What is it?
The Uncanny Valley was a digital art project that explored how early versions of AI interpreted information. Taking news headlines as a prompt, we asked OpenAI to produce a summary image and description. At that early stage of their evolution, they often created some very peculiar results, which we showcase here, for posterity. The project ran from November 2023 to April 2024.
Why “The Uncanny Valley”?
The name comes from the philosophical concept which plots the relation between an ‘object's resemblance to a human being and the emotional response to the object’. Originally applied to robotics in the 1970s, the term seems perfect for how AI gets close to reality, but falls short in often dark and eery ways.
How does it work?
We send popular headlines from the Guardian, to ChatGPT for a text summary and DALL•E for an image. These three elements – authentic headline; AI summary; and AI image – are combined into a draft article. We then review and hand-select the most compelling results to appear on our site. Each post is linked back to the original news article. Finally, throughout the day, we automatically send a tweet for each new post.
Why is any of this interesting?
When we give a news headline to DALL•E, it can often completely misinterpret the subject or create very unfamiliar portraits of familiar faces. Plus, ChatGPT can only access information from January 2022 or earlier. So, when we ask it to summarise today's news, the best it can do is guess. In combination, this often leads to some very strange results.
Why the Guardian?
It has the most reliable API (the way for us to automatically grab it's headlines).
What tech do you use?
Which AI do you use?
We used a combination of ChatGPT and DALL•E.
Can I contact you?
We'd love to hear feedback or ideas to make this digital art experiment better - get in touch with us on X
Who made this?
Hot-Air
Real headlines, regurgitated by AI.
