Hacker News Reader

Those prompts are wild. It is deeply impressive that it works. Think about what would happen if you gave such instructions to a human. Would they be able comply? How big is the overlap between people who are creative enough to produce the kind of pictures Dall-E produces and disciplined enough to follow complex instructions so rigourously?

It also is just straight up impossible to convert those instructions to "regular" code.

View on HN

I can't help but feel that any perceived overlap is coincidental. An illusion similar to seeing a face in an abstract drawing that the artists, or in this case algorithm developers are keen to exploit. Our need to find the fimiliar in something that is ultimately completely alien to our way of thinking.

But with enough existing prompts and training data, it will continue to learn and better trick our senses.

I totally agree that putting those instructions into code would be outrageously complicated, and the biggest strength here is it's ability to the gist of what we are trying to convey.

But it mostly didn't comply. The overlong prompts consistently contain details that are ignored.

Reply on news.ycombinator.com