FAQ
-
WHAT DO YOU MEAN BY PROBLEMATIC Ai?
We define problematic AI as:• AI trained on massive datasets without permission.
• AI that competes with the original training material in the market.
• AI that disregards privacy and copyright.
• AI that fails to account for it’s climate and environmental impact.Examples:
Image generators (INCLUDING those that generate backgrounds, such as Adobe Firefly, Stable Diffusion, Midjourney, Runway ML, Dall-E)
Voice generators (ElevenLabs is a core example)
Music and audio generators (Suno and Udio)LLM and Multimodal models (ChatGPT, Claude, Devin, LLaMa)
-
DOES UNPROBLEMATIC Ai EXIST?
We recognize that artificial intelligence can have many positive aspects, providing us with tools that can make life easier for many, and advancing research in important areas.
This is:
• AI trained with permission.
• AI trained on relatively small datasets.
• AI that has specific goals, with measurable, evaluable results.
• AI that doesn’t directly compete with the original training material.
• Machine learning that is utilised for the greater good within domains such as climate, health, human rights etc.
• Machine learning and use that is considerate, taking climate and environment into account.Examples:
Motion capture in animation and games.
Translation in situations that are not subject to commerical or serious usage.
Soundproofing and acoustic isolation, object detection, classification of objects etc.
AI use in research that follow ethical guidelines based on our values of consent and democracy.