Surely part of the problem is that intelligence seems to be implicitly conditioned on embodiment, to the point that "covering only cognitive tasks" seems inherently ill-defined or arbitrary. Everything we do is a cognitive task. At some point, the criticism will be "sure, it can solve research-grade math problems, but it can't fold my laundry".
Even our large language models have an implicit embodiment in the domain of text (and more recently, multimodal inputs). That seems sufficient for certain things, and insufficient for others. I suspect that AGI that does everything a human can do eventually turns out to be fairly analogous to humans in terms of sensory input and domain output, even if the scale is radically different (e.g. thousands of robots uploading (touch, sight, audio, smell, etc.) sensory data to a single model, and each being actuated individually).
It's a clever move if that's what they're doing. They're restricted in China, and are likely to face stiff competition from Chinese chipmakers in the coming years. Acquiring the largest repository of trainable models and ensuring they run better on Nvidia hardware is probably one of the few moves they have for keeping ahead of the competition. I mean, it would be terrible for the consumer, but it does make me think that NVDA is a decent investment.
Well now that they own it, they can do whatever they like. They don't necessarily have to be optimizing the models for their hardware, they could just setup preferential pipelines to funnel users to their ecosystem. That seems fairly prudent given the challenges coming their way.
It's definitely not. I know Japan and Korea had a similar process when I was living there, and I was under the impression that most places operate this way. Your visa grants you right to temporarily stay in the country, and if you don't have a visa you have no right to be there. Sometimes there's a bridging visa to cover the gap, but a country isn't under any obligation to provide one.
No. Japan certainly doesn’t. If you want to renew your visa you go apply at the local immigration department and they mail you your renewal notice a month or two later. No need to leave the country at all.
It was a long time ago, so I might have the details wrong. I suspect it may have been expats from Korea taking a short trip to Japan to renew their visa, or even folks switching the type of visa they were on.
I've renewed my visa in Japan a bunch of times, including changing work category or visa status, and never had to leave the country for it. It's not standard for working or student visas, though you would have to leave if you were switching from a tourist visa to a different category.
Tariffs are a tax that buys you (at least) three things, all of which are difficult to quantify, and usually overlooked by economists:
1. Security. You make your own bullets. You're able to make and maintain your own military hardware. When another country cuts supply lines in their own self interest, you have options to advocate for yours, either through trade, making it yourself, or, yes, even force.
2. Economic opportunity for your children and grandchildren. And no, not just "factory jobs" as is commonly sneered, but a community of experience and expertise. An engineer that gets to interact with the products they design in that kind of environment is a better engineer. A country that gets to see and interact with the products it makes has a sense of pride and better social cohesion.
3. Social buy-in and consequences for companies. A Chinese car manufacturer doesn't give a shit if cutting costs killed your grandfather. A local company has to, because the executives and workers are all members of the community. I was in Germany during dieselgate, and the sense of disgust at the company was obvious. They felt that.
These things are, imo, important to the health of any country, but they can only really be valued in retrospect. So yeah, tariffs are a tax, but some things are worth paying for even if what's being purchased isn't necessarily obvious or easy to quantify.
Yes those are all good things. Tariffs aren't the only and almost never are the most sustainable long-term way to get them. I'm not old enough to remember the protections against japanese cars, but I know they didn't solve the problems with US auto industry, for example.
I would guess the optimal course to those ends would include some targeted tariffs. I am damn near certain that the optimal course isn't tariff-centric. Every option in a global economy (or any economy) is coupled and never limited to the first-order obvious effect.
As a very simple example that I see ignored often on here: High paying jobs are good, but increase the cost of the ouput. Consumers are also people, they benefit from lower prices. You can never turn one dial independently. There are pretty magical dials though, like education and trade. Some dials are better than other. All dials treated coarsely are worse than dials treated as the multidimensional optionalities they are.
I generally favor increasing productivity and QoL via net-positive interactions. I default to disliking fixed-pie-slicing until convinced otherwise. With all of the caveats to this position as above.
The only repeatable path to industrialization and the benefits thereof is the combination of protectionist policies -- be they tariffs or subsidies -- with a focus on exporting goods and competing in the global market. This is historically true, essentially without exception. You need both, because without global exports, protectionism makes companies lazy, and without protectionism, you risk exposing strategic national assets and supply chains to what is essentially unfair competition.
