steam engines, electricity, cars, trains, planes: revolutionary tech always feels scary and unfamiliar at first but over time, we learn to contain it, harnessing it’s power to change the world AI is no different – it will be contained. http://
the-coming-wave.com
ETHICS
-
AI Will Be Contained Like Previous Revolutionary Technologies
By
–
-
Misuse and Misunderstanding of “Reasoning” in AI Context
By
–
In the context of AI, "reasoning" has got to be one of the most misused and misunderstood term ever, right up with "intelligence".
-
US AI Regulation Gaps: Misinformation, Impersonation, and Bias
By
–
not as familiar with UK law, but lots of gaps in existing regulation in US (eg what to do about wholesale misinformation, what to do about machines that misrepresent themselves as people, bias from uninterpretable systems, etc)
-
Two Models of AI Oversight and Governance
By
–
https://
garymarcus.substack.com/p/two-models-o
f-ai-oversight-and-how
… -
Speaking Against Regulatory Capture in AI Policy
By
–
I spoke against regulatory capture there, explicitly, and suggested some ways of approaching it. so you are misrepresenting me.
-
Social Media Software Should Be Regulated Like Critical Systems
By
–
social media is software and it should have been regulated. airplane guidance software has verification, because lives are at stake. we don’t exempt it because it’s software.
-
Scan AI Models for Data Leakage, Bias, and Performance Issues
By
–
You can now scan your AI models to detect any of the following: • Data leakage
• Biases
• Performance issues
• Spurious correlations
• Overconfidence issues
• Underconfidence issues With only 4 lines of code! Check this open-source library: https://
github.com/Giskard-AI/gis
kard
… -
Creator and Master: AI Ethics Through Shelley’s Vision
By
–
“You are my creator, but I am your master” — Mary Shelley
-
Cruise’s Driverless Cars in Austin Not Ready for Scale Deployment
By
–
20+ @cruise driverless in Austin, TX. [Thanks to Tandy Trower & @garymarcus
] This system is not ready for deployment at scale, and it is not a good business model to be subjecting people going about their lives to disruptive experiments that piss them off. -
LLM Model Alignment: Behavior Customization for Commercial Products
By
–
Alignment is aligning the model's outputs with how you want it to behave! It’s a very complex problem with lots of special cases, but also a very important one when commercializing a product based on LLMs! #ai #llm #llms #whatsai #artifcialintelligence