Midjourney is a proprietary model so I don't feel like there are any useful teaching moments coming from this post. You won't be able to find out how it does anything.
One interesting thing I noticed recently is their model isn't well calibrated. ie prompting with an empty string generates very specific artstation/deviantart style character portraits.
Interesting take, but we've seen ChatGPT posts spammed since the last few days here, too. But no one seems to care that a lot of applications are just APIs to a proprietary model.
Im getting tired of all the AI posts on here where the “AI” is a black box from a provider like ChatGPT/Midjourney/etc. Every post I click on I'm expecting some explanation and nothing.
Closed AI models behind a SaaS API are somewhat ‘revolutionary’, with little to no application of any serious use-case other than summarization or bullshit generation. Even when it generates its output, there is a lot of skepticism and little belief in the trustworthiness of its results.
Other than that, there is nothing revolutionary or a breakthrough in any of this justifying the hype and has failed to be more trustworthy than todays search engines. This AI hype will collapse just as how the GPT-3 hype did and just as fast as the Clubhouse hype did.
What will permanently change everything is open source and transparent AI models that are smaller and more powerful than GPT-3 or even GPT-4.