I mean, I admit that I'm not competitive with people desperate enough to work 12 hour days, 7 days a week, while the grandparents raise the kids in a 10 person household. But I'm probably capable of more sophisticated work due to my education and ability to specialize. Removing all barriers to trade essentially forces companies to offshore and train up nations that may not have your sovereign interests at heart in order to compete and provide a return for investors. It's a short term sugar hit that erodes the nation-state. This is why we strictly limit offshoring for certain strategic industries, but the truth is that all manufacturing industries are strategic to some extent, because they form part of a logistical web.
As for whether or not such a system is sustainable, it's worked for the world's manufacturing powerhouses for decades. The reason they're all falling to China now is because enough of the global supply chain has shifted there that it's now impossible to compete. Assuming those nations don't want to become services-based economies (which has been a disaster for Australia and the UK, neither of which can manufacture a tin of beans), protectionism is the only path forward. And indeed, China is utterly diligent in protecting it's own industries. This isn't fixed-pie-slicing, it's just pragmatic recognition that a government really does want to be able to fix the fighter jets when it needs to, and at some point, that means having enough industry in place to train the guy who does it. And economists, who don't really believe in or care about the problems of nation-states, probably aren't the right people to listen to when that's a concern.
I don't see how any of this works unless the rest of the world agrees to be less productive in the meantime. Apply what you want for the US to every country in the world. If your policy is the solution, then I'd assume all countries would copy it, and we'd have no one to export to. Or if they follow what I consider a more reasonable policy, they will quickly find cheaper places to buy shit, take advantage of basic things like comparative advantage vs. the protectionist (1) US, again leading to no exports.
The US is great, don't get me wrong, but the 1950-70s were based on an extremely fortunate combination of circumstances. It was almost certainly an exception rather than a norm to revert to. The resulting economy and middle class and mobility is aspirational, not easily reproducible. And I am certain not by pure protectionism. I mention this because you said repeatability, which I guess I assumed meant of this time that seems most aligned with what you want.
Edit- just re-read and you agree staying global competitiveness is important. Ok, how? I don't see any route there (in manufactured goods) without going hard on automation. I'm guessing (sorry) you basically think automation is bad because manufacturing jobs. Maybe we actually align here though?
I recommend to too many people the book "how the world got rich" https://a.co/d/0ehTVbK4. It's short, it's about what the dominant, generalized correlates with growth are. My insertion is that economic growth is good for everyone and explains without much exaggeration all good societal progress. Just as one concept from here that explains an enormous fraction of global economic dynamics, look up " catch up growth". Sorry if this comes across as you not already knowing about this, I recommend it because I personally found it such a good resource.
(1) This is me inserting a projection, I don't know for sure. I would need a lot of convincing that protectionism and isolationism makes us more competitive overall. I completely understand all of your arguments about skill and knowledge and that other good stuff I already agreed was good. What I don't agree with is that your approach leads to us being to compete in a global market, especially given domestic political reality. It seems your system relies entirely on being able to export. The one thing we do export btw is services, the thing you shit on. And again, I wish we weren't so dominantly service based. I want all the good things you want that come with manufacturing and more domestic self-reliance. I just don't think you can will into existence of a manufacturing dominant prosperous middle class 60s economy that survives global competition, which your export driven plan requires.
Edit- as another aside, I believe economics is increasingly a captured field. So I probably distrust economists on policy at least as much as you, in general. But when they all agree going back a while with data to back it, I weigh it higher. I also feel bad for economists because historically the actual policy they want isn't what makes it through, to the point the preconditions for their policy to work no longer apply. So again this is perhaps something we agree on. But this doesn't invalidate a bias toward free trade is wrong. Implementation matters. See: the parts I agreed on wrt international trade policy what feels like 20 replies ago.
Edit2 - one counter to my position is that with protectionism and manufacturing we can rely more on internal market demand more because we cover more of the market needs, like China is trying to do. Just thinking out loud trying to figure where I'm wrong. A lot of the good times were driven by internal markets. And this at first glance seems more consistent with protectionism. But you want export driven so I'm not sure.
I just remembered this essay that kind of explains some of my thoughts on how to judge economic policy http://bastiat.org/en/twisatwins.html. Not super relevant but kind of just in case you hadn't read it. Not just the broken window fallacy but the idea of diffuse benefits vs. visible downsides.
Sure, if every country did this for every technology, it wouldn't work. But in reality, they don't. Even ongoing industrialization isn't a given once it's been achieved, because the bar for what constitutes "developed" is always changing, and it takes effort to maintain. We even see deindustrialization, both historically and in the present. There will always be someone to sell to.
