Post AxqgZrUOMFXYsiCtEW by mason@partychickens.net
 (DIR) More posts by mason@partychickens.net
 (DIR) Post #AxqgZogKnOJKAslsvo by emaste@mastodon.social
       2025-09-04T00:33:50Z
       
       1 likes, 0 repeats
       
       Can't wait for September 3rd.
       
 (DIR) Post #AxqgZpoWaQSLgZbwKu by mason@partychickens.net
       2025-09-04T02:13:41Z
       
       0 likes, 1 repeats
       
       @emaste Now let's have it write some software for medical devices.
       
 (DIR) Post #AxqgZqm5105wfHTVAG by feld@friedcheese.us
       2025-09-04T02:18:02.114552Z
       
       0 likes, 0 repeats
       
       @mason @emaste the model Google uses for these summaries is not even close to being the same quality model you use for other tasks 🫠
       
 (DIR) Post #AxqgZrUOMFXYsiCtEW by mason@partychickens.net
       2025-09-04T02:24:24Z
       
       0 likes, 1 repeats
       
       @feld @emaste In fact, I don't use LLMs for any purpose. The environmental cost is too great.
       
 (DIR) Post #AxqgZsWuUN9I6oOPnU by feld@friedcheese.us
       2025-09-04T02:35:54.566949Z
       
       0 likes, 0 repeats
       
       @mason @emaste Google's Gemma model (the one used here IIRC):  > is reportedly only consuming "0.24 watt-hours (Wh) of energy, emits 0.03 grams of carbon dioxide equivalent (gCO2e), and consumes 0.26 milliliters (or about five drops) of water" per the average query.ChatGPT (unclear which model this was):> One hour of Netflix watching equals the energy required for between 300 and 3000 ChatGPT prompts. Ending your shower one second earlier saves enough water to compensate for 40 ChatGPT prompts.I hope to see more independent audits about this soon, but don't believe everything the overzealous tech journalists publish.AI won't break the planet. It will break corporations that over-invested in hardware they don't need. 🥳 What impact it has on society remains to be seen...
       
 (DIR) Post #AxqgZtQDAlNusKGZzk by SlicerDicer@friedcheese.us
       2025-09-04T02:38:43.297929Z
       
       0 likes, 0 repeats
       
       @feld @mason @emaste Google Gemma model is the fastest one I’ve run in TPS but it doesn’t follow instructions well.