We created an Intelligence, we need to regulate it.

2 min read

Notes from my digital room - Issue #9

Hey Digital Folks!

I love to listen to podcasts while I work, and this week I could enjoy a talk with Elon Musk!

One thing that resonated with me is that he said that we should regulate Artificial Intelligence🤖. Maybe it's hard to understand what AI can do already, but here's a short list:

While it might be hard to grasp for everyone what can extensively A.I. do, even myself could develop small artificial intelligence that could slowly "learn" how to play music:

Imagine now a corporation, with a dedicated team and a lot of money. It might be harder to understand why we need to regulate, rather than agree on that after you get it.

What is an A.I.? 🤖

In short, I promise.

Having said that, I would add something: what constitutes AI? Lately more and more people are categorizing more and more algorithms by the tag A.I., but there's very less of an intelligence, rather a simple

if this happens do this, otherwise do that. Start again.

What categorize an A.I. is an actual system of "learning": they are developed with a neural network that, in short, allow them to actually learn, either by being fed by information (think about it as a sort of bias) or they have a series of algorithm that they evaluate the input, applies some modifications, and evaluate the outcome as if it's conform with the expectations or not.

https://s3.us-west-2.amazonaws.com/secure.notion-static.com/1b1da7e6-663b-4ec2-ba99-37fc79ab8420/IMG_20210212_145359__01.jpg?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=AKIAT73L2G45O3KS52Y5%2F20210214%2Fus-west-2%2Fs3%2Faws4_request&X-Amz-Date=20210214T002812Z&X-Amz-Expires=86400&X-Amz-Signature=1d6c10945f0f02287d3c50fb17adfb3f996ad6cfd713957880283867a898cfda&X-Amz-SignedHeaders=host&response-content-disposition=filename%20%3D%22IMG_20210212_145359__01.jpg%22
Image of a Neural Network - [1]M. Taylor, Make Your Own Neural Network: An In-depth Visual Introduction For Beginners. Independently published, 2017.

Think about Neural Network algorithms as approximation functions: they "casually" output numbers and evaluate if that is conformed or not.

Believe it or not, this is similar to what toddlers do: they love to explore the world, and they frequently would happily put anything in their mouth. That's how they evaluate if something is good or not.

What I just described is a really, really extreme approximation of a really complex and fascinating theme, which is to separate normal algorithm from Artificial Intelligence ones.

Have a great Sunday, a great end of the day and see you next week! ❤️