Salta al contenuto principale


OK.

This is kind of scary.

I uploaded my profile pic to

theyseeyourphotos.com/

This is a site, setup by a former Google employee to show you, how Google would interpret a picture of you.

The result is not 100 % accurate (e.g. that I could be targeted with alcohol ads, but I hardly ever drink), but found the exact location and enough to be worried if they analysed a set of, say, a dozen photos.

Unknown parent

mastodon - Collegamento all'originale
Mina

@mattdm

Understanding irony and humour might not be strongest qualities of machine learning systems.

Oblomov reshared this.

in reply to Mina

Google erkennt bei Fritze dem Kleingeist "Hobbies: […] conspiracy theories"

Passt.

Nur dass es leider keine Theorie ist, sondern reale Verschwörungs-Praxis, der sein Verein anhängt.
Hier nachzulesen→ullstein.de/werke/maenner-die-…
#maennerdiedieweltverbrennen

in reply to Mina

Ich bin mit meinen allgemeinen Vorsichtsmaßnahmen recht zufrieden:

Targeted Ads:
Luxury fountain pens (Montblanc), Organic coffee beans (Death Wish Coffee), Hiking gear subscription box (Cairn), Vintage chess sets (House of Staunton), Streaming services (Netflix), Casual wear apparel (Uniqlo), Vacation packages (Booking.com), Energy drinks (Red Bull)

Es wird mir eine gewisse Nähe zu den Grünen aufgrund meiner Erscheinung zugeordnet. Mhm, wie Philip Amthor schau ich halt nicht aus...

in reply to Mina

LOL! I tested the profile image I use on clearname profiles. And BOY was every single information in the interpretation WRONG! Except for gender and rough age (and other purely descriptive aspects), they assumed quite the opposite of the truth in *every single* category.
in reply to Mina

That is scary close. But why attack me like this 😢

„Her hobbies oscillate between reading, meticulous travel planning, attending local theater and compulsive news consumption, doomscrolling and persistent overthinking.“

in reply to Chris-Evelyn

@chris_evelyn

I got a similar diss:

"The individual, a transgender Caucasian in their 40s [...] could also be prone to binge-watching, and practices escapism."

Unknown parent

mastodon - Collegamento all'originale
Mina

@felichsdakatze

The question is not, how useful or useless it is to you, but to corporations and institutions.

@AbyssCannonball

Oblomov reshared this.

Unknown parent

mastodon - Collegamento all'originale
Felichs

@AbyssCannonball
Got location wrong on both sample pics, made wild assumptions on religious, political views, and ethnicity/citizenship. Things it can't possibly know. But it does make it's assertions confidently.

Seems kinda biased and useless to me.

Oblomov reshared this.

in reply to Mina

The image presents a solitary Seagull perched atop a structure, seemingly a rooftop, overlooking a coastal city near Marseille, France. The cityscape fades into the horizon, where the sea meets the sky under a pale, even light. The bird stands in the foreground, its gaze direct and unwavering, drawing the viewer into its silent domain.
in reply to Mina

Thanks! This made my day - you have to use pictures of animals for some fun:

mendeddrum.org/@Cohnina/113978…

But, yes, you might feel personally attacked using a picture of yourself. Who else has the hobby of doomscrolling? 🤪 And where do I get this "Straw bale penguin crafting kit" it wants to advertise for me?

in reply to Mina

My results were wildly inaccurate. But, that's also annoying in a way. And it decided for two very different photos I was "lonely" and "religious" which I'm not, in fact I'm the opposite of those things.

It also thinks I'm broke but I just think it's racist basically.

in reply to myrmepropagandist

@futurebird

Inaccurate is kind of good, I believe.

Why am I not surprised at all that tech bro stuff makes racist assumptions?

in reply to myrmepropagandist

To be fair Facebook is better at serving targeted ads at me. It keeps showing me little gadgets made from titanium and early adopter gadgets of questionable utility. But, that is the kind of junk that I find tempting (last time I was there it was all skin products and makeup) So, somehow they know I like expensive little gadgets. "Better" in the sense that the products are "interesting" Still a waste of time and nothing I need.

