Post AvippVhttapgk9QXtg by BubblegumYeti@mastodon.social
(DIR) More posts by BubblegumYeti@mastodon.social
(DIR) Post #AvippULAyb3qVNmsq0 by kibcol1049@mstdn.social
2025-07-01T13:52:45Z
0 likes, 0 repeats
As AI increasingly powers critical decisions in finance, infrastructure, and beyond, understanding how these systems arrive at their conclusions becomes crucial. While AI’s “black box” decision-making may be acceptable for simple tasks like email generation, it raises serious concerns for high-stakes applications such as loan approvals or power grid management.
(DIR) Post #AvippVhttapgk9QXtg by BubblegumYeti@mastodon.social
2025-07-01T13:56:07Z
0 likes, 0 repeats
@kibcol1049 does it really?
(DIR) Post #AvippWvPMrEQWKkqae by kibcol1049@mstdn.social
2025-07-01T15:11:28Z
0 likes, 0 repeats
@BubblegumYeti We'll have to see. Here's hoping there are enough safety features built in.
(DIR) Post #AvippXhySI50wxTdI0 by lproven@social.vivaldi.net
2025-07-01T15:43:35Z
0 likes, 1 repeats
@kibcol1049 @BubblegumYeti There aren't. There are none. I am dead serious. What is currently called "AI" is neither the _A_ nor the _I_. It's not intelligent -- it can't even count -- and it's not artificial. "Artificial" means "built by artifice". It's from the same root as "artificer", a skilled craftsman. "Generative AI" is just very very clever automatic predictive text _and nothing more_ built from giant tables of closeness of words in databases with _millions_ of dimensions. They are called "large language models" and they are all built from a simple algorithm called "the transformer". The @financialtimes has a really good, readable, simple explainer here:https://ig.ft.com/generative-ai/It's free. Well worth your time.It's clever but it's no miracle. No humans can directly read LLMs or ever will. All they can do is feed them back into themselves: get the bot to analyse its own output and second-guess itself.(They are already ecologically catastrophic to start with, but this process makes them much much worse.)