> In ~1997 this level of compute power would have been enough to get on the TOPS500 list of the fastest computers our civilization is capable of creating.... now something that consumes only 54 sq mm of silicon that gets put into a $15.00 gadget.
... and it's incapable of doing anything useful without even bigger computers somewhere else.
I often wonder if there is anything else in society that could scale in a similar way to moores law.
Is there anything else we could use a lithographic-like process to make massive numbers of very cheaply, yet still find buyers for?
Floors of a modern skyscraper are nearly that - they're made with moulds that get reused again and again on every floor. But to get decent scaling, you really need to make instances in at least 2, and preferably 3 dimensions in one go.
We've already done it with things like the printing press that can churn out books very cheaply - but it turns out that we've now reached the limit of the numbers of books people want.
Humans also reproduce exponentially - with a sufficiently high birth rate, numbers of people can rapidly grow. One needs enough space to support that growth, and we are sadly lacking in that... Perhaps Mars will allow us to practice our fast-scaling of civilization?
Plastic molding of different types (injection, extrusion, blow molding, etc) enables massive quantities of physical objects to be produced at a staggeringly low cost and high rate of speed. This has of course enabled the production of many things that are beneficial to our lives. It has also increased the amount of waste that humans produce, to say the least.
> Perhaps Mars will allow us to practice our fast-scaling of civilization?
For a few decades (assuming fixed rate), until the population doubles and we'll be overpopulating two planets. We need to either slow down on the exponent - not just population, but also markets - or figure out interstellar travel rather quickly.
Astronauts on the ISS have proven that living in space even for a few months is extremely unhealthy. Seems like that might be a harder problem to solve than building the space cities.
High energy cosmic rays are tough to shield against. On Earth there’s 60 miles of atmosphere, in space you would have to reproduce that with several meters of dense material.
You'd have to pack nuclear reactors to provide power (though that's a given for any kind of serious activity in space anyway), and your scientific and operational capabilities would be limited/made difficult by the lack of sunlight.
If you can lift exatons of steel into space, you can probably already afford to solve the problem of direct synthesis of all, and every fertilizer, and probably nearly every other chemistry process just by throwing more energy at it.
Sure. But you even quoted me saying not just population, which is the key point. Stabilizing population doesn't buy us much if the market - and thus resource exploitation - is still growing exponentially.
Economic growth doesn’t imply increased resource use. Actually it’s more likely to be the opposite - growth is more likely to be in services and moving to electricity, which is renewable. Our use of many nonrenewable resources has peaked and gone down as population has gone up.
> Our use of many nonrenewable resources has peaked and gone down as population has gone up.
Our use, or the growth rate of use? It's good when the latter goes down, but we're not out of the woods as long as it stays > 1.
> Actually it’s more likely to be the opposite - growth is more likely to be in services and moving to electricity, which is renewable.
Most services have consumables, and electrification of everything is very production-intensive. And while digital services may be growing faster than manufacturing these days, manufacturing is still growing.
Yes, in theory, we can have economic growth without commensurate manufacturing growth, but I think we're still couple abstraction layers short in our economy before we could think about running it completely on imaginary things.
> Humans also reproduce exponentially - with a sufficiently high birth rate, numbers of people can rapidly grow.
With a sufficiently high birth rate, daycare, cheap real estate, schools with less than 100 pupils per class, free higher education, and solid, assured pensions, and state elderly care.
Because economic pressures are much stronger than humanitarian pressures, so standardized things are optimized to be as small and cheap as possible. Hence slums, standardized testing and junk food.
What’s actually happening in the US is houses are getting larger because city governments are run by old people who nitpick all apartment buildings (because they think they’re ugly) and approve unlimited expensive single-family houses, which have only gotten larger over time.
There's a lot of incentives in the US, but to simplify it: per-house-price and education funding are linked, via property taxes. Taxes paid by one home vs cost to educate average number of children per one home.
If we funded more local public education directly from non-property-tax sources (or made public education more efficient), a lot of the resistance to lower-cost housing would disappear.
The Eastern Bloc kind of tried, you can see how it went.
Saying it being a half-success would be a complement.
If a Union's citizen could've choose in between Union's kindergarten, and an American one, he wouldn't need to choose, though some people say daycare wasn't as bad in Western countries of the bloc.
