The AI Crackpot Index

(Apologies to John Baez)

A simple method for evaluating any claim about artificial intelligence or machine learning

  1. A -5 point starting credit.
  2. 1 point for every statement that is widely agreed on to be false. 
  3. 2 points for every statement that is clearly vacuous. 
  4. 3 points for every statement that is logically inconsistent. 
  5. 5 points for each such statement that is adhered to despite careful correction. 
  6. 5 points for each use of the terms “AGI”, “singularity”, “superintelligence”, “intelligence explosion” etc. as if they refer to specific future events.
  7. 5 points each time a claim is justified largely by appeal to popular media examples (eg “a Terminator-like scenario”).
  8. 5 points for each argument to the effect of “Artificial intelligence isn’t really intelligent” or “Machine learning isn’t really learning”. 
  9. 5 points for using the term “AI” to describe any advanced software system or algorithm, regardless of the details of its operation. (20 points for using the term “AI” to describe systems that use humans-in-the-loop to achieve core functionality).
  10. 5 points for dismissing any discussion of algorithmic bias and fairness as “woke”
  11. 10 points for every description of some AI as “aligned” without discussing what it is aligned to, or how alignment is calibrated.
  12. 10 points for each mention of Roko’s Basilisk or Paperclip Maximizers.
  13. 10 points for each mention of The Chinese Room argument.
  14. 10 points for each mention of The Simulation Hypothesis or the “multiverse” 
  15. 10 points for each use phrases such as “taking over the world” or “kills all humans”.
  16. 10 points for each claim that AI has achieved “consciousness”, “self-awareness”, “sentience”, or similar suggestions without providing definitions or evidence. (5 more points if “evidence” consists of only the most superficial behavioral similarities)
  17. 10 points for each mention of "Bostrom", "Yudkowsky", “Musk”, or any person associated with the “Intellectual Dark Web”. 
  18. 20 points for every claim to have invented a revolutionary new AI system without providing any technical details.  
  19. 20 points for each insistence that preventing eschatological (“end of the world”) scenarios should take precedence over all other practical and ethical concerns.
  20. 20 points for treating a single chatbot query as definitive research.
  21. 30 points for treating any discussion of ethical and regulatory oversight, as is common practice across many industries, as a nightmare scenario of government overreach and stifled innovation. (10 more points for comparing such oversight to Nazis, stormtroopers, or brownshirts). 
  22. 40 points for anyone using or suggesting the use of AI to classify and/or rank people by intelligence, race, sex, gender, beauty, or other social categories (eg, “criminality”). 
  23. 50 points for drawing any connection whatsoever between AI and blockchain/cryptocurrencies/NFTs or other known scams.