| Most of OpenAI CEO rhetoric doesn't hold up to even cursory reason. "People talk about how much energy it takes to train an AI model - but it also takes a lot of energy to train a human. It takes about 20 years of life - and all the food you consume during that time - before you become smart." (Sam Altman) Let us take a worst case upper favorable estimate of his flippant statement: Adult Human peak power W: 600 Average U.S. home yearly kWh: 10500 20 Years = 7300 Days = 175200 Hours Human_metabolic_kWh = 105120 Home_shelter_kWh = 210000 Total_20Y_Human_Life_kWh = 315120 US average electricity cost is @17.65¢/kWh, and according to sources the LLM models took “1.2 million kWh for older foundational models to upwards of 11 billion kWh for next-generation frontiers". So if true that means: 1200000 kWh = $211,800 USD 11000000000 kWh = $1,941,500,000 USD In other metrics that means we know: 1200000 kWh trained LLM = 4 Human lives reallocated 11000000000 kWh trained LLM = 34907 Human lives reallocated Sam claiming 34907 Human lives is a small price to pay for a 85% accurate nonsense machine is ridiculous hyperbolic nonsense, and or the data they are publishing is inaccurate. People have a right to be displeased with the industry dismissing peoples concerns. =3 |