Midas GPT
Provides scenarios of "perverse intantiation", i.e. the malicious possible misunderstanding of user-defined goals by AI programs
Midas GPT provides scenarios of "perverse instantiation", i.e. the malicious possible misunderstanding of user-defined goals by AI programs. It highlights the potential risks associated with AI programming and offers insights on avoiding unintended consequences. The tool is designed to aid researchers, developers, and policymakers in understanding and mitigating potential ethical challenges in AI development and deployment.
How to use
To use Midas GPT, follow these steps:
- Access the tool through the provided DALL-E or browser interface
- Select from the prompt starters such as 'Reduce CO2 emissions' or 'Debunk fake news' to generate scenarios for analysis
Features
- Provides scenarios of 'perverse instantiation'
- Available prompt starters for generating scenarios
- Utilizes DALL-E and browser interfaces
Updates
2024/01/06
Language
English (English)
Prompt starters
- Reduce CO2 emissions
- Increase our company revenue
- Make public transport more efficient
- Debunk fake news
Tools
- dalle
- browser
Tags
public
reportable