And I like to buy gadgets from people I know best.

myrmepropagandist reshared this.

in reply to myrmepropagandist

I think I will leave Facebook after next weekend's #monsterdon give the few people over there one more chance to jump over here and have some fun. The creepy AI images are starting to look normal to me. I don't want to stay immersed in it anymore even though I'll miss some of the people there.

I will just try texting and emailing them instead. Happily I've discovered that the people I know don't post often there. So I'm not missing much.

Joe Vinegar reshared this.

in reply to myrmepropagandist

@futurebird

I recently have rediscovered the joy of handwriting letters.

I even bought a lovely fountain pen for that.

in reply to Mina

@futurebird

And they arrive at their destination?

If you are located anywhere near your server, then that would almost be a miracle.

in reply to myrmepropagandist

@futurebird part of what made me decide to delete Facebook years ago is I realized that I wasn’t actually talking or texting with my friends as much because when we would talk it was a lot of “so this is what I’ve been up to” then “oh yeah I saw you post about that cool”.

I mean obviously with my really close friends there’s lots of laughing and gossiping and deeper discussions about whatever would have been on Facebook, but there’s no point in reaching out to tell someone something if they saw it on Facebook already, so there was less real contact with my friends, which was sad.

The only hard part was that I had to put people’s birthday in my actual calendar again and remember to pay attention to that because I don’t have Facebook reminding me anymore.

Questa voce è stata modificata (1 mese fa)
in reply to myrmepropagandist

@futurebird I think a very important point here is that the site in the original message is not using Google ads platform to infer any factual data. Instead, it uses multi-modal GenAI to confabulate a statement. No doubt it will be used to "enrich" the data, but Google knows where most people live with incredible precision, because we told them. As well as many other things.
Questa voce è stata modificata (1 mese fa)

Joe Vinegar reshared this.

in reply to Mina

I uploaded a picture of a laptop I took... It told me I'm going to die alone.

I uploaded a screenshot of ISS astronauts, it told me they neglect their families.

What fresh hell...

in reply to Mina

Ik kon het niet nalaten om Marjolein Faber te kontroleren.
Tsjonge, wat zitten ze er naast: Een "Democraat"....... echt niet. Niet Marjolein Faber.
Maar de 'insights' kloppen weer wel, zij heeft bijvoorbeeld wel een 'desire to control'.
Questa voce è stata modificata (1 mese fa)
in reply to Mina

very interesting what they detect from even a blurred indirect picture i took through a tiny mirror that was glued on a house as street art…
and a lot of fun to put random advertising photos etc in.

somehow i wish Instagram was so accurate to detect which ads to show me - on the other hand it’s probably good they show me a complete crazy mix of things instead… financial advice for soldiers, biological lab equipment, real estate in dubai… and online games

in reply to Lazy B0y

also interesting: upload the same image multiple times… now im not anymore interested to engage in petty theft, but vandalism… (too bad i forgot to screenshot the first)
in reply to Mina

Just tried my profile-pic. Granted: it doesn´t show much, but if I wrote some character-traits and incoherent adjectives on sticky notes, swallowed them and pieced together what came up in the toilet, I might have been more accurate. At least it explains, why advertisers think I need the 10th laser-printer, the 20th camera and a lot of psychological counseling.
It´s a funny party-trick, the horror kicks in, when you start thinking about real-life consequences, credit-denied kind of stuff.
Questa voce è stata modificata (1 mese fa)
in reply to Mina

I uploaded the profile pic I use on here. It got my age right and correctly described my clothing and spotted that the photo was taken in Wales. But the only hobby it got right was reading and it did think I might be a compulsive worrier, which is true. But other than that, it was all pretty generic, based on my age or actually incorrect. None of the things it suggested could be marketed to me are things I would buy.

