fido, we believe this approach is a dead end. Instead, we need to teach machines how to reason like we do.
Most of us on the Fido team are foreigners and English is our second, third or fourth language. We all have learned new languages using grammar books. This is how a sentence should look like. This is how you form a question. This is how you use conditionals. You use relative clauses for this and that. If you want to recommend something or warn someone, you can use these constructions. And so on, and so forth.
This is our secret sauce. We have taught our AI how language works step by step, the same way we learned.
It's very novel but also very intuitive and scalable approach. Imagine that you want to know if a certain place is safe for your kids. It's hard to believe, but every current approach to answering that question requires you to label tons of examples to train a system dedicated to answering only that question. What is worse, such a system won't tell you anything except an arbitrary confidence score it has been trained for. Not very efficient, right?
Fido AI already knows how people can express their fears, warnings, recommendations, and more. It automatically reads and understands billions of reviews and opinions about any places, and can easily tell you if a certain place is safe for your kids or not. It can also tell you why it made those recommendations. But it can do a lot more. It can understand your needs and find the best places for you based on the world's collective intelligence.
The third wave of AI is rising right now. Fido AI is constantly learning more and more by reading millions of reviews, opinions, forums, blogs and tweets every single day.
If you want to learn how we are "Freeing Language from the Fundamental Prison of Deep Learning," check out our blog.
An independent benchmark study compared the accuracy of both Fido and Google language parsers. It concluded that Fido’s technology made four times fewer errors than Google's AI on human-written reviews and opinion.
If you would like to try the parser in action by yourself -- leave us your e-mail: