Anyone know of any examples “in the wild” of product content (e.g. PDP level, data and descriptions) and category content (including category images) using e.g. ChatGPT and Dall-E?
A/B tested against human content?
I’m aware of products that can extract structured product data from free text (Oracle has had such a product in their catalogue for about a decade, bought in naturally) but was wondering about state of the art for key merchandising activities (for own channels or marketplace/dropship channels)
(From blog author some time later) In case you don’t read the comments, the post linked above was indeed written by AI.
It did a sterling job, if somewhat bland, except it invented the Hither Green Viaduct which totally doesn’t exist.
I have no idea why ChatGPT created this nugget of misinformation, but you can now find an image of this supposed engineering marvel in Google Image Search.
Such are the dangers of ‘fake news’, which it turns out is much quicker to compose than the real thing. The future is artificial creativity, and the future is coming fast.
I was also inspired by news reports this morning of the Hollywood writers strike, in part wanting guarantees that the studios won’t use AI to create scripts
This reminds me of a story where a mapping company put a completely fake town (Agloe) on a map to prove copyright infringement, so that if any other map showed that they would know it was copied from them and not original. That happened, and the second company countered that it was real - and when the original company went there they found a town had created around the spot where they marked it on the map. (It’s a plot element in the fiction book The Cartographers, based on the real thing that happened.) So basically, Hither Green Viaduct is the new Agloe, just created by AI instead of as a copyright detection, and because it has been declared, others use that as a reference and it just becomes “real”.
The London A-Z street maps have also had “trap streets” for decades. Has become a bit of a sport to find these (although there is also enough change in London that sometimes the map is just out of date, sometimes you can see this on historical aerial photographs).