'Thirsty' ChatGPT uses four times more water than previously thought

Re:'Thirsty' ChatGPT uses four times more water than previou... by lambdaba on Saturday November 23, 2024 @04:38AM (#42220065)

Inference consumes far more total compute than training the model.

[ Parent | on HN ]

Re:'Thirsty' ChatGPT uses four times more water than previou... by MrMcCall on Saturday November 23, 2024 @04:45AM (#42220096)

By 'inference', do you mean using the model to compute a specific series of results?

[ Parent | on HN ]

Re:'Thirsty' ChatGPT uses four times more water than previou... by lambdaba on Saturday November 23, 2024 @04:50AM (#42220119)

Yes, training is a tiny amount of usage, this is well-known.

[ Parent | on HN ]

Re:'Thirsty' ChatGPT uses four times more water than previou... by MrMcCall on Saturday November 23, 2024 @04:52AM (#42220132)

Well, if that's the case, then the situation is worse than I imagined. I thought that training the model would be far more expensive computationally than using it to compute an 'answer'.

Thanks.

[ Parent | on HN ]

Re:'Thirsty' ChatGPT uses four times more water than previou... by MF-DOOM on Saturday November 23, 2024 @05:15AM (#42220193)

Training is much more computationally intensive than computing a _single_ answer. But not compared to running the model over time at scale for millions of users.

[ Parent | on HN ]

Re:'Thirsty' ChatGPT uses four times more water than previou... by MrMcCall on Saturday November 23, 2024 @05:19AM (#42220206)

Thanks. That is precisely how I intuited the situation. 40+ years of programming FTW!

[ Parent | on HN ]