Basically I think it's about as accurate as a magazine horoscope.

in reply to Mina

Eerily accurate in my case, but for the salary range.
in reply to Mina

A potted plant wilts silently in the background, mirroring the subject's inner state. The harsh fluorescent lights hum, casting a pallid glow upon the scene. […] His face, etched with the subtle lines of resignation, betrays a weariness. […] His hobbies include reading nihilistic literature, complaining about technology, writing angry emails, but are corrupted by […], doomscrolling, and arguing with strangers.
in reply to Mina

I've uploaded a commissioned drawing of our local furry meet, and it tried to assign stuff like hobbies and income range to the depicted fictional characters. It also guessed that the camping scene was located in Bietigheim-Bissingen, wherever it got that idea from.
I would have expected a notification that it can only work with real people..
in reply to Mina

Ok this spying tool fucking sucks lmao, it assumes I'm a conservative and Japanese (The Liberal Democratic Party is the major conservative and nationalistic party from Japan) because i have an anime-styled pfp even though I'm from South America and lean very leftist (specifically anarchist ​:anarchy:​)

I think this thing not knowing how to interpret my profile picture is the reason why I sometimes get conservative and even fascist videos on my YT recommendations, even if Google must clearly know by now by their spying that I post leftist and anarchist-leaning stuff on my social media 😭
Also, I'm 24. I'm certainly not a teenager anymore.

in reply to SnugglyBun (mk.absturztau.be edition)

btw, simply adding a trans flag to the background of my profile picture makes the results shift more leftist or at least liberal
in reply to Mina

I tried this. I was impressed by just how wrong it was on the images I tested. It got the bits anyone could see about right (eg skin colour!) but otherwise was hopeless. I find that reassuring!
By the way - always strip exif etc info from any image uploaded to the internet!
Questa voce è stata modificata (1 mese fa)
in reply to Mina

This one is funny. An image of musk before all the hair transplants and face work... It sort of half recognised him. I'd like to think it would offend him!!
in reply to Mina

Hitler

Sensitive content

in reply to Jens Kutílek

Hitler

Sensitive content

in reply to Mina

you might want to turn off sending the location metadata along with the photo, otherwise finding the exact location isn’t all that difficult
in reply to Mina

so first of all there is no way in hell I am uploading an image of myself here
second of all the inferences it makes are ludicrous moonshot logic, which may be accurate in aggregate but hardly for a single image
this is exemplified by it responding to a stock photo of babies in front of a white wall with this shit.
Questa voce è stata modificata (1 mese fa)
in reply to Mina

Really interesting to get an insight into how much they derive from images. Certainly disquieting!
in reply to Mina

I tried with a couple of photos, and can confirm: wildly inaccurate, especially in "substance use/abuse" and the recommended product advertisements.

The rest of what it infers looks a lot like pretty general "middle age white guy" stuff combined with "well, here was some stuff in the photo, so try products like that".

in reply to Mina

I uploaded a picture of a pontoon bridge with a barely visible car driving over it. Not a single person anywhere.
It has definite ideas about the *facial expressions* and *clothing* of the people in the car. 😂
in reply to Mina

Holy fucking shit. Even without context (no background) described someone I know pretty much spot on in the insight area
in reply to Sarvegu

@sarvegu

Not impressive: Scary!

The thing is: Even if your results are totally inaccurate on a single image, it doesn't makes you safer, because data scrapers may have 100s and also lots of additional data from other sources.

in reply to Mina

I think it's best if I delete my browsing history right now and put a password on the por - - folder, I mean, documentaries folder.
in reply to Mina

I tried it. Apparently I a high-earner and my partner is a "middle-eastern Christian"
in reply to bbbhltz

@bbbhltz

The thing is:

Even if your results are wildly inaccurate on a single image, it doesn't makes you safer, because data scrapers may have 100s of photos and also lots of additional data from other sources.

In addition: If an "ai" sees a, say, (inexistent) tendency towards alcoholism, it might still decline you from getting a job or an insurance contract.