You're focusing on the US, when what I'm saying is historically true for essentially every nation that ever industrialized. Many of those countries are still industrial powerhouses today. The US had a great run from the 50s to the 70s, but it was the world's largest exporter by 1914 thanks to WWI, and didn't relinquish that spot until 2003. Similarly, Japan and Korea didn't have any such luck and still managed to both develop and maintain high living standards even through economic hardship, thanks to their industrial foundations. Taiwan uses its chip industry, again heavily export based, as a shield. China is following the same playbook today, first during China shock 1.0 20 years ago, and now with higher value goods. The only real exceptions to this rule are Singapore and Hong Kong, both of which exploited the fact that they were strategic waypoints in global shipping. Even then, Singapore is one of (if not the) largest oil refiners in the Asia Pacific region, and exports a tremendous quantity of refined petroleum products. Singapore isn't strictly protectionist, but it makes heavy use of cheap Asian laborers who will never, ever be citizens, and it engaged in massive skills transfers from other nations during the 80s and 90s (my own family was part of that wave). And of course, the major European powers all built themselves on the back of protectionism and exports. The erosion of that directly resulted in the crises we see today.
I have no issue with automation, and I agree that investors and business will always want to exploit cheaper economic conditions elsewhere. What I'm saying is that any government that wants to preserve living standards and maintain an industrial base needs to have the backbone to tell those interests "No". They are part of the social fabric of a country, and the backbone of it's defense. Investors will make less money, but I don't see that as a bad thing. They would actually have to allocate money more effectively within a country to turn a profit.
I'm not shitting on services, they're necessary. But services can't manufacture bullets when you need them, and countries always need them eventually. Not to mention, we live in a time when even services being offshored, so what's left? We build houses and make coffees for one another, while banks and mega corporations that are beholden to no border lord over us?
For what it's worth, I agree with you that economics is a captured discipline, and I applaud your devotion to testing your own ideas. But the stuff I'm pointing out isn't economics, it's history -- things that actually happened. In my opinion (and it is just that) any economic theory that is at odds with history is straight up garbage.
Thing is, absolutely none of what you’re saying applies to Trump’s tariffs. Because he’ll remove them two weeks from now if the right person flatters him the right way.
So there will be no build out of manufacturing capacity. You’d be a fool to spend tens of millions on that when at literally any moment the market will revert back to the norm.
This administration’s policy making is insane. I’m surprised folks are still trying to sanewash it.
I didn't say anything about politics. Whether or not industry believes this is a permanent state of affairs, or destined to end with this administration, has no bearing on whether or not tariffs can be a useful tool. My issue is with trivially wrong economic dogma.
This. Involve your kids in your hobbies. If they aren't interested, that's okay too -- you can be involved in theirs. Some hobbies have a finite life anyway. You aren't going to be a gymnast forever, for example, and that's okay.
In my experience, good software engineers are often better "engineers" than those working in more physical disciplines, because they get a lot more practice. But traditional engineers have a far better culture when it comes to testing and validation. Since the cycle time is longer, and the cost of mistakes is higher, analysis and testing are generally baked in from the beginning. In software, it's easier to skip that stuff. But software engineers who do get indoctrinated into that culture learn most of the same lessons that traditional engineers do (i.e. how can a component be tested and maintained? what makes for a good design?) and the speed of development means they get more exposure to those types of challenges in general.
In my experience this works both ways too. Good engineers working on physical projects have picked up on some of the useful practices that good software developers have adopted for maintaining progress in the face of ambiguous requirements until they can be clarified and allowing as much flexibility as possible without necessarily compromising quality in the meantime. I believe these relationships can be summarised as something like "Good people keep open minds and learn from what has worked well for others".
The problem for regulating software development is still who gets to formally determine who the "good" people are. This kind of thing should clearly be objective and evidence-based but what useful evidence do we have available?
In physical engineering disciplines there are often clearly evident problems if something was built without being adequately specified by the responsible engineers. In a disastrous case a bridge might literally fall down but you're also going to see that a bridge wasn't designed properly if it's distorting in ways it shouldn't under loads that it should be able to support. There are lots of experienced engineers who have proven records specifying buildings or planes or ships that need to not break using established and peer reviewed techniques.
In software we can all agree catastrophic failures that result in loss of life or half the Internet going down are obviously bad. For something controlling a life-saving medical device or the launch authorisation system for the nuclear missiles we can probably all agree that the answer to what quality level we want in the software is "the best quality we can achieve". But those systems have unusually serious consequences if anything ever goes wrong and probably also very high development budgets that can justify such an extreme position on quality. In general we don't have clearly defined levels of software where different trade-offs between cost and risks and other factors might be considered reasonable and acceptable. Nor do we have well tested and universally accepted standards for how to reliably achieve a specified quality level.
