Altman fluffs superintelligence to save humanity as OpenAI slashes prices

>GPT-4o query consumes around 0.3 watt-hours for a typical text-based question, though this increases substantially to 2.5 to 40 watt-hours for queries with very long inputs.

Given a context window of 128 kilotokens, and how the energy usage range (0.3 to 40 Wh) increases by a factor of 133 from "single query" to "worst case," this suggests linear scaling of energy usage with context length.

Given the common user pattern of "chat with the LLM" the 0.3 Wh figure seems reductive. As the chat grows in length, the length of the "query" becomes the length of the entire chat.

Assuming context compression for long chats, the figure we're looking for per-question energy use should be the energy use for a query half the length of max context tokens.

For GPT-4o: 20 Wh

I absolutely believe inference is cheap and efficient, but the net impact of training the model should be also added to the equation. Not just the cost of inference.

Is this information readily available somewhere?

Business man lies to sell product.

I am having a hard time figuring out when Elon and when Sam said something nowadays

> "Maybe we will go from solving high-energy physics one year to beginning space colonization the next year; or from a major materials science breakthrough one year to true high-bandwidth brain-computer interfaces the next year," he said. Maybe. Or maybe not. We note that fellow futurist Elon Musk, who once cautioned about releasing the AI demon, predicted in 2016 that humans would land on Mars by 2025.

"Maybe"

What about the definition of "AGI" being changed again to mean something else? What does "AGI" actually mean anymore or was it hijacked again for the purpose of manipulation?

But this is the most important sentence of the entire article:

> Tech leaders simply pay no cost for misprediction.

They don't care if they are wrong on their predictions. So focus on what they are doing rather than what they are saying.

I'm sorry, but I'm not buying into that level of optimization. Earlier this year they struggled to make a profit on Pro at $200, granted that was for unlimited usage, but now they've optimized the design/implementation sufficiently to slash prices by 80%?

I’m interested in how good it is at coding, only a little interested in other stuff.