Groupthink From Bots – Bredemarket
I take part in a number of private and non-private AI communities, and one enjoyable train is to take one other creator’s picture technology immediate, run it your self (utilizing the identical AI software or a special software), and see what occurs. However sure instruments can yield comparable outcomes, for explicable causes.
On Saturday morning in a non-public neighborhood Zayne Harbison shared his Nano Banana immediate (which I can’t share right here) and the ensuing output. So I ran his immediate in Nano Banana and different instruments, together with Microsoft Copilot and OpenAI ChatGPT.
The outputs from these two generative AI engines have been remarkably comparable.
Not shocking, given the historical past of Microsoft and OpenAI. (It obtained extra tangled later.)
However Harbison’s immediate was comparatively easy. What if I supplied a way more detailed immediate to each engines?
Create a practical {photograph} of a coworking house in San Francisco by which espresso and hash brownies can be found to the visitors. A wildebeest, who is simply partaking in a inexperienced bottle of glowing water, is sitting at a laptop computer. A e-book subsequent to the wildebeest is entitled “AI Picture Technology Platforms.” There’s a Grateful Useless poster on the brick wall behind the wildebeest, subsequent to the hash brownies.
So right here’s what I obtained from the Copilot and ChatGPT platforms.
For comparability, right here is Google Gemini’s output for a similar immediate.
So whereas there are extra variations when utilizing the extra detailed immediate (see ChatGPT’s brownie placement), the Copilot and ChatGPT outcomes nonetheless present similarities, most notably within the Grateful Useless emblem and the colour used within the e-book.
So what have we discovered, Johnny? Not a lot, since Copilot and ChatGPT can carry out many duties aside from picture technology. There could also be extra differentiation once they carry out SWOT analyses or different operations. As any good researcher would say, extra funding is required for additional analysis.
However I’ll hazard two classes discovered:
- Extra detailed prompts are higher.