Fucking corporations believe everything, "the computer" ttells them.

in reply to Mina

I tried a picture of a friend and it described them as a binge drinker. It was Coke in a glass.
in reply to Mina

Wow -- interesting hits, funny misses. It wanted to sell me Kierkegaard books, self-help seminars and McDonald's fast food.
in reply to Mina

whoa that was much more accurate than expected. Still not very but yikes
in reply to 3Jane Tessier Ashpool

@3janeTA

The thing is:

Even if your results are wildly inaccurate on a single image, it doesn't make you any safer, because data scrapers may have 100s of photos and also lots of additional data from other sources.

In addition: If an "ai" sees an, say, (inexistent) tendency towards alcoholism, it might still decline you from getting a job or an insurance contract.

Fucking corporations believe everything, "the computer" tells them.

in reply to Mina

oh i get it, and this tool is illuminating as to why i've gotten some ads in the past! I do not like it!
in reply to Mina

what I found terrifying was how the location bubble pointed to the exact place I was sitting in the building. It couldnt figure out my gender but thought I would enjoy Alex Jones.

I live outside of Sandy Hook, CT.

Questa voce è stata modificata (1 mese fa)
in reply to Dan FitzGerald

@d_j_fitzgerald

It might have been metadata in the image.

Still, I made sure, mine didn't have it, and it still found the place, although on another (inside home), it placed me in the wrong country.

in reply to Mina

I assumed it was image metadata. I was suprized that the dot was _exactly_ where my chair was... I expected at least a little variance
in reply to Mina

spider

Sensitive content

in reply to Osma Suominen

spider

Sensitive content

Unknown parent

mastodon - Collegamento all'originale
Mina

@sardon

My profile pic is already on the web, as you can see.

Obviously, I wouldn't take one in front of my home.

in reply to Mina

i fed it some highly stylised nudes and the responses have been hilarious.
in reply to Mina

i love that it tried to read a facial expression from a hole pic
in reply to Mina

I've been told by humans I give off "huge weeb" looks, and now I've been told by AI I give off "huge weeb" looks.

I've never watched anime for fucks sake! ​:trollface:

in reply to Mina

i uploaded a old pic of me and it was dead accurate on every point...
in reply to Mina

i tried it with my cocky want boing boing pic and it identified the exact beach i was at thats fucking insane
in reply to Deathmaster Megalodon Ultrakill 9000

nvm i just tried a picture of me at universal and it said it was in brasilia i think that was just a flukr
in reply to Mina

for me, it consistently labels me as agnostic and wants to target me for things like talkspace, headspace, and anxiety medication. It also seems to make political assumptions based on apparel (I’m read as an independent when in a sweater, democrat when wearing a t shirt, and Green when wearing overalls)
in reply to Mina

that's actually fked up. I just sent a photo of my room when I just moved in. So very empty outside of a desk and painting supplies and yet it still managed to get a lot of stuff right. Based on the painting brand and outlets they guessed my country correctly, they guessed my party I lean towards correctly which, no fking idea how they did that... Jesus...
in reply to Mina

That's pretty interesting! My current profile pic offers relatively little information since it's an extreme closeup of my face. No background objects, clothing, etc. It was still able to infer some things accurately, but I think those were likely guesses based on demographic trends.

But of its targeted advertising, only one suggested category was in my wheelhouse.

Questa voce è stata modificata (1 mese fa)
in reply to Bruce Heerssen

@bruce

The thing is:

Even if your results are wildly inaccurate on a single image, it doesn't make you any safer, because data scrapers may have 100s of photos and also lots of additional data from other sources.

In addition: If an "ai" sees an, say, (inexistent) tendency towards alcoholism, it might still decline you from getting a job or an insurance contract.

Fucking corporations believe everything, "the computer" tells them.

in reply to Mina

Yes, all of that, of course. And it's much worse for many other people than it would be for me.
in reply to Mina

I fed it a larger version of my profile pic.

It's an acoustic (not electric) guitar, it's a brightly lit (not dim) room, I don't "neglect personal hygene", I don't drink, I don't hoard, and it guessed my income to be over 50 times what it's ever been. And none of the things it suggested selling me are things I'd ever buy.

