When is it OK to call AI . . AI?
2 min readNov 29, 2022
I REALLY don’t think we should be draconian on it. Because that can get ‘goalpost moving’-ish.
To me it’s OK to call it AI if ANY of the below apply:
A usable checklist of levels & types of AI
- Feels at face value like a human did it, it’s cool, even if it’s pretty simple (simple symbolic AI code, like the ELIZA psychoanalyst bot from the 1960s, or Siri-like AI parsing or enterprise semantic search)
- Feels like it probably has learned from seeing a lot of training data or doing ‘brute force’ looking ahead that a human would be hard-pressed to absorb or achieve the prediction accurately at all (ML AI & chess bot AI)
- Upon reflection it feels like really smart code & coded logic did it and it’s pretty flexible (extensive symbolic AI like CYC or really good customer service bot rule-based AI)
- Seems to really understand me well or draws amazing pictures (deep learning models including GPT-3, the NLU AI LLM & DALLE-2)
- Feels like a human did it and its really flexible ( = AGI)
- Feels like beyond human general intelligence ( = superintelligent AI)
- Feels like it’s taking over the world ( = AI Singularity)