They do it because it makes customers stickier, I think.
All telcos offer huge discounts to sign up, but those discounts aren't sustainably profitable. So prices usually rise afterwards... And then customers move.
They are betting that if you sign up to cable + mobile, you will be less likely to switch for the competing offer, and pay the higher long term price.
He's well compensated because he owns a lot of the company, much at much cheaper valuations, right?
For all the complaints about him you can't argue he hasn't delivered a lot of value. Lots of corp CEOs make more with falling company performance and valuation, because they aren't really paid for performance.
The conclusion I've come to is that AI is still revolutionary, and the buildout is probably necessary to some degree, but it is way overestimating the demand and pricing power providers will have.
SW can use a ton of AI. Lots of other applications can't, and they don't need the super smart frontier versions either. We just don't need all that much compute to serve AI for these applications, and people won't need to pay the margins necessary to get access to that AI. SW alone is not going to be enough to pay for all the investment.
I don't buy brand loyalty either. If I can pay 1$/m for the same thing from someone as I pay OpenAI or Anthropic 10$/m for, I'll do that, which almost zero barrier to entry.
The buildout is paying stratospheric prices on the assumption they can charge that much on the other end, but if demand tapers off (or model efficiency doubles again), prices will start falling. Everyone talks about Jevon's, but I don't really see why I'd spend more on AI than I need to do the things I want. I already run out of stuff to ask it to do on a basic subscription, I hardly need more.
The crux is this: I can spend x for y results. If I spend more, I probably don't get 10x for 10y results - I probably get, say, 2y results. So on that equilibrium I just don't spend all that much more because the reward for all that extra spend is not worth it. And if models get better this tradeoff starts to look worse, since I get more of the super-effective benefits for cheaper. Now I'm spending 0.1x for y results, and spending even x for 2y might not make sense, let alone 10x for 3y.
The problem is, that its not just AI for a niche thing, its AI together with compute.
People are already using AI to build them some things with CAD. Electronics too. But now what does AI need to be very good in it without an expert? Simulation and Optimizers. This again needs more compute.
And the agentic layer will be added to all big products out there. Everything you will do will consume AI tokens
To be fair, Jevons suggests that as prices fall, novel use-cases are unlocked. The things you or others currently do with AI might not need frontier models or loads of tokens, but if prices fall or capabilities increase, new applications might arise to exploit that.
Yes, this reasoning has been used to justify the absolutely gigantic amount of spending going into AI right now.
I'm not convinced it's true. It just seems to be taken for granted that demand for intelligence is unlimited and that its cost limited, which just doesn't seem true to me.
I don't understand how anyone can call these AGI. I've always had the view that AGI == Independent human-level intelligence.
At some point someone moved the goalposts for this to include LLMs that emphatically are not that. I think it was Altman when they came out with Astra?
Please do a youtube series where you ask random people on the street some questions and have them go against Opus 5.5. The model is definitely smarter than me in many ways- it would obliterate me in trivia, could write a poem in iambic pentameter before I could remember or look up what exactly the rules for it are, would probably dominate me at chess or football strategy, would do a better job at taxes, and for sure writes better code outside my small little slice of specialty. It definitely parses and diagnoses logs better than me unless I wrote all the loggging statements myself.
I have to ask, what humans do you know that have more general intelligence? This isn't a dream team where you can say "Magnus Carlson can beat it at chess, and Linus can write better C code", this is a comparison across all domains and levels of intelligence. Even its IQ is estimated at 141, which is seriously higher than the average human. The average human's reading level is atrocious. The average human cannot write a standard 5 paragraph essay about a random topic.
If I choose an arbitrary video game, board game or whatever can Opus play it?
My understanding is that models will often still forget games rules etc unless you design a harness specifically to assist them and prevent them going off the rails. I was going to use chess as an example but it sounds like they are good enough to do that now, albeit I'm not sure if they have a special harness for it.