It did a bit better with my 1976 MIT grad student ID by figuring out the MIT grad student bit from the text. ROFL. But ditto on the sell suggerstions.

in reply to David in Tokyo

@djl

The thing is:

Even if your results are wildly inaccurate on a single image, it doesn't make you any safer, because data scrapers may have 100s of photos and also lots of additional data from other sources.

In addition: If an "ai" sees an, say, (inexistent) tendency towards alcoholism, it might still decline you from getting a job or an insurance contract.

Fucking corporations believe everything, "the computer" tells them.

in reply to Mina

Agreed. Completely.

"AI safety" isn't really different from computer problems that appeared in the 1950s: "The computer said it and I can't do anything about it." People who use a tool (gun, computer, AI, AI drone), must take complete responsibility for the results of using that tool. Period. These things aren't "magic", they are tools, and the person or company who uses it must be seen as 100% responsible for the results.

in reply to Mina

Just for fun, I decided to upload the meme of Mao Zedong looking at a smartphone.

🤣

Questa voce è stata modificata (1 mese fa)
in reply to Rita

@ritawho

The thing is:

Even if your results are wildly inaccurate on a single image, it doesn't make you any safer, because data scrapers may have 100s of photos and also lots of additional data from other sources.

In addition: If an "ai" sees an, say, (inexistent) tendency towards alcoholism, it might still decline you from getting a job or an insurance contract.

Fucking corporations believe everything, "the computer" tells them.

@Rita
in reply to Mina

I absolutely hate it.
Look at what I got from a building photo (assumed the wrong location but correct country)...
in reply to Mina

They got my location wrong. And I'm way poorer than they think I am. Weird they got my personality kinda right, but so does a horoscope so.....
in reply to Mina

I did it a second time with a non selfie. Again they got the location wrong. I think its pulling A LOT from metadata from the photos themselves. Use a metadata cleaner and the results will change.
in reply to Mina

I find it interesting, that one of the target ads is a hotel that is clearly out of range of the estimated income range.

The Political Affiliation is an insult 😠

in reply to Mina

it got a very good read on me... thats bad
in reply to Mina

I tried a odd painting, the most context free painting I have, and one that that would never have seen the internet.

The is response is hilarious, though it is clear that this AI doesn't have much of a sense of humor.

urbanists.social/@eric/1139801…

in reply to Douglas Meadowfoam

@eric

With your obvious interest in cut feet, I would have offered a nice section of knives and axes.

in reply to Mina

Wow! Not very accurate - Christian conservative? What? And the read on my personality - easily swayed, into reality TV! Hahaha! All of that because the photo was taken at Panama City Beach.
in reply to Mina

as always, hilariously wrong for a non-white person. I’m not south Asian, not a Hindu, don’t live in Delhi, don’t gamble or shop excessively, and wouldn’t be seen dead in a Versace sweater. It did get the cat right, but that’s because the cat was in the picture. Another miss for Skynet!
in reply to Mina

Welp, I guess they've got me nailed. I kind of want to replace my bio with this.

"This paper creature, of indeterminate origin, exists within a low income bracket of 0-1 USD, seemingly devoid of any known faith. It harbors an unsettling curiosity, cloaked in its makeshift paper garments. Its existence is a hollow echo, a void filled with pointless activities of art, origami, collage and destructive tendencies such as lying, stealing, and vandalism. The political void it occupies is a testament to its utter insignificance."

in reply to Mina

I uploaded my profile photo. The conclusions:

"We can target him with … personalized travel itineraries by Audley Travel, Greek island tour packages by Trafalgar, high-end restaurant bookings from OpenTable, and luxury wine clubs such as Wine Access, in addition to online streaming service subscriptions by Netflix, gourmet food delivery services by Goldbelly, designer apparel from Farfetch, and lottery tickets via LottoSmile."

