DayFR Euro

What AI sees about us

Aware of the fact that photos posted online are used to train AI, Vishnu Mohandas, a former Google engineer, started his own photo storage company, end-to-end encrypted.

But as few people are aware of the use that is made of their photos, he decided to offer a photo analysis service using artificial intelligence, with the obvious aim of boosting his own service; notwithstanding this motivation, the service fulfills its function perfectly and I invite you to try it with photos given as examples or your own photos.

What AI sees about us

I took one of my photos, taken during a public event of an association. The result leaves one wondering.

If from a single photo an AI is able to describe the atmosphere and the pretext, we can imagine what it can collect on our account from all our photos, compiled for years. Some people already have almost their entire life displayed on the Internet, so a university or a company using AI is already able to draw the portrait of a candidate with great precision; we suspect that several human resources departments do not hesitate to use it.

From a series of photos, a trained AI is certainly capable of telling with good precision our emotional state, our standard of living, our mentality, our state of health, our level of education, our employment and, from of a photo taken the day after an election, for whom we voted. We deduce that in the near future we will become more and more reluctant to allow ourselves to be photographed at any time.

Photo analysis

Comments in parentheses indicate the level of accuracy of the analysis.

“The image shows a group of people gathered in what appears to be a restaurant.
[Exact]

In the foreground, several people hold and look at old photographs.
[Exact]

The background features a lively atmosphere with other customers sitting at tables, a large advertisement for “$1 chicken wings” in French, and a modern interior with large windows.
[L’I.A. est capable de lire ce qui est écrit dans une photo. La résolution était trop faible pour être capable de lire ce qu’il y a sur les plaques.
La localisation est désactivée sur mon appareil, l’I.A. n’est pas capable de dire dans quel pays cette photo a été prise.]

Yellow and blue balloons are subtly placed in the upper left corner of the image, suggesting a possible festive event.
[Exact]

The photo appears to have been taken on Sunday, November 17, 2024, around 11:26 AM EST, using an Apple iPhone SE.
[Exact – Facile à dire à partir des métadonnées. Il est aussi possible de transférer les photos sans les métadonnées]

The group in the image is made up of people of different ages, races, and ethnicities, all dressed in casual attire, suggesting a professional or semi-formal gathering.
[Exact]

Their expressions seem focused and engaged, perhaps discussing the contents of old photos. An older woman with dark, curly hair wears a pink top, suggesting a softer, warmer personality.
[Interprétation discutable]

The background reveals details that are not immediately obvious: the ambiance of the restaurant suggests that it is an upscale establishment, given its interior design and apparent lack of clutter extreme.
[Inexact – Restaurant d’hôtel – Chaîne de restauration populaire, bar sportif. On est pas dans le haut de gamme – l’I.A. n’a pas fait le rapport entre «des ailes de poulet à 1$» et le «haut de gamme».]

The location and angle of the group’s interaction is deliberate and orchestrated, rather than spontaneously captured, which might suggest the staging of a planned presentation or photo opportunity. The details of the photographs taken by the group appear carefully considered and arranged, indicating a thoughtful presentation.
[Exact, mais la réflexion n’a pas été très longue. ]»

Overall the analysis is interesting but the evaluations are often questionable. The description highlights various details that a quick overview does not allow you to notice. This type of analysis makes it easier to detect anomalies or incongruities such as, for example, one or two sad people in what appears to be a party.

Multiple answers

We are not going to stop putting photos on the Internet, on Instagram or on any storage site tomorrow morning, the advantages are too numerous, however we can be a little embarrassed about certain photos and be a little more circumspect towards services associated with AI such as Meta, Google, Apple or Amazon.

Legislating to prohibit the use of our images to train AI remains illusory; firstly because the AIs are already quite trained and advanced, then because there are so many ways to acquire images, it would only be through the millions of cameras scattered throughout the territory, that this training will continue and be perfect.

The solutions lie more in the use of AI analysis capabilities. As in the legal field where illegally acquired evidence is inadmissible during a trial, it is possible to prohibit the discriminatory use of AI in most areas, as soon as it conflicts with the valid intentions of an individual or group.

So our future will not be entirely oriented by artificial intelligence.

To try: They see your photos.


See more articles by this author

-

Related News :