I am a student of last year of computer engineering and lately I have been very interested in AI. Fields such as ML and DL seem very disruptive to me.

A few months ago I saw an interview of Bill Gates in which he was asked the following: if now you had 20 years in which area of ​​technology would you focus or what kind of company would you start? He replied that he would try to teach to read to an artificial intelligence algorithm and so it would draw all the knowledge from all possible books.

Bill Gates: If I were starting a company today, it would use AI to teach computers how to read

The issue is that currently nobody has managed to do it in a very efficient way. Do you think that with the AI ​​techniques that exist now such as Supervised Learning, Unsupervised Learning, etc., is it possible to achieve something similar? If so, why type of algorithm or technique do you bet?

Thank you. It is only scientific curiosity.


With a bit searching in the internet, it's possible to identify the statement of Bill gates in the mainstream media.[1] In an interview, he has explained, that if he would be young again, he can imagine to start an AI company which develops a software which can read all the books ever written. Before it's possible to analyze this statement from a technical perspective, it make sense to do a survey which papers about “neural network can read” were written already. In Google Scholar, there are some examples given, in which neural networks were trained in reading stories in a book, and a question&answering system based on neural networks was introduced too, which tries to parse the entire Wikipedia.

This helps a lot to answer the original question, if a neural network can learn to read in a general sense. According to published research in the domain of neural networks for natural language parsing, there are some activities available which are trying to realize the goal of letting a neural network read books. The obstacles are located in the detail problem of training these networks in a short amount of time. Usually the researcher are confronted with the problem, that the parameters of the neural network are complex, but the computation resources to determine the connections between the neurons are limited. So it remains an open challenge to solve the issue.

[1] CNBC story about Bill Gates, https://www.cnbc.com/2019/06/24/bill-gates-says-hed-start-an-ai-company-today.html

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.