It's not about how "smart" they are, or how good they are at recall. It's about whether I can teach a model to understand and execute an arbitrary, general task regardless of whether they have been trained on that task before.
Disregarding the above, the fact models still require a human prompter/decision-maker is defacto proof they aren't AGI in my view. If they were humans would be obsolete in decision-making.
I guess im doing a terrible job of explaining my view.
If these were AGI we wouldn't be talking about them helping engineers, engineer jobs would just not exist. The AI would be superior in every way that mattered. That's not the case yet, so IMO we don't have AGI.
they aren't as intelligent as a human, but they can throw shit at the wall faster, and get something that sticks faster. they are superhuman in areas that require trial and error, like hacking.
everyone stopped talking about the Turing test once ChatGPT came out and broke right through it while still being very very dumb. AGI is just too blurry to define.
The ~grid scale design of this for reference is somewhere around 200mx200mx50m for above ground, or ~100m cube (or bigger) for quarry infill. Such a design is going to get crushed by battery/hydro over the hours/days, but wins on cost for seasonal usage.
The biggest expense will be the pipework inside that pile. That you won't be able to service easily. And you'll also need a LOT of insulation around it to prevent heat from leaking out too quickly.
Even better than that: the thermal time constant is proportional to the square of the radius. There's the square/cube effect that causes a factor of r ratio in the volume to surface area, but there's also a factor of r from the temperature gradient (which, from the center of the pile to the surface, is inversely proportional to r).
Homework: compute the heat flux from a 40 meter radius sphere, given the surface temperature 600C and the ambient temperature 0C. Now compute the amount of insulation needed to keep its temperature above 500C for 3 months, given that dry sand has an R value of 0.25 W/mK.
Note that the thermal time constant of a given geometry scales as the square of the radius. This is why subsurface geothermal heat is there even though it was "charged" gradually over millions of years. So it's not hard to achieve very long time constants just by making the pile a bit larger. They've done these computations. The result is the minimum practical size is a pile that provides ~100 kW(th) constant output.
So the idea is not practical for storage for an individual home, but could work on the scale of a village or small industrial enterprise. It would allow a solar microgrid to be completely disconnected from the larger grid, even at high latitude (assuming one could get a sufficiently good heat engine to work at that scale; it's a bit small for steam turbines.)
If you look at very large piles of dirt, like large earthen dams, if arranged in a suitable geometry they would have thermal time constants (assuming no fluid flow through them) in the thousands of years.
A computation for you: compute the cost of a given R value per m^2 for various materials. Dirt is very cheap, so even if it's not the best insulator it comes out on top. You need the thing being insulated to be large enough that you can fit the thick dirt layer around it, of course.
They also state the cost per kWh(th) of storage capacity is about $0.10, or $100,000/GWh(th). This is thermal energy, which is converted to work at maybe 40% efficiency, mind you. There are costs over and above the dirt (for example, pipes laid through the dirt to carry heat in/out).
It's not quite as bad as that, since you don't consume fossil fuel resources when plants aren't running, but this is mostly a benefit for oil/gas plants, not coal (since the coal is likely useless without the power plant consumer anyway)
If you shut down by day, it can take multiple hours for it to be able to provide full load... and multiple hours for it to shut down cleanly. During this time it is producing power at less than optimal rates and/or using the fossil fuels to keep the wheels spinning and in sync with the grid so that they don't have to go through a phase matching period.
(aside for coal - note that it takes more than 12 hours for a coal plant to get running from a cold start)
That's not really a problem. A spin up over a couple hours matches nicely with the slow sunset over a continent. The sun takes ~4 hours to go from a bit above the horizon in the east coast to a bit below the horizon on the west coast (solar panels make some power in twilight).
Wind also helps since wind makes more power at night, as does demand which is dramatically lower at night.
For coal, Great. Making coal economical is excellent. We should have banned it worldwide 20 years ago.
Solar+A couple of hours of batteries is already economically viable. The storage problem that is still expensive is bridging days/weeks of insufficient sun and wind.
