Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Jevons paradox

en.wikipedia.org

Post reply on HN
Print view View original

151–160 of 165 posts

Re: Jevons paradox

#151

Post by sangnoir »

ahmeneeroe-v2 wrote: post #139
aurareturn wrote: post #137

Earlier quoted context omitted.

This felt false to me, LED bulbs are generally 10x more efficient, so the average number of lights would have had to increase 10 fold, and this is ignoring the savings from not having AC cool those other 50 watts a bulb that went to heat. Jevons Paradox does not state that there is a 1:1 relationship between efficiency increase and consumption. A 2x increase in efficiency does not always result in a 2x increase in co…

Yes, I don't have 10x more lightbulbs but now household appliance has an LED. I just installed new GFCI outlets and all of them have LEDs too!

The comparison becomes meaningless when we count the milliwatt, sub-Lumen LED in an outlet wirh incandescents. "I had 8 incandescent lightbulbs, but now have 1000s of LEDs- including outlets, TV backlights, DVR status" isn't comparing like with like.

Re: Jevons paradox

#152

Post by ahmeneeroe-v2 »

sangnoir wrote: post #151
ahmeneeroe-v2 wrote: post #139

Earlier quoted context omitted.

Yes, I don't have 10x more lightbulbs but now household appliance has an LED. I just installed new GFCI outlets and all of them have LEDs too!

The comparison becomes meaningless when we count the milliwatt, sub-Lumen LED in an outlet wirh incandescents. "I had 8 incandescent lightbulbs, but now have 1000s of LEDs- including outlets, TV backlights, DVR status" isn't comparing like with like.

Yeah probably a meaningless comparison for Jevons Paradox (and its application to NVDA) since that is about resource use and LEDs are not a resource

Re: Jevons paradox

#153

Post by aurareturn »

empath75 wrote: post #123
crazygringo wrote: post #113

Earlier quoted context omitted.

No, this isn't directly about Nvidia or stock prices at all -- it's about the emergence of much lower-cost AI models. So the way Jevon's paradox could apply is that if an AI model only costs 10% of the energy+GPU to train (or run), then we might not get 90% energy+GPU savings... we might actually wind up consuming more energy+GPU on the whole because the models become so much cheaper, we use them more than 10x as muc…

The way the market reacted is completely illogical. If Nvidia announced a new chip that was 10x more effective at the same price, the market would be like WOW they're going to sell so many of those! But a software advance that achieves the same effect somehow causes the opposite reaction.

Exactly. The only thing that matters is AI adoption rate and training scaling laws right now.

Re: Jevons paradox

#154

Post by mrbungie »

scotty79 wrote: post #81
kgwgk wrote: post #49

Earlier quoted context omitted.

Is that really the analogy you want to use? « In 1873, greed, speculation and overinvestment in railroads sparked a financial crisis that sank the U.S. into more than five years of misery » https://www.smithsonianmag.com/history/robber-baron-gamble-r...

And yet, in history, railroad is one of the few examples of bubbles that didn't ultimately popped. You can still lose money in the greed-fear cycle even if the road leads up to unimaginable heights.

A bubble around a technology can pop, and yet the tech can be ultimately valuable in the long term. That does not mean it wasn't a bubble nor it didn't pop.

Re: Jevons paradox

#156

Post by scotty79 »

kgwgk wrote: post #93
scotty79 wrote: post #81

Earlier quoted context omitted.

And yet, in history, railroad is one of the few examples of bubbles that didn't ultimately popped. You can still lose money in the greed-fear cycle even if the road leads up to unimaginable heights.

I guess if it « didn't ultimately popped » neither did the dotcom bubble - nor any housing bubble.

Yes. Both dotcoms and housing still bring immense profits to investors.

Re: Jevons paradox

#157

Post by HarHarVeryFunny »

aurareturn wrote: post #118
HarHarVeryFunny wrote: post #105

Earlier quoted context omitted.

The models will all get more efficient, but NVIDIA as the high cost provider of compute (as well as cloud services that use them) - 10x markup on H100s - may not benefit from increased demand. If the models turn out to be a commodity, then low cost providers will be the winners, likely running on chips like TPU and Trainium (Amazon/Anthropic) that don't have the 10x markup.

Low cost providers will not be the winner because those will become commodity. It’s the leading model, the one closest to AGI that will make all the profit. Nvidia has products to serve lower end as well. There’s no reason why Nvidia can’t dominate in the commodity sector. And if Nvidia can’t keep their high margins, they will make it up with significantly more volume (Jevons Paradox in play).

I think there'll continue to be models at different levels of capability and price - and not at all obvious whether it'll be the Ford or Rolls Royce that'll make most money.

My guess is that is that what passes as "AGI" will also become commoditized, maybe even free to consumers (funded by advertising, business referrals to restaurants, etc).

Re: Jevons paradox

#158

Post by HarHarVeryFunny »

blagie wrote: post #117
HarHarVeryFunny wrote: post #115

Earlier quoted context omitted.

> I would not be surprised if there were a 10x increase in use of lighting with LEDs. Has the average homeowner added more ceiling lights or lamps because of different bulbs? Has the average office added more lighting because of LEDs? Factories? In order for an overall 10x increase in usage, there needs to have been some MASSIVE new usage or expansion of lighting, somewhere, to make up for all of these enormous use c…

> Has the average homeowner added more ceiling lights or lamps because of different bulbs? As much as I appreciate the snark and sarcasm, at least where I live, the equivalent is happening. Old home and commercial construction would typically have one central light per room. Homes would have a single light fixture with 1-3 lightbulbs at the center of each ceiling and were pretty dimly lit. Commercial would be similar…

I can accept that new construction may have more lighting, but nowhere near 10x, and certainly not 10x overall when you factor "legacy locations" where it is mostly just a bulb replacement.

Re: Jevons paradox

#159

Post by inglor_cz »

aurareturn wrote: post #134
inglor_cz wrote: post #129

Earlier quoted context omitted.

Maybe installed capacity in lumens.

I can’t find the numbers you’re looking for. I did find that the total consumption of light increased 100,000x between 1700s and 2000 in Uk. Chart: The consumption of light in the United Kingdom increased over 100,000 fold between 1700 and 2000. https://onclimatechangepolicydotorg.wordpress.com/low-carbon... Also keep in mind that better lighting efficiency enables people in poor countries to have lights where they c…

Absolutely, yes. Many more people can now afford light.

My grandma (born 1926, died 2016) was born in Eastern Slovakia in the pre-electric era, the region she lived in only got connected to the grid in 1945.

In her youth, the whole family would simply have one kerosene lamp and gather around it after dark.

Re: Jevons paradox

#160

Post by mitthrowaway2 »

woadwarrior01 wrote: post #110
ApolloFortyNine wrote: post #104

>Energy-efficient lighting, such as LED bulbs, has become a popular solution to reduce energy consumption. However, it serves as a notable example of the Jevons Paradox. While LEDs significantly lower the cost of lighting per unit, this reduction often leads to increased usage. For instance, people may install more lights in their homes or businesses, use them for longer durations, or adopt decorative and outdoor lig…

This is exactly the case. Anecdotally speaking, growing up in the 90s, I remember my parents adding lightbulbs and fluorescent tube lights to their shopping list every couple of months. We used to have a small stockpile of these in the pantry. Fast forward to today, the last time I bought lightbulbs for my house was over 5 years ago and that was to upgrade from regular led lamps to Philips hue.

The discussion was about energy efficiency and energy consumption, not light bulb durability and replacement rate.
Post reply on HN