Teaching AI about war: Why democratising history leads to better AI | Nick Hodder | TEDxWinchester

Key Concepts:

• Class transition and "class homelessness"
• Museum accessibility and inclusivity
• Personal vs. "museum-worthy" history
• AI in museums: computer vision and large language models
• Data bias and narrow historical narratives
• Importance of diverse perspectives in understanding war and conflict
• AI's potential for linking information and enabling participation
• Challenges of AI: misinformation, hallucinations, and certainty vs. uncertainty
• Human element: relatable storytelling, joy, and community building

1. Class Transition and Museum Accessibility

• The speaker describes their experience of becoming middle class in 1975, highlighting the cultural differences and feeling of not fitting in with either working-class or middle-class peers.
• Example: Feeling "class homeless," being "too rough for the cul-de-sac and too fancy for the council estate."
• This experience leads to a discussion of museums as spaces that can feel exclusive and intimidating to those who don't feel they belong.
• The speaker notes the feeling of being judged for not understanding the "intellectual blanks" in museums, contrasting this with the "casual confidence" of those who seem to have inherent cultural knowledge.
• Example: Feeling pressured to know about wine and using terms like "Malbeck" without understanding them.

2. Personal History vs. Museum Collections

• The speaker contrasts the grand narratives of museums with the personal histories and memories that are often overlooked.
• Example: The speaker's grandmother's war stories and the war memorial with their surname engraved on it.
• Argument: These personal stories are just as valuable as the artifacts typically found in museums, but they are often not collected or interpreted.
• Reason: Museums have limited storage space and prioritize objects that fit existing narratives.
• Problem: This creates a gap in representation and excludes many perspectives.

3. AI in Museums: Opportunities and Challenges

• AI is being used to analyze and understand museum collections, particularly images.
• Example: Imperial War Museums using AI to identify and describe Spitfire aircraft with increasing accuracy.
• AI can help link information, find new connections, and enable participation from diverse backgrounds.
• Challenge: AI is trained on existing data, which may be biased and incomplete, leading to narrow historical narratives.
• Example: Many images in museum collections tell a "white middle class western story."
• The speaker references Yuval Noah Harari's "Sapiens" to illustrate the vast amount of untold stories in history.
• Argument: It's crucial to train AI with diverse perspectives to avoid perpetuating existing biases.

4. The Human Element: Storytelling and Community

• AI is powerful, but understanding war and conflict requires more than just technology.
• Argument: Museums need to engage with everyone, from staff and volunteers to visitors, to create a sense of belonging.
• Key: Enabling contributions and collaboration from the widest range of people.
• The speaker emphasizes the importance of human storytelling, relatable language, and humor in creating connections and building community.
• Argument: Joy and laughter can create human connections that lead to shared knowledge and a sense of belonging.
• AI can't replicate the human element of storytelling and community building.

5. Synthesis/Conclusion

The speaker argues that museums have a responsibility to be more inclusive and representative of diverse perspectives. AI can be a powerful tool for analyzing collections and making connections, but it's crucial to address data bias and ensure that AI is trained with a wide range of perspectives. Ultimately, understanding history requires more than just technology; it requires human connection, empathy, and a willingness to tell stories in a relatable and engaging way. The goal is to create spaces where everyone feels they belong and can contribute to a shared understanding of the past.