Insight
One atomic statement made by an expert, a student or a publication, with its evidence status.
Expert Open for expert review · What do these labels mean?
Prompting for a Henning Larsen building in straw and biogenic materials returned something close to their own existing Bestseller project — the model is biased toward existing data, which makes it structurally poor at innovation.
— Paraphrased from the lecture
| Where in the source | mid — anchor: “difficult to innovate around it” |
| Confidence | high |
Context
In the part of his talk on how the studio works with AI, Jakob Strømann-Andersen (Henning Larsen) gives a “fun example” of prompting ChatGPT for a Henning Larsen building in straw and biogenic materials, to illustrate the bias of AI models.
From the transcript
This is a ChatGPT. We kind of did a prompt to do [a Henning Larsen] design building based out of straw and biogenic materials. And […] this come out of it. Maybe it’s that it’s not like you have seen some of our [straw] building that is like the Bestseller project. So it’s also biased. It used data that is there. It’s also difficult to innovate around it, so you need to understand the mechanism behind the AI engine before you can actually work with it.
Automatic transcript, presentation, middle part. Lightly cleaned: repeated words and fillers removed, omissions marked […], editorial insertions in [brackets].
Connections
Something wrong or missing? Suggest a correction