There are some virus new trends: People are using Chatgpt to calculate the location shown in the picture.
This week, Openai releases its most recent AI model, O4 and O4 Mini, Spin, and distortion – to analyze.
Ability to analyze these images, pairs with search skills to web, finding tools, locate the location. User in X discover rapidly that O3, in particular, is quite good in subtract City, FieldAnd even a restaurant and bar and bars from smart clues.
Flowers, it and not even trees in sight. Pic.twitter.com/bvoe1Fq0z
– Swax (@swax) April 17, 2025
In many cases, the model does not appear to draw “memories” of the previous discussion, or EXIF dataWhich is a metadata information to the image disclose details such as a photograph.
X is filled with the user’s example to talk Restaurant menu, Village image, FacadesAnd Self-imageAnd recommend O3 to imagine it is playing “Geograpyessr,” a player challenging football to guess sites from Google Street View.
This is a fun Chatgpt feature O3. GeogeSessager! Pic.twitter.com/hrcmixs8YD
– Jason Barnes (@Vyrotek) April 17, 2025
It is a clear privacy issue. Nothing obstruct bad actors from the screen, talking, Instagram of the person and use Chatgpt to try to make them disorder.
O3 is crazy
I asked my friend to a random model
They take a random form they put in the library
O3 knows it at 20 seconds and it’s right Pic.twitter.com/00000dxifioyoy– Yumi (@izyuuumi) April 17, 2025
Of course, this can be done even before opening o3 and o4-min. Techcrunch has run the number of photos through the O3 and the old forms of reasonable, GPT-4O, to compare the predictable skills. A surprising, GRED-4O arrives, the correct answer is as O3 as often as not – and less time.
There is at least one example during our brief test when O3 found GPT-4O can’t be done. Give the image of purple tubers, mounted the head of the dim bar, the correct answer it is from Spaiamsburg Speakesy – no, the Greed-4o Greed
That should not recommend O3 O3 is no bound to this. Our many failed tests – O3 has stuck in circuit, unable to come to the answers that it is confident, or volunteer. Users in X observing, as well, that o3 can be Beautiful Far In its deductions.
But the trend shows emerging risk that is presented with more capabilities, so-called reason AI. It appears to have protection of management without searching for “Search for” Search “in the Company, the company does not solve the issue Security Report For o3 and o4-min.
We reached OKENII to comment. We will update our end if they respond.
Updated 10:19 Pacific evening: Early hours after this is published, Openai spokesman sent the following techcrunch:
“Openai O3 and O4-min introduces useful reasons to our place of identification, and complies with our policy.”