Related information :

Declaration of Cooperation on Artificial Intelligence: part of the text related to it: “In October 2017, the European Council asked the Commission to present a European approach to Artificial Intelligence (AI). The Commission has announced that it will adopt a Communication on AI in April 2018. This Declaration builds on the achievements and investments of Europe in AI as well as the progress towards the creation of a Digital Single Market.“ Funny enough, this is also in October 2017 that I got the idea of contributing to the field of AI, via a website or a blog, and in April 2018 that I just posted my first article in this blog, blog I’ve created a week ago!

Approximation of the mapping function. We can consider that the computer has learned enough from the training set, i.e. has reached a good-enough approximation of the mapping function from how to generate the correct output from the input/inputs, when this approximation is not too far from the original output. How much this “good-enough” means depends on the task at hand and how much the user wants it to be precise/accurate to the original output. Therefore, the margin of errors is variable and could go let’s say from few percents (~ 5 %) to almost perfect. For example, in the field of image recognition, the best AI so far could reach an approximation of 3% compared to the original output (i.e. the computer still makes 3% of errors in generating the right output, which is better than humans on average, who make 5% of errors. This approximation is calculated by comparing the performance of the computer to the original output.

High computational power: I think this is important to understand as it might have its relevance to understand the main difference between humans and AI so far. High-computational power computer can achieve complex (maths) calculations very quickly. And the more you increase this computing power, the fastest it gets at computing more and more complex calculations. Thus the two variables which increase with the level of computational power are speed and complexity of the calculations performed by the computer. In brief.