I would still admit, Union's secondary education was a rare somewhat half-success story, and it only looks so good if you compares it to USA, but not so much other Western bloc countries.
China tried mass housing along Union's model, and it failed miserably (though this failure could've probably been called a success by standard of places like California.) It took it to adopt capitalist real estate model to get where it is now, and even today, Beijing has no viable recipe how to house 60% of population that got to stay behind in the industrialisation drive.
The bling-bling Chinese 1st tier cities most Westerners get to see only represent how 5% of China population live, and even they get to house most of its citizens in rather gloomy "bedroom district" type suburbs which most outsiders lauding "Chinese housing model" never see.
The quality of housing in China drops with inverse square root of a distance from nearest 1st tier city, and from downtown to suburb.
> Why haven't we better stardized housing, education, and food to scale?
In other words, some most essential living needs are hardest one to scale, and mostly not for physical, but for social, and organisational reasons.
I don't think it is that odd, it's just a case of costs, scale and who you're designing for.
A TOPS500 computer is gonna set you back a few million $, so you're only going to have one of them and it's economic to employ a team of well qualified people to write software and interface relatively directly with the hardware.
Now that same computing power can be purchased by ~anyone and most of those people are not specialists and have essentially no ability to write software for the platform, so (from a design perspective) it's a waste of time designing an interface that allows them to do so.
If your goal is to have a small powerful platform aimed at people who want to run software on it directly that's completely doable. Just see the raspberry pi and the raft of similar devices now available.
Sort the opposite effect with software. Something like "In ~1997, some of the largest desktop software applications would have been only a fraction of the size of today's smartphone apps."
I was thinking of the MS Office Android app. You can get it running on a $50 tablet. How powerful would it be vs let's say Office '97 on a Windows 95 PC purchased for around $2000 25 years ago?
1) the age (now?) says “hundreds of thousands”, up two orders of magnitude from this. 2) it’s referring to the WiFi IP core, not the SoC itself, which seems an appropriate estimate. Just reading the 802.11 standard would account for a chunk of that time.
Even if we could, in 2021 it's far easier to pick up a more powerful, more appropriate SBC and simply use that for a project. A Raspberry Pi 4 is only $10 more and immensely more capable, as well as well-supported and heavily documented.
Unfortunately, it's not in manufacturers' best interests to make these devices hackable. All costs considered, Amazon is likely selling these near at-cost. They're not selling the hardware as an independent product. They're selling it as an extension to their larger platform (Alexa). As soon as a product becomes hackable, some portion of sales are lost to enthusiasts who buy one thinking they might hack it some day. If this gets out of control, eventually the company is subsidizing enthusiasts and the whole proposition becomes a net loss for the company.
The whole thing might even run at a net loss with corporate arguing it is an investment.
Amazon knows voice agents are on track to become a primary human-computer interface, a user agent for the internet, and Amazon wants to own a big share of that pie, because that is a gatekeeper tech. Those who own it can ensure people buy on their marketplace, and companies that want to sell to these consumers must pay a fee. Amazon can't reasonably compete with google or apple in the "voice agent on the smartphone" segment until some antitrust forces those to offer a choice of agents, so they do the smart thing: push hard for a quasi monopoly on ambient microphones.
Over time there could be an antitrust to tell Amazon they must open "their" voice agent infrastructure for the competition, and they will use this "investment" as an argument why that would be unfair to them. Even if, they will rather add an API limited to accredited competition instead of giving consumers root on the hardware.
The whole concept is hostile to the consumer by design: imagine a keyboard would include a non-replaceable typing agent that can only type on websites that are partners with the keyboard manufacturer (as in "buy the skill for your keyboard to type on HN") and all your inputs are first sent to the manufacturers servers. Hey you get auto-complete and the keyboard is cheap. That is the state of voice agents.
> It's easy to imagine how many of these will be electronic waste in not that many years.
I'm pretty in favour of requiring some level of "the user can run whatever they want, supported but at their own risk" for every (reasonable) device for exactly that reason.
It’s super impressive that someone can just look at a die image and know what blocks are what. Kinda wished I knew more to about ASICs to be able to do the same.
... and it's incapable of doing anything useful without even bigger computers somewhere else.
Progress is so odd sometimes.