Mistral AI
Note: This topic was mentioned during the call but not discussed in depth.
Victoria (Spain) recommended Mistral AI as an alternative to US-dominated AI models, offering more European perspectives.
Why Mistral Matters
Victoria observed:
"AI always ignores us, Europeans, the middle ground ðŸ˜"
"I can talk from experience, I always have to remind the AI to consider European options in their answers"
Her recommendation:
"@Pete Kaminski Mistral AI"
The European Perspective Problem
Victoria identified that mainstream AI models (like ChatGPT, Claude) have US-centric biases:
- Training data dominated by US sources
- Default assumptions reflect American culture
- European contexts get ignored or relegated to "other"
- Need to explicitly prompt for European perspectives
Mistral as Alternative
Mistral AI is a French AI company offering:
- European-developed models
- More European content in training
- Different cultural assumptions
- Alternative to US tech dominance
Broader Issue
Pete Kaminski acknowledged:
"It wouldn't be easy, but it could be done to collect European-centric information to train models. Europe does have a wealth of information going back centuries that would be super valuable."
Challenges:
- Much European content not digitized
- Language diversity
- Different data privacy laws (GDPR)
- Resource requirements
Connection to Curiosity
This relates to Cultural Dimensions of Curiosity:
- If AI shapes how we ask and answer questions
- And AI has cultural biases
- Then it affects what curiosity paths seem available
- Access to diverse AI models enables more curiosity
The Training Data Question
"What we need is a non-US dominated internet for our training models 😂"
This raises questions about:
- Digital colonialism
- Whose knowledge gets preserved and amplified
- Cultural diversity in AI
- Right to see your own culture reflected