The US uses around 4TWh of electricity a night. A battery with that capacity, to get through the night with no generation, would cost over a trillion dollars. Of course thered still be some generation at night, but still, using a battery just to get through the night is not even close to viable
That sounds like a lot right? But it isn't large compared to US wars to protect energy resources (>$4T just for post-9/11 wars), and the payback it would generate over decades to come. In fact, this level of investment is a total no-brainer, compared to the benefits it will generate (forget about even the climate benefits - just the other economic and defense benefits.)
You generally get more return on investment if you invest in building infrastructure in your own country rather than investing in blowing up infrastructure in a different country. So I’m implying that solar and batteries are a lot cheaper than the war in Iraq.
1 week national battery storage is totally doable. All we need to do is convince the rest of the planet to stop making GPUs, phones, computers, and EVs for the next decade, and then give us all their rocks.
78.4TWh is about $19T (all in costs) for a national battery grid for 1 week storage, could be built over ten years but would need the US to build and own double the current global output, not even considering where all the raw materials and labor comes from.
And if we subtract all those datacentres crammed with Nvidia GPUs ? ;)
If we're being honest here, the problem with the US is that it has always treated any sort of energy source as being infinite and has always been quite extravagant in its use of energy.
Whether a preponderance for gas-guzzling SUVs or datacentres the story is always the same. The US mentality is there will always be infinite oil in the pump and infinite electrons in the cable.
Is that actually that much more than building 400GW worth of nuclear generation capacity (assuming 10 hours of nighttime)? I think that, that simple calculation shows that it is viable. Even a trillion is 1/32th of the US GDP. Construction over a decade should not stress the US economy at all.
I also think your trillion number is at least double the raw cell cost today.
Just by scaling up the existing technologies the full cost including installation and maintenance of a grid-scale battery in 5 years will be below 100 usd/kwh. So the price is max 400 billions or 1% of US GDP.
Current li-ion pricing is $19 per kwh (BYD's latest cell's pricing), so a single battery that covers all 4TWh would come in at 76 billion dollars. Still a shit ton but reasonable.
US has ~250 million cars. If they are all EVs (100KWh), it is 25TWh capacity.
Reserving 20% of that capacity to supply power back to the grid is 5TWh.
EVs can be charged for free during the day, ~20% of solar is lost due to curtailment. Even if a significant percentage (10 - 20%) of the people decide they are going on long drives from Florida to Vancouver and turn off supply-power-to-grid-for-money in the app, its fine, there are plenty of cars. It is the cheapest, most resilient, distributed and self-healing energy ecosystem. Only one problem, it won't make the super billionaires money. Unless it can be packaged in a way to make the super billionaires money, its a no go in US.
This infrastructure buildout costs nothing, people buy new cars all the time (~20M/year), just buy EVs instead. Because people are providing the capital and the storage infrastructure, they should be paid for it.
I think this is a pretty decent idea and hopefully within a decade at least california will be doing it. The problem though is that the biggest advantage of an EV is that you charge it overnight and never need to worry about going to the gas station or running low on power. I think a pretty significant number of people would want 100% charge every morning which makes the total addressable battery size a good bit smaller.
EVs with 300 miles of range are pretty common. The average round trip commute is 42 miles, i.e. you have on average more than 7 days before you run out from a full charge. Turning that into ~5.5 days in exchange for chopping tens of dollars a month off of your own electric bill because you're paying the daytime solar price instead of the nighttime something more expensive price is going to be very popular.
And when you know you're taking a long trip you can still do a full charge without needing to do that the other 99% of the time.
USA should explore Pumped Storage Hydropower more, it has same or better efficiency than battery storage and can use existing facilities like hydroelectric dams
Dispatch time, total installed capacity, and regional power grids.
Batteries have a much faster dispatch time than pumped storage. Pumped storage is good for offsetting reduction in other production capacity, and its dispatch time is faster than peaker plants, but its not as fast as a battery.
