Because maybe you are thinking in storage LOTS of energy?

Think about it. A battery is amortized with usage. You have a range of batteries, but usually the best for BES is very stable batteries that can support 5.000-10.000 cycles.

So even if they battery + infrastructure costs you 300 $ per kwh, each cycle depends on
150/10.000 = 0,03 $/kwh cycled. Cheap enough.

But... how much time takes to do 10.000 cycles?

If you do a cycle per day, that's 10.000 days or around 27 years. Let's round to 30. Still reasonable.

But you only have one season change per year. Amortize on 10.000 years?. Hell, no.

What you need is a technology that decouples power and energy. In hydrogen the energy stored is managed by the storage/tank/underground reserve, while the power is managed by the fuel cell/electrolyzer.

So you can have a giant tank of one or two months of consumption, while you only have some megawatts of power. The storage can have gigawatts hour of energy inside.

If you want gigawatts hour of storage, you need gigawatts hour of batteries, to just use them one cycle per year.

Hydrogen is a lot cheaper for seasonal storage, while batteries are way better than hydrogen in daily cycling.

So both have sense for different scenarios.

You are wrong. Hydrogen is one of the most efficient chemical P2F solutions available, and that's the reason it has so much attention.

Of course is inefficient if you compare against NON chemical storage like batteries or air pressure. But in the chemical conversions, is one of the best. All are inefficient, but hydrogen is the best of them.

The problem is that the overall efficiency includes other complexities like storage AND very expensive infrastructure at current prices. And that's the reason it still fails. BUT there are potential use cases where hydrogen will excel.

And maybe season balance / energy reserve could work well. Not for daily storage, that's clear, but when you need to store lots of energy AND with reasonable efficiency, there is potential.

Being precise, using certain electrolyzers, underground salt caverns storage + high temperature fuel cells it will result in a reasonable efficient way to transfer some energy from summer to winter. And with scale, I don't see why the systems can't reach reasonable prices.

Pretty sure technology won't stop.

This is a market thing. Too many competitors and ruin each other in a too fast race.

Typical of a maturing market. Less capitalized business that don't have some exclusivity in the market burn all their case in the race and broke.

After sometime and all concentrate, the renewable will continue to be driven by real efficiencies more than a crazy race.

LLM are just a kind of AI that it's mainly pattern recognition. That's the reason depends in huge amounts of data.

But Models ChatGPT are already beyond LLM. They are already a "mix of experts". A model that route to other models that compute the problem.

The problem is that we are still working on good models for true abstraction.

On first sight, Hierarchical Reasoning Model sounds a very promising stepping stone in the right direction.

https://www.youtube.com/watch?...

I'm pretty sure is not as simple as the video shows, but it seems a really breakthrough beyond LLM models.

4 hours is fine enough to avoid the duck curve problem.

I would say, ask ChatGPT is similar to ask another person. One that knows a lot about everything but not in deep as well.

The problem is not that. It's the same with real people.

The problem is if young people uses this as a offload mechanism during their learning time. Learning requires effort. It's just like copying the work of another school pal. You don't do the effort, you almost learn nothing.

In the day by day basis, most people has no time to learn about other things, so they ask other people. In that circumstance, ChatGPT don't damage any skill, because the skill wasn't there in any case, and if you don't have ChatGPT, or other people to ask, you would simply ignore the question because you just lack the time to see about that.

But it's different if you are supposedly learning or working about something. You can use ChatGPT as a teacher, if you ask AFTER TRYING, then it will help you to get better, or you can ask the answer or do your work directly. In that case, you will lack effort and become weaker.

The problem, as always, it's not the tool but how it's used. Still the abuse and easy access make it a problem. You can't take the easy road if that road doesn't exists. So I can understand the problem.

Gas + wind is the fastest deployment to increase winter production.

As wind is not warranted, you need to deploy full gas power, BUT wind helps to lower the gas consumption by a lot. Solar helps a lot in the other seasons, sometimes making gas unnecessary.

So gas + wind + solar is the usual new power. Later batteries can lower the gas need even more.

Still, AI is growing incredible fast. Network managers will need to stop/delay some projects to ensure the network integrity is not compromised.

That's something I would like to have benchmarks. A new performance comparison between Linux/Proton vs Windows 10 instead of Windows 11.

I guess they will be more or less equal.

Well... You can say it's a "research" zone, but it has huge implications.

If the AI is mainly an advanced parrot, once the data become too small, or self-feeding, it has a huge problem of becoming wrong.

I have to say that these critics are too focused on LLMs while more and more AI is currently mixing different concepts while LLM remain more in the "talkative" layer more than in the "thinking" layer.

But it remains an acceptable concern that a lot of enterprises can be selling automated AI, and if that AI depends too much from previous human data, if AI removes more and more humans from the job market, the model can crumble.
Also we can having artificial expectations on the future expecting higher and higher results while the problem is that it's reaching that levels through copying so the higher it gets, the less data is available, so it will have difficulty to get even better results.

If it's like they said or not has a huge implications for the future of the AI.

That's said, AI is not only LLMs. And the advanced high pattern recognition is very close to abstraction, so LLMs can be smarter than they think.

Still, I agree in moderate caution and don't get blind by the results. We need to understand how much true intelligence has the models. Not just memorization, but understanding.

A rental and a hotel aren't obviously the same.

Hotel require to offer more services, while at the same time, a rental is a temporary lend of rights and obligations to a third party that doesn't exists in a hotel.

Different things.

That's said, because they share a lot of the same market, and because hotels have to offer more services, they are clearly more expensive, so house rentals are cheaper than hotels.
But at the same time, as in holidays people spend more, they are a lot more expensive that regular rentals, pushing regular rentals higher.

