AI is evolving towards human-level intelligence at breakneck pace – or have we already reached it? ChatGPT is very powerful in its own right though it may not reach full human level intelligence or AGI levels. Still, if that's what's available to the public, is it possible that powerful government agencies or tech giants have created full AGI in secret?

Several people have asked me this question so this video dives into some arguments for why AGI may or may not have been created in secret. Some of the discussion is based on my experience as a PhD student at an Ivy League institution.

This video also defines various tests for AGI and artificial intelligence. This helps answer the question, is there a Turing test for AGI? We also discussed briefly what might be left before AGI is created in the public sphere.

#agi #ai #secret


What's Left Before AGI? PaLM-E, 'GPT 4' and Multi-Modality
   • Video  

Has GPT-4 Ignited the Fuse of Artificial General Intelligence?
https://www.datanami.com/2023/03/30/h...

Morgan Stanley note on GPT-4/5 training demands, inference savings, Nvidia revenue, and LLM economics
  / morgan_stanley_note_on_gpt45_training_demands  

AGI: What, When and Why
https://blogs.perficient.com/2023/04/...

0:00 Intro
0:20 Contents
0:27 Part 1: What is AGI anyway?
0:47 Turing test
1:32 Other proposed tests
2:10 Does AI pass these tests today?
2:56 Applying tests to GPT-4
3:49 Part 2: What's left before AGI?
4:28 Scaling hypothesis
5:19 Exponential growth in intelligence
5:56 AGI continuum
6:14 Do we already have AGI?
6:55 Part 3: Has AGI already been created in secret?
7:19 Government agencies and contracts
8:55 Academic research
10:19 Industry labs
11:32 Summary: government, academia, industry
11:52 Nvidia GPU argument
12:57 Conclusion
13:34 Outro