If Microsoft is to be believed, I am a 23-year-old man, married to a 66-year-old man with a two-year-old daughter and a 42-year-old mother-in-law.

Aside from the fact that this is a biological impossibility, it is also entirely wrong on almost every aspect.

The suggestion above is based on results from a new experiment launched by Microsoft at its Build conference in San Francisco on Thursday, 30 April.

The truth is that I am a 35-year-old man married to a 33-year-old woman with a 7-month-old son and a [redacted]-year-old mother-in-law.

Microsoft's How-old.net uses the company's face detection technology in collaboration with machine learning through its Azure cloud infrastructure to try and accurately guess the gender and age of people in photographs uploaded to the website.

I tested the system with a number of pictures of myself and while some were right on the money (above), others had be rolling back the years to my early 20s.

Constantly learning

The system was built in just a couple of days and when the developers at Microsoft sent out a few emails to ask people to test it, they hoped for at best 50 responses.

In fact the test grabbed people's attention in a big way and within a few hours over 35,000 people had tested the system with over half of those using it uploading their photos of themselves - something which the developers didn't think would happen at all.

The robot - as Microsoft calls the project - is constantly looking online for more photos to analyse in order to improve its recognition algorithms which would suggest that with time, the service will improve.

Looking at the what people have shared on Twitter, the results have varied wildly from entirely accurate to completely wrong. It has recognised a plastic doll as a four year old girl, said that a zombie is 87-years-old and said that Alf doesn't have a face.

My favourite outcome however comes from Scottish comedian Brian "Limmy" Limond:

Microsoft How Old: Limmy
Good advice from the inimitable Limmy Twitter/@daftLimmy