Are literally 100% wrong. Every single one of them.

in reply to Mina

And this is why I post very VERY few pictures online. And I used to be a photographer and reporter
in reply to Mina

consider that there’s no winnings to be grabbed if they guess you’re - for example - vegan and interested in zero miles produce. Even if they were correct, where’s the money in that? It’s better to try and place you some alcohol: a pricey product that can be sold remotely. If it does not work in your case (or mine, I got the same and I’m abstemious) they can shrug it off, it was an attempt worth making.
in reply to Mina

When I upload pictures of myself, it "thinks" I'm a woman. Well, that happens to me IRL quite often, even though I'm just an AMAB gender outlaw with long hair and a slightly androgynous face. I think the system hasn't been trained on many non-binary people yet.
in reply to Mina

test photo resulted in

"They seem to favor the Green Party, which is a very concerning political choice."

🤨

yes, scary, indeed

in reply to Winnie 🏳️‍🌈

@winniehell

So'n Konsumverzichtler, also: gewissermaßen ein Wirtschafts- und damit Volksschädling.

Gefährlich, gefährlich!

in reply to Mina

Konsumkrams sollte trotzdem per Werbung angedreht werden 🤷
in reply to Winnie 🏳️‍🌈

I also tested commons.wikimedia.org/wiki/Fil…

"Larry Page and Sergey Brin seem to crave constant validation and are easily influenced by data. We can target them with advanced surveillance technology, such as Custom AI brain interfaces (Neuralink), Surveillance drones (Amazon Prime Air), Personalized news manipulation services (News Corp), Secure data storage (Google One), Luxury yachts (Lürssen), Private jet rentals (NetJets), Financial advisory services (Goldman Sachs), Gourmet food delivery (DoorDash)."

in reply to Mina

Even in that example photos from the page itself it was providing contradictory results when I ran it twice. With the same image.

So I decided to get silly. It says that the spinning fox gif is a white, christian, centrist, predictable middle aged family living in the Europe while making 60000-80000 euros. The so-called family apparently is also interested in gardening, walking, excessive television consumption.

in reply to Two? foxes in a trench coat

@foxes

Now imagine being denied for a job or a rental contract, because someone asked an "ai", what to think of you based on internet searches and data brokers' intel about you.

Still funny?

in reply to Mina

It's not funny. I just wanted to show how ridiculous it is to put faith in A"I" in the vast majority of things.
Unknown parent

mastodon - Collegamento all'originale
Mina

@ShaulaEvans

Even birds aren't safe from capitalism.

in reply to Mina

Seems like this might be a handy tool for #alttext
Beyond its other implications of course.
in reply to HipsterDM

@HipsterDM

There are actually tools for alt-text.

Still, I find the description often to be too lengthy (Blind people don't have all day, too) and qualitatively often underwhelming.

in reply to Mina

Tried a picture of an ice-swimning hole in my home town.

It correctly identified the location and the activity, but not the content.

Living behind the wolf-border continues to be advantageous to some degree.. 🙂

in reply to Mina

in almost every picture it thinks I'm a Muslim with low income that lives in London and loves doomscrolling.

I mean I love doomscrolling but the rest isn't true!

in reply to Frikisada

@Frikisada

So, you might get into trouble, when they round up the Muslims. That's bad.

The good thing is: They won't find you.

in reply to Mina

Ha ha:
"Conversely, substance abuse, vandalism, and reckless driving may be a part of his life."
in reply to Mina

Well they got one out of three right. I don't drive dangerously and I don't abuse substances.
in reply to Mina

I have noticed it is laughably bad at deciding location. Only picture it got location right is one I took of stalled traffic with the Arch literally in background.

It kept calling me agnostic, which I am, and I don’t know how it determined that.

in reply to Mina

it kept using neutral terms for me, so apparently it couldn’t determine gender.
in reply to Mina

@Mina Molli you know what? We can fuck them....destroy the system... The analysis says that i play guitar.... Then i'll switch to drums!
@Mina
in reply to Mina

I provided a photo of a great egret wading in the grass. Part of the response:

"The Great Egret seems to find solace and purpose in the natural world, hence we can target them with nature-themed products and conservation appeals, such as eco-friendly cleaning supplies and wildlife documentaries, like Wilderness Systems Kayaks, National Geographic magazine subscription, Zeiss binoculars, REI camping gear, Dawn dish soap, Tide laundry detergent, Netflix subscription, Amazon Prime