4 ms·
The less I know about a thing the more useful an LLM seems to be. I’m working with a new-to-me enterprise code base, the LLM helps me find related (and duplicat
by cmiles74 1y ago
The less I know about a thing the more useful an LLM seems to be. I’m working with a new-to-me enterprise code base, the LLM helps me find related (and duplicate) code. Even here it’s usefulness has an expiration date, eventually I’ll know where stuff lives and I’ll use it less and less. Life experience tells me I’m not unique and I suspect the constant cram-AI-into-the-thing is because the vendors are hoping, eventually, they’ll find a use-case for LLMs that sticks.
- giancarlostoro 1y agoThis is the correct way. Use the LLM dont let it become the only way you work.
- immibis 1y agoYes, they've legitimately good at a few things (e.g. very fuzzy search), but that doesn't justify the amount of investment.
- goku12 1y agoThe environmental damage is even worse than the investment. Those who have the money to invest in it usually care only about the returns and not the environment.
- tempodox 1y agoHardly anyone even mentions it in discussions. Even as far as these tools are actually useful, nobody ever asks whether it’s worth the environmental costs.
- cshores 1y agoFor Google's Gemini LLM, the energy impact is negligible, with the average prompt consuming the equivalent energy of just three seconds of a microwave's operation.
- immibis 1y agoHow much energy is otherwise consumed by a Google search?
- cshores 1y agoI’m not entirely sure, since it seems that a very slimmed-down version of Gemini has been attached to search. It’s definitely not the full Gemini 2.5-Pro that engineers use to carefully reason through answers. Instead, it relies mostly on tool calling to stitch together a response.
- tempodox 1y agoAll those data centers full of GPUs aren’t running on solar or wind power.
- cshores 1y agoI did a bit of research on the environmental impact with regards to the United States. Recent numbers suggest that ChatGPT handles about 2.5 billion prompts per day worldwide, with roughly 330 million of those coming from the United States. Since the U.S. population is about 335 million, that works out to about one prompt per person per day on average, though actual users issue several times more. On the energy side, Google recently estimated that an average Gemini inference consumes around 0.24 Wh, which is roughly the same as running a microwave for a single second. Older rule-of-thumb comparisons put the figure closer to 3–6 seconds of microwave use, or about 0.8–1.7 Wh per prompt. If you apply those numbers to U.S. usage, you get somewhere between 79 MWh and 550 MWh per day nationally, which translates to only a few to a few dozen megawatts of continuous load. Spread across the population, that works out to between 0.09 and 0.6 kWh per person per year — just pennies worth of electricity, comparable to a few minutes of running a clothes dryer. The bigger concern for the grid is not individual prompts but the growth of AI data centers and the energy cost of training ever-larger models.