It's a bad consequence of a market behavior. Not law misuse required.

At the same time, there are also abuse of this system, making pseudo-hotel services under the disguise of rentals, skipping hotel regulations.

Also too much tourism generate bad externalizations, so I can understand why lots of people complains. So I see a reasonable regulation can help, like require a "license" which price depends on the pressure of the tourism (small number, pretty cheap or even free, large number, expensive), and making a clear distinction between hotel and rentals.

That said, I don't trust politicians to make a "reasonable" regulation. Probably just taxation and useless bureaucracy.

The thing is, having backup fuel generators are cheap, and if the operation is a minor fraction of time, even expensive e-fuels/biofuels are acceptable.

So if you want a 100% renewable grid, you also can do it. Through Power to Fuel technology for the last 10% fraction of consumption.

That's said, while fossil fuels are available, and lots of things require electrification, the money spend on electrification will reduce CO2/fossil fuel consumption faster than focus on the remaining 5-10% of the electricity production.

But even that fraction will be covered sometime. It's not that you can't. It's just is more cost effective to focus on other decarbonization efforts.

AlphaEvolve has developed new ways to do matrix calculations faster than the known algorithms.

For now, I know it will be plagued with problems. High costs, lots of bugs, etc.

But it's just matter of time it will work well.

Then what?

Then it depends in the next bottleneck. What limits our consumption?
There are different possible bottlenecks. One is wealth distribution. It doesn't matter if there are lots of people if robots owners accumulates most of the wealth so consumption of the most is limited or reduced, so the total consumption doesn't increase. That's turn into a dystopia.

Another is resources. We are in a transitional model from limited resources using in a inefficient way towards a new close circle renewable usage of resources. In the meantime, resources can limit the total production capability. In fact, massive automation can make this worse, so it's need to advance here as quickly as possible.

An third is just human desires. A different way of living and culture can just limit our own consumption.

If the three occurs. Wealth redistribution, sustainable resource usage and a rejection of consumerism culture, then the consumption per person in Earth can peak. And if population peaks, also peaks the total production. And as AI and robots advances, our society will need to change as the old job centered model won't work anymore. We won't produce to consume. Robots can do that. We will only work for enjoyment and self-fulfillment, because it's healthful for us (in a moderate level).

There are multiple dystopian futures, but also utopian futures as well, where our moral values changes and humankind can refocus to new goals and ways of living.

I don't expect a soft transition, but still I'm optimistic with the final outcome after some social tensions. More like the previous century. (Let's hope it doesn't include World Wars)

AI is coming and they can't stop it, only slow it.

The copyright system has benefits but also problems.

The idea of copyright is to help creators, inventors, etc. so bring up the idea, which is one of the most difficult steps is correctly rewarded instead of allowing anyone to just copy the good idea and doing a better implementation sometimes not through a refinement but just better manufacture capabilities, creating a barrier between creators and producers.

BUT, that model runs over premises that are crumbling each day it passes.

First, original ideas used to be owned by small inventors or creators. Nowadays most of them are owned by big companies, just because the real creators are employees or because they buy the patent.
Patents & copyrights are more and more using as ways to block innovation instead of spreading it.
Sometimes when the rewards reach the creators (let's say, for example, a group of researchers of a drug inside a big pharma), most of the revenue is dragged by the pharma instead of the researchers.

And now, with AI, while we aren't still in the moment where AI are good as creating new styles or ideas (although there are some success), they clearly allows to DROP the replication costshttps://slashdot.org/story/25/03/27/0023207/openais-viral-studio-ghibli-moment-highlights-ai-copyright-concerns?utm_source=rss1.0mainlinkanon&utm_medium=feed# to a ridiculous degree.

If the old model worked well, rights should drop the cost by creation, while the total revenue could remain flat or even increase thankful to the increase of total works created. BUT that's not how they try to move here.

The people with AI wants free access. Creators wants restrictions.

We clearly a need a new model. One that reward each person, BUT NOT IN EXCESS, and don't try to stop the unstoppable. That only will move the generators from one place to another.
Once the AI tech works enough well, the impact on creation will be enormous, reducing producing costs by a lot.
Now we have images and rudimentary tools. But in the future we will have plenty of tools to make fix a style, build characters, develop scenes through a combination of sketches and tools to tell the AI how the animation should work, not just words..

In the end, a very small of workers will be able to do plenty of work.

That's a huge advantage. We shouldn't stop that revolution. We just need a model that reward each worker rightfully. Style creators too.

The old model require a huge update or just a replacement by a new model.

Maybe it has reached a peak, maybe not.

It's clear that if the population peaks, and without good AI, the total brain will peak one way or another.

It's not the population peak, not yet. But adding people is just one of the factors, so other things can accelerate the peak a little.

BUT, that doesn't mean that's a stop for humankind. So that peak would be forever.
To grow exponentially we clearly needs grow, but we need to create more "space" to grow without collapse our environment.

That's exactly what a space program can do... in a long... very long term. So, for some decades, maybe centuries, our civilization can reach a brain peak. Well. AI progress seems optimistic so we could expand through that for some time. In long term, adding more population will be a must, and that require to build more sustainable places. And probably Earth is not the best place if we want to ensure that our homeworld biosphere doesn't collapse.

Automation and AI can helps us to break that limit in the meantime until a lot of space infrastructure is built. Build computer centers on space would be easier than manned colonies, but in the end, it's almost the same. It require we industrialized space anyway. And that would take time, so that temporary peak is very probable, no matter if it's in the recent past or in the near future.

But we can break that limit. It's not forever if we don't loose the will.