Hey everyone, this is Olga, the product manager for the summary feature again. Thank you all for engaging so deeply with this discussion and sharing your thoughts so far.
Reading through the comments, it’s clear we could have done a better job introducing this idea and opening up the conversation here on VPT back in March. As internet usage changes over time, we are trying to discover new ways to help new generations learn from Wikipedia to sustain our movement into the future. In consequence, we need to figure out how we can experiment in safe ways that are appropriate for readers and the Wikimedia community. Looking back, we realize the next step with this message should have been to provide more of that context for you all and to make the space for folks to engage further. With that in mind, we’d like to take a step back so we have more time to talk through things properly. We’re still in the very early stages of thinking about a feature like this, so this is actually a really good time for us to discuss here.
A few important things to start with:
- Bringing generative AI into the Wikipedia reading experience is a serious set of decisions, with important implications, and we intend to treat it as such.
- We do not have any plans for bringing a summary feature to the wikis without editor involvement. An editor moderation workflow is required under any circumstances, both for this idea, as well as any future idea around AI summarized or adapted content.
- With all this in mind, we’ll pause the launch of the experiment so that we can focus on this discussion first and determine next steps together.
We’ve also started putting together some context around the main points brought up through the conversation so far, and will follow-up with that in separate messages so we can discuss further.
Summarization is one of the things LLMs are pretty good at. Same for the other thing where Wikipedia talked about auto-generating the “simple article” variants that are normally managed by hand to dumb down content.
But if they’re pushing these tools, they need to be pushed as handy tools for editors to consider leveraging, not forced behavior for end users.
Summaries that look good are something LLMs can do, but not summaries that actually have a higher ratio of important/unimportant than the source, nor ones that keep things accurate. That last one is super mandatory on something like an encyclopedia.
The only application I’ve kind of liked so far has been the one on Amazon that summarizes the content of the reviews. Seems relatively accurate in general.
If we need summaries, let’s let a human being write the summaries. We are already experts at writing. We love doing it.
Removed by mod