We recently announced our OpenOps framework to bring AI-acceleration to the Mattermost platform with a variety of different LLM backends, including Azure AI and OpenAI.
As part of this initiative we are moving to “Default Web Visible” (or “DWV”) for information that’s not confidential. This lets us more easily evaluate the ability of current and future LLM models to incorporate DWV data into workflows–starting with web-aware systems like Google Bard.
For example, if we’re starting an internal initiative to train our sales engineers on how to use AI features in OpenOps, we can discuss such an initiative in DWV systems, like Reddit or the Mattermost Forum, where the information can be picked up by AI models through web search rather than requiring specialized ingestion processes needed for private data.
The additional benefit of defaulting to DWV systems is the ability to be more transparent with our open source community, who can both more easily view our thinking and discussions on DWV platforms, as well as learn about our initiatives through future web-aware LLMs.
We welcome feedback to this approach across our communities. Do you think it’ll be effective? Any caveats you think we should consider? Are there other models we could learn from?
Please feel free to respond, like, report or comment,
