𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗔𝗜 𝗱𝗼𝗲𝘀𝗻’𝘁 𝗴𝗶𝘃𝗲 𝘆𝗼𝘂 𝘄𝗵𝗮𝘁’𝘀 𝗿𝗶𝗴𝗵𝘁 — 𝗶𝘁 𝗴𝗶𝘃𝗲𝘀 𝘆𝗼𝘂 𝘄𝗵𝗮𝘁’𝘀 𝙡𝙞𝙠𝙚𝙡𝙮. 𝘓𝘦𝘴𝘴𝘰𝘯𝘴 𝘧𝘳𝘰𝘮 𝘢𝘯 𝘪𝘯𝘢𝘥𝘷𝘦𝘳𝘵𝘦𝘯𝘵 𝘦𝘹𝘱𝘦𝘳𝘪𝘮𝘦

I asked GenAI to create an image of a watch showing 𝟴:𝟭𝟱. It gave me 𝟭𝟬:𝟭𝟬.

So I tried again. And again. And again.

Six separate attempts. All showing the 𝘴𝘢𝘮𝘦 time.

And it’s not just 8:15; try asking for any time, and you’ll probably still get 10:10.

𝗪𝗵𝘆? Because GenAI isn’t pulling from real life. It’s pulling from what it’s 𝘴𝘦𝘦𝘯 𝘵𝘩𝘦 𝘮𝘰𝘴𝘵, and for watches, that’s not photos of 'real' people wearing them, it’s ads.

As Richard Shotton explains in 𝗧𝗵𝗲 𝗖𝗵𝗼𝗶𝗰𝗲 𝗙𝗮𝗰𝘁𝗼𝗿𝘆:

“𝘈𝘭𝘮𝘰𝘴𝘵 𝘦𝘷𝘦𝘳𝘺 (𝘸𝘢𝘵𝘤𝘩) 𝘢𝘥 𝘴𝘩𝘰𝘸𝘴 𝘵𝘩𝘦 𝘴𝘢𝘮𝘦 𝘵𝘪𝘮𝘦… 𝘢 𝘧𝘦𝘸 𝘮𝘪𝘯𝘶𝘵𝘦𝘴 𝘦𝘪𝘵𝘩𝘦𝘳 𝘴𝘪𝘥𝘦 𝘰𝘧 10:10. 𝘛𝘩𝘢𝘵 𝘴𝘱𝘦𝘤𝘪𝘧𝘪𝘤 𝘵𝘪𝘮𝘦 𝘸𝘢𝘴 𝘤𝘩𝘰𝘴𝘦𝘯 𝘣𝘦𝘤𝘢𝘶𝘴𝘦 𝘵𝘩𝘦 𝘩𝘢𝘯𝘥𝘴 𝘤𝘭𝘦𝘢𝘳𝘭𝘺 𝘧𝘳𝘢𝘮𝘦 𝘵𝘩𝘦 𝘭𝘰𝘨𝘰.”

But here’s the twist: These watches don’t have logos.

𝗪𝗵𝘆 𝗻𝗼𝘁?

Probably because, unlike the time, there 𝘪𝘴𝘯’𝘵 a common logo GenAI can fall back on.

So the hands are still smiling obediently at 10:10, but framing…nothing.

It doesn’t know what makes sense.

It knows what it’s 𝘴𝘦𝘦𝘯 𝘵𝘩𝘦 𝘮𝘰𝘴𝘵.

Even when you ask for something else, you still get 10:10.

The most likely answer isn’t always the right one.

But it’s the one you’ll get.