Re:OpenAI Livestream: ChatGPT Images 2.0
by runarberg on Tuesday April 21, 2026 @07:46PM
(#47856463)
Also, ignoring training when talking about the environmental costs is bad faith. Without training this image would not exist, and if nobody generating images like these, the training would not happen. So we should really ask, the 10 seconds it took for inference, plus the weeks or months of high intensity compute it took to train the model.
[ Parent
| on HN ]
Re:OpenAI Livestream: ChatGPT Images 2.0
by ToValueFunfetti on Tuesday April 21, 2026 @09:24PM
(#47857502)
You'd want to compare against the fraction of training attributable to the image
Re:OpenAI Livestream: ChatGPT Images 2.0
by runarberg on Wednesday April 22, 2026 @12:09AM
(#47858884)
That is not how environmental damage works. I don‘t count my BP’s oil production towards the fraction of the deep water horizon disaster. I don‘t count a suburban home in Santa Monica towards a fraction of the 2025 California wildfires. The claim is normally that BP’s business practices resulted in the the Deep water horizon disaster, and that California suburban development resulted in the 2025 wild fires spreading out of control.
Here the AI industry is responsible for the environmental damage of AI training. And when you enumerate the damage caused by AI, you include the damage caused by AI training, just as you would include Deep water horizon for the damage caused by oil, and you would include the LA fires as the damage caused by suburban sprawl.