Conversation

Notices

1. Embed this notice
   Stephen Sekula (steve@mastodon.cooleysekula.net)'s status on Tuesday, 01-Jul-2025 03:15:11 JST Stephen Sekula

   in reply to

   • Aram Sinnreich

   @aram Joking aside, I find it profound that such basic deep neural nets naturally exhibit hallucination. Since only about 25% of humans will hallucinate once in their lifetime, but artificial neural nets of the modern variety hallucinate strongly ... it makes me wonder what biological neurological mechanisms must be in place to inhibit or offset the active hallucination response.

   I feel there is much to learn even from primitive tech. But I agree we should not over-hype it.

   • clacke@libranet.de is my main likes this.
   • Embed this notice
     Stephen Sekula (steve@mastodon.cooleysekula.net)'s status on Tuesday, 01-Jul-2025 03:15:12 JST Stephen Sekula

     in reply to

     • Aram Sinnreich

     @aram It still boggles my mind that people hype the fundamental excellence of modern machine learning when, in fact, the technology that goes into it is still largely based on the experiments of Hubel and Wiesel which were focused on the visual cortex of the common housecat.

     Personally, while it's breaktaking how effective these technologies are ... there is a part of me that still goes, "But this is based on a housecat visual cortex (glances uncomfortably at his 2 treacherous cats)."

   • Embed this notice
     Aram Sinnreich (aram@aoir.social)'s status on Tuesday, 01-Jul-2025 03:15:14 JST Aram Sinnreich

     It’s exhausting to read endless credulous coverage of “AI” when we don’t really understand the human brain, or human intelligence, well enough to treat many neurological and psychological conditions, let alone “match” it with circuitry.

     For instance, I have a loved one with Alzheimers. Will Bezos’ giant data center get Alzheimers? No? Then it won’t “match” the human brain. GTFO with this nonsense.

     https://futurism.com/the-byte/amazon-anthropic-ai-data-center