Hacker News Reader

Comment by swyx | Hacker News Reader

(i'm the author) it was 100% human, i'm afraid to say. i was trying to summarize a lot of content in a concise article.

An erroneous summarization is:

  The Tiny Teams concept has resonated so strongly that it’s 
  pretty clear it is the next major transition of the org 
  chart as we go from level 2 to 3 AGI.

LLM's, nor any offerings based upon them, qualify as Artificial General Intelligence[0]. So to assert there is an existing "level 2" AGI, let alone a progression to "level 3" AGI, is nonsensical.

0 - https://en.wikipedia.org/wiki/Artificial_general_intelligenc...

Quoting wikipedia:

>Researchers generally hold that a system is required to do all of the following to be regarded as an AGI:[34]

> - reason, use strategy, solve puzzles, and make judgments under uncertainty > - represent knowledge, including common sense knowledge > - plan > - learn > - communicate in natural language > - if necessary, integrate these skills in completion of any given goal

Modern AI can do all of those.

Can it though? Everything I've seen and experienced is that LLMs are very good at making it appear to do those things, but the amount of times I've gotten stuck on "you're absolutely right!" When correcting the LLMs suggests that it can not reason by any means, nor does it learn. Otherwise, an LLM would never get stuck in a loop.

Of the reproduced attributes listed (reformatted and numbered):

  1 - reason, use strategy, solve puzzles, and make judgments under uncertainty 
  2 - represent knowledge, including common sense knowledge 
  3 - plan 
  4 - learn 
  5 - communicate in natural language 
  6 - if necessary, integrate these skills in completion of any given goal

Only 2 (partially) and 5 are applicable to LLM's, as they are statistical text (token) generator algorithms. Very useful ones, no doubt, but not ones which satisfy all of the above.

Reply on news.ycombinator.com