Skip to content
Next Episodes
Mon: 7:30pm
Tues: 7:30pm
Classic Episodes
Mon: 1:30pm
Tues: 1:30pm
Is How Normal Am I AI safe? The AI can predict your BMI, but does it store your data? Credits: Social Media
By Jane Corscadden

Is How Normal Am I AI safe? The AI can predict your BMI, but does it store your data?

By Jane Corscadden

Is How Normal Am I AI safe? The website, hownormalami.eu, asks for camera access, but does it use any further data and information? Its AI is powerful enough to predict users’ BMI, age and gender with relatively high accuracy, which has sent Reddit users into a flurry of speculation about the website.

What is How Normal Am I?

Its a website that requires camera access to analyse users faces and tell them how normal their features are. It can also go further than this, guessing all sorts of information such as your BMI, age, and weight.

FUZHOU, CHINA – AUGUST 28: A citizen scans her face to go through a ticket barrier of a subway station on August 28, 2020 in Fuzhou, Fujian Province of China. (Photo by Zhang Bin/China News Service via Getty Images)

How Normal Am I is described as an art project run by tech critic Tijmen Schep. The project has been funded by the European Unions Sherpa research programme, which looks at the impact of data analytics on ethics and human rights.

The concept behind How Normal Am I is that it uses artificial intelligence (AI) to demonstrate how face detection algorithms are increasingly used to judge you.

This will hopefully educate users on the numerous ways AI face detection is used in everyday cases, such as in preventing crime and finding missing persons, and how these biases can affect the process.

Is How Normal Am I AI safe to use?

The website has raised concerns with some users regarding data collection and privacy. But are these concerns grounded in truth?

On the terms and conditions page, the website states that no personal data is sent to their server in any way and all the face detection algorithms run on individual users’ computers, confined to their browsers.

At the end, users can (if they choose to) share some data anonymously, but this is completely optional.

This makes the How Normal Am I AI as safe to use  or even safer  than most things online nowadays.

The website’s terms and conditions also go into detail about the machine learning sources used to build the AI algorithm that reads people’s faces.

For instance, they used a third-party machine learning model that was pre-trained from an open source project on Github. They use this purposefully  to mimic the AI used by most organisations.

As most places use third-party AI, they arent training their own machine learning models, which can often lead to biases.

Essentially, How Normal Am I kept these errors in on purpose, to showcase the issues of the one size fits all models used by organisations around the world.

Users shouldnt be concerned about data and privacy issues when using the How Normal Am I AI. Its sole use is educational, and its creators clearly state it doesnt hold onto personal information (unless you give it specific permission to).

What concerns do users have?

The concerns raised about How Normal Am I are similar to those shared by anyone who knows about AI.

Essentially, as an educational tool, How Normal Am I is meant to show the flaws of most machine learning models  especially those that arent programmed in-house.

For instance, AI that is only trained to look for white faces may not recognise an image of a black person. This creates a racial bias that may have serious implications in areas such as law enforcement.

These are major issues that need to be fixed. But thats what How Normal Am I, and the EUs Sherpa programme, are for  to point out these ethical issues in tech and fix them.