I'm too lazy to double check the numbers, but as far as I remember, Germany in order to increase it's average generation by 10% had to expand capacity by 70% in solar plus wind. With stats like this, there's a thin line between progress and waste. And all this while we have nuclear.
(How the world really works, Vaclav smil if anybody is less lazy than me)
No, capacity factor is a distraction. The only meaningful question is whether it is cost effective or not compared to other production methods,, and solar - taking into account the low capacity factor - is starting to look very good.
I would say as electrician in Bavaria: there are enough empty roofs for solar. Especially in poorer neighborhoods. I saw similar numbers and they are scary: to reliably replace conventional power plant one needs 20x the power of wind and solar. And this hardware must be imported from China, there is no large scale production of solar equipment in Europe.
I love the water/ice metaphor, but the author tends to completely ignore the physical world.
Example with cardiologist - we all know what happened to the radiologist prediction.
Example with defence (or war) becoming mostly a case of having a better AI model: well, try to win without a solid, distributed production capabilities, energy access and safe supply chains, in a geographical disadvantage.
Embodiment is coming, but it will require moving a lot of atoms.
Also, even in text heavy domains, a lot of knowledge is not written down, often of purpose (especially in legal), and that's the juicy part...
For a normal person, the only real words in this sentence are "get", "with", and "image", but the last one does not mean what they would think it means.
Even WIX needs some level of tech savviness, usually beyond 90% small business owners.
And Instagram? Well, one of the main points of having a restaurant is to tell your friends about it, so the Instagram profile is more important than actually having a real restaurant.
Goons work MUCH better than rockets for intimidation, and actually scale much better.
Rocket is obvious and spectacular. Those are for amateurs.
A journalist got beaten up to the brink of death and will never walk again by 'unknown perpetrators'? Well, it's a dangerous country, and he had it coming, maybe some concerned citizens went a bit too far, but our dear leader cannot watch over everybody.
Scaling: do you think other journalists will not take notice?
And he will still be alive to reminder them how they may end up.
If you want to see how far imagination can go here, look up Artyom Kamardin and think how would you behave after hearing his story .
The pyramids in the article are missing "energy" and "capital": in the world where intelligence becomes a commodity only those two matter. Capital to buy the hardware and install it, and energy to run it. Models already are a commodity, and "physical is the new king".
As a side note, if you believe that because of the agents doing most of the work we will face the problem of what do we do with the all the free time (with presumably UBI in place), please contact me, I have a bridge to sell you.
Exactly this. General purpose intelligence and automation allow a clean break between capital and money as we understand it.
Money is used only to pay wages. It has intermediate uses, storage, leverage, etc but at the edge all you can do with money is pay wages. Nobody pays the dirt when you take out the metal, nobody pays the forest for the trees, nobody pays the chickens for the eggs or the cornfields for the crop. Ultimately it’s wages all the way down.
If you don’t have to pay wages, you don’t need money, you just need self replicating automation, energy, and access to land and resources mine or farm the raw materials you need.
If you zoom out to space, it’s essentially grey goo with maybe some humans at the top for a while at least.
Inside the gilded walls, if you want something, you don’t buy it, you build a factory to build it, even if it’s a one off.
If you need money for something because you don’t have enough reach and power yet, you just mine gold or bitcoin.
You don’t build products to sell, you don’t need customers. You just need energy, resources, and the kind of power that comes with 20 million self replicating robots to project your will. You don’t need government, and you certainly won’t be funding it. Government is a really complex system to administer a monopoly of coercive force for the common good. You have your own monopoly of force operating for your good.
The difficult part in the capital flywheel has always been humans in the sticky parts. Take them out and that baby will hummmmm.
Pesky humans outside the gilded walls will be accommodated in the same way we accommodate ants at a construction site.
It's not certain this is the future: the obvious trade off is lack of flexibility: not only when a new model comes out, but also varying demand in the data centers - one day people want more LLM queries, another day more diffusion queries.
Aaand, this blocks the holly grail of self improving models, beyond in-context learning.
A realistic use case? More efficient vision based drone targeting in Ukraine/Taiwan/ whatevers next. That's the place where energy efficiency, processing speed, and also weight is most critical. Not sure how heavy ASICS are though, bit they should be proportional to the model size.
I heard many complaints about onboard AI 'not being there yet', and this may change it.
Not listing middle east as there is no serious jamming problem there.
In a not-too-distant future (5 years?) small LLMs will be good enough to be used as generic models for most tasks. And if you have a dedicated ASIC small enough to fit in an iPhone, you have a truly local AI device with the bonus point that you get something really new to sell in every new generation (i.e. acces to an even more powerful model)
Yes but not in five years. The chips will be dirt cheap by then. We‘ll get “intelligent” washing machines that will discuss the amount of detergent and eventually berate us. Toasters with voice input. And really annoying elevators. Also bugs that keep an extremely low RF profile (only phoning home when the target is talking business).
