3 ms·

The human brain is dramatically more energy-efficient than AI models like ChatGPT. Human brain: Uses about 20 watts of power. ChatGPT (GPT-4): Running a singl

The human brain is dramatically more energy-efficient than AI models like ChatGPT. Human brain: Uses about 20 watts of power. ChatGPT (GPT-4): Running a single query can use hundreds of watts when accounting for the entire datacenter infrastructure (some estimates suggest 500–1000 watts per query on average, depending on model size and setup). If we assume: 20 watts for the human brain thinking continuously, 1000 watts for ChatGPT processing one complex query, then the human brain is about 50x more energy-efficient (or 5000% more efficient) than ChatGPT per task, assuming equal cognitive complexity (which is debatable, but good for ballpark comparison).

JustFinishedBSG 1y ago
Watt isn’t a measure of energy. Without how long it takes for a human and ChatGPT to solve the task then the comparison doesn’t teach anything
- roschdal 1y ago
  You're absolutely right — watt is a unit of power, not energy. To make a meaningful comparison, we need to estimate how much energy (in joules) each system uses to solve the same task. Let’s define a representative task: answering a moderately complex question. 1. Human Brain Power use: ~20 watts (on average while awake) Time to think: ~10 seconds to answer a question Energy used: 20 watts × 10 seconds = 200 joules 20watts×10seconds=200joules 2. ChatGPT (GPT-4) Estimate per query: Based on research and datacenter estimates, GPT-4 may use: Around 2–3 kWh per 1000 queries, which is 7.2–10.8 megajoules Per query: 7.2 MJ 1000 = 7200 joules 1000 7.2MJ =7200joules per response (lower bound) 10.8 MJ 1000 = 10 , 800 joules 1000 10.8MJ =10,800joules per response (upper bound) Comparison Human: ~200 joules ChatGPT: ~7,200 to 10,800 joules Conclusion: The human brain is about 36–54 times more energy-efficient than ChatGPT at answering a single question. Or in percent: 3,600% to 5,400% more efficient
halyconWays 1y ago
There's massive evolutionary pressure for maximizing energy efficiency in brains. I'd like to see LLMs procreate and select for energy efficiency while, ideally, minimizing insanity and maintaining g-factor.
homebrewer 1y ago
You should then probably count the whole body, which consumes approximately 200 watts last time I checked.
ryanianian 1y ago
Watt is not a unit of energy but instead a unit of power. A brain may need 20 watts, but it may use 20 watts for a lot more time than ChatGPT would. The brain may ultimately be more power-efficient, but the units you want are watt-hours.
TobTobXX 1y ago
This is fascinating. I mean it's not an argument against LLMs (I have only one brain, even though I'd like to have more). But I really hope that we'll learn much more about how our brains work.
- Legend2440 1y ago
  The brain is more efficient because it physically is a neural network, as opposed to a software simulation of one.

JustFinishedBSG 1y ago

Watt isn’t a measure of energy. Without how long it takes for a human and ChatGPT to solve the task then the comparison doesn’t teach anything

roschdal 1y ago
You're absolutely right — watt is a unit of power, not energy. To make a meaningful comparison, we need to estimate how much energy (in joules) each system uses to solve the same task. Let’s define a representative task: answering a moderately complex question. 1. Human Brain Power use: ~20 watts (on average while awake) Time to think: ~10 seconds to answer a question Energy used: 20 watts × 10 seconds = 200 joules 20watts×10seconds=200joules 2. ChatGPT (GPT-4) Estimate per query: Based on research and datacenter estimates, GPT-4 may use: Around 2–3 kWh per 1000 queries, which is 7.2–10.8 megajoules Per query: 7.2 MJ 1000 = 7200 joules 1000 7.2MJ =7200joules per response (lower bound) 10.8 MJ 1000 = 10 , 800 joules 1000 10.8MJ =10,800joules per response (upper bound) Comparison Human: ~200 joules ChatGPT: ~7,200 to 10,800 joules Conclusion: The human brain is about 36–54 times more energy-efficient than ChatGPT at answering a single question. Or in percent: 3,600% to 5,400% more efficient

halyconWays 1y ago

There's massive evolutionary pressure for maximizing energy efficiency in brains. I'd like to see LLMs procreate and select for energy efficiency while, ideally, minimizing insanity and maintaining g-factor.

homebrewer 1y ago

You should then probably count the whole body, which consumes approximately 200 watts last time I checked.

ryanianian 1y ago

Watt is not a unit of energy but instead a unit of power. A brain may need 20 watts, but it may use 20 watts for a lot more time than ChatGPT would. The brain may ultimately be more power-efficient, but the units you want are watt-hours.

TobTobXX 1y ago

This is fascinating. I mean it's not an argument against LLMs (I have only one brain, even though I'd like to have more). But I really hope that we'll learn much more about how our brains work.

Legend2440 1y ago
The brain is more efficient because it physically is a neural network, as opposed to a software simulation of one.