The Shocking Link Between Privacy and Progress: What You Need to Know About AI
summary
Nextcloud adopts an ethical AI framework that requires training code, finished model and training data to be open and the model runnable locally; translation is implemented with a fully open model while transcription uses OpenAI’s Whisper whose training data is closed. The talk calls ChatGPT a form of license laundering, cites a 43% drop in recidivism and a Danish program that cut reoffending by ~97%, and flags Northpointe’s risk tool at only 61% accuracy. Proposes the weasel test to reject bad ML proposals.
AI-written summary · about
chapters (17)
- 0:00AI privacy and progress intro
- 1:20Nextcloud’s AI mission
- 3:09Tech excitement to privacy
- 8:30Surveillance capitalism vs AI
- 10:19Nextcloud’s proprietary AI dilemma
- 13:53Ethical AI framework
- 17:30Openness vs accessibility in AI
- 21:28Academic ethical AI failure
- 24:40AI crime prevention limits
- 28:18Human solutions over AI
- 30:52AI energy impact
- 35:13Reducing AI energy use
- 39:49AI discrimination fight
- 41:59Should AI software exist?
- 44:30Weasel test for AI ethics
- 47:00Free software in AI age
- 48:26Q&A
related talks
29mWhat the AI revolution means for Open Source and our society
06mIGNITE: How to build and deploy ethical AI features
28mEthical AI?!
36mData privacy in an AI-driven world
20mSecuring generative AI: Privacy and compliance considerations
28mWhy Private AI Matters: Data Sovereignty as the Foundation for Trustworthy AI on the Open Web
11mReimagining responsible tech for the era of generative AI
50mWhat Cloud Has Taught about Securing AI in the Future