2020 numbers had 22,894 MW of capacity. More storage capacity for power is needed.
The "where are things located" is in the mountains where there's a significant elevation change. Most of the country is too flat to have pumped storage be practical. Installed pumped storage in Michigan (about 2GW of capacity there) does little for the power demands of North Dakota.
And then... there's also building new ones.
Even though FERC has issued licenses for three proposed pumped storage projects since 2014, no new pumped storage projects have come online in the United States since 2012. EIA’s most recent survey of planned power plant additions shows no pumped storage projects under development.
At 5:12 a.m. on December 14, 2005, the northwest side of the upper reservoir was overtopped when water continued to be pumped from the lower reservoir after the upper was full. This led to the catastrophic failure of a triangular section of the reservoir wall and the release of 1 billion US gallons (3,800,000 m3) of water in 12 minutes.
...
As of 2011, Ameren was reported to have paid a total of close to US$200 million in settlements related to the 2005 breach.
> The Taum Sauk pumped storage plant is a power station in the St. Francois mountain region of Missouri, United States about 90 miles (140 km) south of St. Louis near Lesterville, Missouri, in Reynolds County.
It has a hydraulic head of 860 feet and that represents the challenge - you need to find places in the US where you can get several hundred feet of elevation in a short distance.
Hypothetically, lets build one in Minnesota... Taum Sauk Mountain is the highest point in Missouri. The highest point in Minnesota is Eagle Mountain (in the boundary waters - you can't build there, pretend we can) ... the elevation to the top os 550 feet (about half what Taum Sauk provides) over 3.5 miles (Taum Sauk is just over 1 mile). ( https://maps.app.goo.gl/vHfWarwubaLvsf5z7 )
I hate these glib comments that don't remotely reflect reality.
Go look at the goods people got in countries that didn't practice capitalism. They were worse in almost every way, more expensive if you could get them and often the things people wanted were completely unavailable because there was no market incentive informing production.
In my experience, this sort of comment is usually followed up by some personal preference of the commenter. Great. I don't want the same thing, so I guess something has to mediate the resource allocation and effort of society towards providing that want/need at a certain quality/price.. and guess how we do that?
Are you looking at the USSR and countries that followed its model?
I don't think you can look at any authoritarian regime and make assumptions about how an economic system performs. The autocratic features tend to cause society to break down.
For an actual anarchist production model, look no further than Open Source. It's built by people who self organize, to solve problems that are real (and often to make silly, inefficient things). It's often (but not always) motivated by something other than profit. It's people who believe in a credo organizing themselves around that credo and building the tools to sustain that credo.
But imagine if open source was behind an autocrat, there's no way it'd be nearly as rich or flourishing.
Voluntary hierarchy is different from autocracy. If the community really felt Linux was better worthy Linus, they could make that happen. Linus isn't operating secret police that scare the shit out of people.
> I don't want the same thing, so I guess something has to mediate the resource allocation and effort of society towards providing that want/need at a certain quality/price.. and guess how we do that?
This. Freedom of choice is the simplest way to describe the advantage of capitalism (free markets) over socialism. Obviously, markets don't solve everything, but the US certainly is not some anarcho-capitalist's libertarian fantasy.
I guess the lived memory of socialism's failures has passed and experience is the best teacher.
> "goods people got in countries that didn't practice capitalism" are generally built to last.
No they are not. They were simple and people were poor so were forced for fix them for eternity, don't use them and absolutely not abuse them. Nobody gave a fuck about their craft in USSR. Everything was made from cheaply and from scraps.
Few things that might've survived (and I don't see _anything_ in my parents or in-laws homes) are literally prime example suvivorship bias.
Building all of the systems to train and serve our models, internal tools and infrastructure, (lots and lots of) research, our product and API surfaces, and more. We have a lot of problems that we can solve with code and there are all sorts of new projects spinning up every week.
reply