The Price We Pay for Free AI
A couple weeks ago a new trend started to show up all around my Instagram feed. I follow many activists and influencers coming from many countries - and they all seemed to be using the same thing. For a computer science student I am very out of touch with new trends. The viral new softwares that show up on my feed are n than not very unsafe when it comes to data protection. Nonetheless, as I was working on a “painting” robot myself, the colourful avatars, cyborgs, and faeries with faces of my favourite internet people sparked my curiosity immediately. In case you have been living under a rock - just like me - let me introduce you to Lensa AI.
How Your Favourite Influencer Turned Into a Fairy
Lensa AI is a photo editing app that was introduced in 2018. It is one of two products developed by a Russian company Prisma Labs - whose mission statement is to “Democratise Photo and Video Editing”. Sounds very noble, as does any mission statement of any company ever created. But what does it look like in practice?
Lensa AI is a free app that you can download on your phone. Then you have to pay 4$ for it to do the magic - you can upload up to 20 pictures of your face (or any face you would like to see transformed) and you will receive 50 avatars - you can be whoever (or whatever) you want.
Sounds all fine and dandy, doesn’t it? Well, let your favourite computer scientist explain to you why Lensa might not be as fun as it seems.
Where Do Your Photos Go?
The problem with a lot of technology is that they usually don’t even hide the fact that they track your behaviour online. It’s all there. All of it. And if you only look you might find paragraphs like this:
“We also may use [automatic data collection] technologies to collect information about your online activities over time and across third-party websites or other online services (behavioural tracking). If you do not want us to collect this information, do not download Prisma or delete it from your device.”
I mean, they do seem very fair - if you do not want your data to be collected, do not use the app. Simple.
As far as I am aware, Lensa promises to delete your information within 24 hours of you uploading your photos. But you did agree to the terms and conditions, didn’t you? You may have not even noticed it, but you had to, at some point. After digging into the T&C you accepted, you can find paragraphs such as:
“We retain your Personal Data for as long as we need it for the purposes for which it was obtained or until you ask us to delete it. We also may retain your Personal Data for a longer period of time on the basis of our legitimate interests, for example, as necessary to comply with our tax and legal obligations, to resolve disputes, and to enforce our agreements. Even if we delete some or all of your Personal Data, we may continue to retain and use aggregated or anonymous data previously collected that can no longer be used for personal identification.”
So you know. Maybe your data does not really get deleted after all.
What would your data be used for, though? You have already finished using Lensa, you have your cool, colourful avatars on your phone. The photos you uploaded are virtually useless. And to you, they probably are. But, looking at a long bullet point list of reasons we can find within Prisma Labs T&C, your data can be used, amongst others:
“To provide, improve, test, and monitor the effectiveness of Prisma.
To develop and test new products and features.”
What that tells you is - you are basically paying money to the developers of Lensa to develop their product further. Your photos will most likely be input into the massive training set of photographs and art that the algorithm trains on.
Art Classes for Computers
Maybe you have asked yourself a question - how exactly do all those fancy softwares know how to paint. The thing is - they don’t. What they do is they look at millions and millions of data available to them. Where do they get the data? That one’s easy - from the same place you and I would look for it - the internet. Prisma Labs claims that Lensa itself was trained on “unfiltered internet content”. What does it mean?
In essence, Lensa could have trained on and looked at any given artwork available online - without regards for copyrights. A lot of artists have indeed expressed their concern about their data being stolen - this is basically what happened. There is virtually no way of finding out. We can only look at the styles of certain avatars and wonder - is that a copy of someone’s art project that they spent their money and time on, and now are not receiving the credit where credit is due? Maybe.
Another thing worth noticing is that the internet is biased. We are biased. People of Colour, just to give an example, often get underrepresented or misrepresented in media, and that reflects in the data available for Lensa. It is not hard to find stories of People of Colour being very disappointed with the results they got from the software, as their faces or features were, essentially, made to look more white than the original models.
All in all - please use cool new technologies. By all means. Artificial Intelligence is super fun and has a huge potential. However, let’s use it in a mindful way and use software providers we can trust. And please, above all - read the fine print.