AI-Plans.com Critique-a-thon September 2023: Announcement & Call for Participation
Sign up: https://docs.google.com/forms/d/1vJ7ZhOSjYKbp06alPJ1kgDWtPqlNdgnp0uBzbflvfa8
Deadline for Applications: September 19, 2023
Sign up here: https://docs.google.com/forms/d/1vJ7ZhOSjYKbp06alPJ1kgDWtPqlNdgnp0uBzbflvfa8
Event Dates: September 17 - 27, 2023
Hello, I'm Kabir, the founder of AI-Plans.com, a platform for the analysis and improvement of AI Alignment plans,
I'm excited to share that we're holding a Critique-a-thon, with a prize fund of $1000.
This event is designed to help people of any level of expertise create high quality critiques of alignment plans, and we're offering prizes for the highest quality submissions.
The final, refined submissions are reviewed by expert judges.
Judges include:
- Dr Peter S. Park, a Harvard Mathematics PhD, and MIT postdoc at the Tegmark lab
- Charbel-Raphael Segerie, head of the AI Unit at EffiSciences, leader of the Turing Seminar and the bootcamp ML4Good
- Dr Linda Linsefors, co-founder of AI Safety Camp and AI Safety Support
It’s also a great way to refine your understanding of AI Alignment and learn new arguments for and against the vulnerabilities in AI Alignment plans, since in the debate stage, you swap positions, so if your partner is more knowledgeable than you and made a lot of points you don’t have a counter for, you can now make the points and find out if there’s a counter for them.
This is one of the ways it’s designed so that specialist knowledge is not a limiting factor- the main deciders are effort, reasoning ability and the clarity and precision of your writing.
The final, refined submissions are reviewed by expert judges.
Time commitment we're looking for: at least 40 minutes each day, from the 17th to the 27th
Prize Distribution:
First place: $400
Second Place: $250
Third Place: $150
Two Honourable Mentions: $100 each
Each participant who makes submissions will also get a verifiable participation certificate.
Winners will get special certificates, which are also verifiable.
Event Structure:
September 17-18: We create a Broad List of Strengths for AI Alignment plans and improve the existing List of Vulnerabilities. We’ll draw inspiration from various resources, including but not limited to:
September 19-21: Match strengths and vulnerabilities to specific alignment plans. For each plan, label up to 5 strengths and/or vulnerabilities and quote the relevant area of the plan. Include your confidence level for each label as a percentage.
September 22-23: Pair up with another participant and debate the strengths and vulnerabilities identified in the previous step. One participant will defend the labels, and the other will question them. The next day, we’ll swap positions, with the one defending now making the case against the label being true and the one who was making the case against is now making the case for.
This will help refine the critiques and strengthen the arguments and also reduces any disadvantage from a lack of knowledge on a topic, since if your opponent know a lot more about the topic and make .September 24-25: Rotate partners and repeat the debate stage with a new partner. End this stage with a final summary write-up.
September 26-27: Provide feedback on each other's arguments. Point out any areas of improvement, such as faulty logic, questionable assumptions, or lack of evidence.
Final Stage (two weeks): Expert judges will evaluate all submissions based on accuracy, evidence, insight, and communication. Cash prizes will be awarded to the standout critiques that demonstrate top-notch critical analysis and make the best contributions.
I look forward to seeing your critiques and insights! If you have any questions, please feel free to contact me.
If you’re a researcher and would like to be a judge, please get in touch at kabir03999@gmail.com.
Apply for the Critique-a-thon here: https://docs.google.com/forms/d/e/1FAIpQLSdDCrSrrCjTJLqfkvuQcXzlwf6OILT2LPHuz3Pzu8PLTXvQhA/viewform?usp=send_form