What is the Role of "AI" in the Centers for Biodiversity Informatics of the Future?

What this was about

Yoder reflected on the sociocultural impact of AI on a biodiversity informatics centre, drawing on the experience of the endowed Species File Group (about 10 staff and about 100 collaborative projects), which is considering growing from a group into an alliance. He set out ten observations: storage rather than compute is the bottleneck; standards matter more than ever; centres must spot and harden community-built AI tools; administrators, the public and researchers are ambivalent about AI; data protection walls are both necessary and futile; AI noise makes 'open' costly to maintain; and risk should be spread across networks of people. He concluded that centres are long-term havens for people who use technology, and that AI should serve life.

Why it matters. The talk offers a practical, non-technical view of how AI affects the running, funding, security and mission of long-lived informatics centres.

In the room

• The Species File Group has been endowed for about 15 years, has about 10 full-time staff and about 100 collaborative projects, and is exploring a move from group to alliance to share risks and benefits.
• Mathematics, where AI output can be verified with the Lean system, shows that principles proposed for organisations in response to AI are largely just good principles for any centre.
• The biggest bottleneck is storage, not compute. Aggregating data into commercial clouds enables AI models that communities may resent, so the community needs shared or 'infinite' storage.
• Standards are more important than ever: to parse data with AI you still need a standard to parse into. Standards are 'a sea of stability' in the AI wave.
• Citizen scientists and taxonomists now build long-needed tools with LLMs; centres must notice them and harden them for the long term.
• Administrators are enthusiastic but visionary rather than practical; social media and some colleagues (e.g., journal editors) are hostile, and corporate partnerships are contested.
• Walls and silos are inevitable, required and pointless: data leak eventually, so hiding protected-species locations is futile, yet Cloudflare-style walls are now ubiquitous.

Notable moments