Perceptible latency is somewhere between 10 and 100ms. Even if an LLM was hosted in every aws region in the world, latency would likely be annoying if you were expecting near-realtime responses (for example, if you were using an llm as autocomplete while typing). If, say, apple had an LLM on a chip any app could use some SDK to access, it could feasibly unlock a whole bunch of usecases that would be impractical with a network call.
Also, offline access is still a necessity for many usecases. If you have something like an autocomplete feature that stops working when you're on the subway, the change in UX between offline and online makes the feature more disruptive than helpful.
It doesn't have be to true for all models to be useful. Thinking about small models running on phones or edge devices deployed in the field that would be a perfect use case for a "printed model".
The real benefit, to a very particular type of mind, is that the alignment will be baked in ( presumably a lot robust than today ) and wrongthink will be eliminated once and for all. It will also help flagging anyone, who would need anything as dangerous as custom, uncensored models. Win/win.
To your point, its neat tech, but the limitations are obvious since 'printing' only one LLM ensures further concentration of power. In other words, history repeats itself.
The edge is ownership - of GPUs, capital, connections and distribution channels (this includes SEO).
Also, SEO will be meaningless if LLMs will be the main discovery channels. Much less transparent, and we are already seeing that, for example the "what about South African ***code" grok system prompt manipulation from a few months back.
Anecdotal evidence to confirm: I had two false alarms from an unrelated MRI scan, and beside wasting a lot of time on diagnosing them - it was also extremely stressful.
My father is a part of "full body PET scan every 3 years" program as part of post - cancer treatment, and it worked twice: early detected lung and prostate tumors, both removed.
> My father is a part of "full body PET scan every 3 years" program as part of post - cancer treatment,
These treatments are wonderful and it is great that they exist. But many people fail to understand the difference in terms of pretest probability, etc.
I can absolutely see the heavy psychological impact pending biopsy results may have. People are quick to discount these issues when you raise them as a concern, but only if they never went through this stress themselves
Sorry, I missed a key detail. What you are describing isn’t the ‘whole body mri’ I’m referencing by to.
People in high risk situations like multiple myeloma, or various metastic diseases, or system conditions are a whole different category and there is clear benefit to screening them.
> My father is a part of "full body PET scan every 3 years" program as part of post - cancer treatment, and it worked twice: early detected lung and prostate tumors, both removed.
My mum gets scanned a little more frequently than that, following treatment for an inoperable tumour in her lung around five years ago. During treatment she was getting scanned every three months or so, and it was remarkable watching this thing go from the size of a tangerine, to actually expanding a bit and looking "fuzzy" once the drugs kicked in, to being the size of a plum, then the size of a grape attached with a little thin thread of tissue, to being a thing the size of a pea. Now there's a tiny ripple of scar tissue that no-one wants to investigate further, because if it's not doing anything let's not poke at it.
There is a roughly pea-sized "thing" on her adrenal gland that was a bit worrying because anything like that is going to get intimately involved with your lymphatic system and then it's going to metastasise like hell. But it neither got bigger nor smaller in the nearly six years since the first scan, so it can't be that important.
This is one of the great things about the NHS, especially here in Scotland where we have (possibly as a result of the weirdly high levels of cancer) some of the best oncology services in the world.
If we'd lived in the US, the insurance companies would have taken one look at an 83-year-old about to become a grandmother and sent her home with a bottle of morphine to die. As it is, she's doing very well and got to see both her grandchildren start school.
> in Scotland where we have (possibly as a result of the weirdly high levels of cancer)
Interesting, I had no idea and just looked this up[0]:
> Scotland had the highest overall incidence (446 for men and 379 for women per 100 000), and Wales had the second highest rate (450 and 366 per 100 000), compared with 394 and 338 per 100 000 in England and 394 and 345 per 100 000 in Northern Ireland.
Thanks, she had little to no side effects from the experimental immunotherapy drug. She'd said at the outset she didn't want to be a "chemo zombie" having been through all that about 20 years ago (she's had four different cancers throughout my lifetime, treatment getting better every time).
It's fucking expensive, but immunotherapy is really a miracle cancer treatment. We're not quite at "Oh you've just got a wee bit of cancer, we can give you something for that if you see the pharmacist on your way out" but we're not far off.
Thank you for writing this. It was also my first impression after seeing not only spec driven dev, but agentic systems that try to mimic human roles and processes 1:1. It feels a bit like putting a saddle on an automobile so that it feels more familiar.
reply