⨑ MOOD::0

↻ HABITS [Meditate::] [Exercise::✅] [Personal Work::✅]

Tuesday, January 10, 2023

<< Yesterday | Tomorrow >> Retrospective:: Last Month | Last Year

📅 Daily Questions

🌜 Last night, after work, I

🙌 One thing I’m excited about right now is

🚀 One+ thing I plan to accomplish today is

• [ ]

👎 One thing I’m struggling with today is

📝 Ephemeral Notes

• There has been a dichotomy of approaches to AI: Statistical approach and Symbolic approach
  • The statistical approach is what is used by ChatGPT
    • It pulls patterns from the training data and uses it to generate very “human” like texts
    • However, it doesn’t mean the output is correct it is just statistically possible
  • The symbolic approach is what is used by Wolfram Alpha
    • It converts natural language into precise computation language and applies algorithm or built-in knowledge to solve it
  • The statistical approach is very poor at handling math
    • It can also appear correct because it can give a plausible explanation that sounds correct but is actually wrong
    • Or when not understanding the underlying facts can lead to wrong answers
  • Can one train a big enough neural network to do anything?
    • Fundamental facts about computation (computational irreducibility) makes it clear that this is not possible

Notes created today

List FROM "" WHERE file.cday = date("2023-01-10") SORT file.ctime asc

Notes last touched today

List FROM "" WHERE file.mday = date("2023-01-10") SORT file.mtime asc

created: 2023-01-10 11:30