From Studio Wibbly..
Putting legal ethics aside for a moment regarding how OpenAI trained its models on Ghibli and manga content, you must admit that the results are rather impressive and demonstrate how far generative AI has progressed.
When it comes to training AI on public and copyrighted content, we’ve effectively allowed search engines, since the early days of the internet’s consumerisation (circa 1993), to crawl websites and index (including copyrighted) content for users to find information. Without this, achieving the level of accessible information we enjoy today would have been extremely challenging. Personally, I wouldn’t have learned nearly as much as I know now without platforms like AltaVista back in the day—and later Google and Microsoft Bing.
There have been attempts to make search engines pay for collecting extensive snippets of information, and these efforts are ongoing. However, I struggle to see how this could work without diminishing the utility of finding information. One viable approach is monetisation through partial indexing—allowing enough content to be searchable while placing full access behind paywalls. Platforms such as Medium and Substack are increasingly adopting this model, and search engines are beginning to support such content better. It’s entirely possible that AI could follow suit (if it hasn’t already), though AI models can also be trained offline.
Let us not forget examples like Napster and LimeWire—applications that enabled illegal music file sharing and were immensely popular. What followed? The emergence of services like Rhapsody, Spotify, Tidal, Deezer, Apple Music, and Amazon Music. Even Napster transitioned into a legal streaming service for a time. These solutions transformed an industry plagued by piracy into one generating revenue for creators (albeit with complications). AI is still in its infancy—akin to Napster’s early stage.
Regardless of these debates, AI is here to stay whether we like it or not. I suspect existing search engines will eventually evolve into fully AI-driven systems with reasoning capabilities to enhance relevance.
On a lighter note, OpenAI’s latest model is remarkably adept at generating comic strips. Mild spoilers for the Apple TV show Severance follow: I found the Muppet-themed comic amusing because the system steadfastly refused to use prompts like “Kermit” or “Fozzie” but rendered results when I used “generic frog puppet” or “fuzzy bear puppet” instead. It clearly understood my intent. Interestingly, Star Wars posed challenges I couldn’t yet resolve, whereas Star Trek worked seamlessly.
With a bonus – “kind” of fixed the Muppets one. Still looks a bit odd, but it’s A.I., I can’t leave it notes.
