Conversation

Notices

1. Embed this notice
   Paul Cantrell (inthehands@hachyderm.io)'s status on Monday, 19-May-2025 08:20:18 JST Paul Cantrell

   in reply to

   • Nick Lockwood
   • Matt Massicotte

   @nicklockwood @mattiem

   The trouble with that kind of heuristic is that every statement of the form “this time generic human-like intelligence is just around the corner!!” has also aged very poorly. For 60+ years running.

   I think it’s reasonable to guess that the past pattern will repeat: new tech will peel off some layer of the problem previously considered to be too difficult, we then discover that the next layer down is far harder than we realized, the tech becomes an assistive tool, and the human work changes but continues.

   • Embed this notice
     Nick Lockwood (nicklockwood@mastodon.social)'s status on Monday, 19-May-2025 08:20:19 JST Nick Lockwood

     in reply to

     • Matt Massicotte

     @mattiem the two statements are not contradictory - the former only implies that AI can't write an entire piece of software by itself, and the second only implies that it can be used as a tool to aid a human in the writing of software.

     That said, I am no longer all that confident that AI can't write software by itself. In the last few years almost every statement of the form "AI can do X but can't do Y" has aged very poorly.

   • Embed this notice
     Matt Massicotte (mattiem@mastodon.social)'s status on Monday, 19-May-2025 08:20:20 JST Matt Massicotte

     I regularly see both “AI cannot write software” and “I am using AI to write software successfully”.

     The emotional allure of the former is a big reason why the scientist in me refuses to discount the latter.

     (I do not use AI myself)