> The problem for regulating software development is still who gets to formally determine who the "good" people are. This kind of thing should clearly be objective and evidence-based but what useful evidence do we have available?
There is zero useful evidence because there is no one to collect it.
The Institution of Civil Engineers was founded in 1818, after decades of random civil engineering societies in Britain doing the exact same thing we are now (running around like chickens with their heads cut off). It wasn't until after the ICE's Royal Charter a decade later that civil engineering began to get really systematized into the "real engineering" we know today and that charter effectively established them as a regulatory body that allowed that to happen.
There is zero useful evidence because there is no one to collect it.
I'm not sure that is entirely true. There have certainly been a few people who have attempted to study what did or didn't work in industrial settings - either pure academics or people working in industrial research labs. But I agree that currently we have nowhere near enough data to form robust conclusions about almost anything in this field and I think this is the strongest argument that the industry is not ready for any kind of licensing and regulation regime.
It's probably worth distinguishing software engineering from the ability to program. Programming is a skill that is similar to reading and writing in that it's a type of literacy, and it shares a common theme with literacy in that even though most of us can read and write, not everyone is good at it. Professions that leverage that unit of literacy build on top of it, and take literacy as a given.
For instance, a lawyer spends most of their time reading and writing, but no one would ever say "you can read and write? Have a crack on our legal team". It's a particular type of writing, and the writing is really a means to an end. This is a distinction that is lost in software development. Every engineering discipline now learns programming, but the ability to write code shouldn't be a license to write software in the same way that knowing how to read and write doesn't just grant you the ability to join a legal team and start writing contracts.
Fossil fuels contribute to climate change. Methanol is likely to power global shipping in the coming decades, and there's zero reason it can't also run cars. The reality is that hybrid vehicles have all of the same advantages that BEVs have, but they can also be refueled quickly without heavy reliance on centralized infrastructure. Aside from being the reason that we're all driving ICEVs in the first place, it's a compelling enough advantage that such vehicles will continue to be with us for a long time.
If you would like bet how many methanol vs battery electric cars will be on the road in 2030, 2035, and 2040, I am happy to take the battery electric side of a Long Bet [1], with $10k to the winner's choice of 501c3 charity. Let me know. If battery electric vehicles are already 1/4th of global annual light vehicle sales, and there is ~20M-30M of battery electric vehicle manufacturing capacity globally, with no methanol vehicles on the road today, what does the future look like?
BYD intends to be the largest automaker in the world in five years, and makes EVs exclusively [2]. They have already surpassed Ford in international vehicle sales. Their flash charge charging system can charge the vehicle's battery from a 10% to 70% state of charge in five minutes, and from 10% to 97% in nine minutes. Fossil fuels for light vehicle use cases are dead, and the state of the art for EVs and charging will only continue to improve.
> Aside from being the reason that we're all driving ICEVs in the first place, it's a compelling enough advantage that such vehicles will continue to be with us for a long time.
Take a look how many refineries are left in the world. No new ones will be built, and the existing ones will only get enough investment to keep them operating. How quickly they are taken offline is how fast ICE vehicles are phased out, because you're not running your ICE vehicles without refined petroleum products [3] [4]. They are the weak link in the chain.
> WoodMackenzie: A total of 101 out of 410 refineries around the world are at risk of getting shut down over the next decade. This number represented 21% of global refining capacity. Inflated operating costs of refineries are an especially important risk factor for future profitability as well as their investments in decarbonization.
I must be hallucinating all of those hybrid BYDs I keep seeing on the roads, then. You say 'EV', but the distinction that matters is battery electric versus high voltage hybrid, and low voltage hybrid. And hybrids, which are uncontroversially ICEs, will be with us for decades. Yes, even the non-plugin kind.
The world is not running out of fuels any time soon. If petrol goes away, I'll convert my cars to LPG. If that goes away, I'll convert them to ethanol or methanol. The whole reason ICEVs became a thing in the first place is because they don't require centralized infrastructure to be useful. Incidentally, this is the same thing holding BEVs back: until you solve the street parking situation the way China has, the market will be reluctant to switch.
As for your bet, I'll make you a promise instead: I'll put a few extra miles on my car today just for you. And who knows? If this becomes a social movement, maybe we could see new refineries open to meet the demand.
The issue with ICEs is the fossil fuel part, not the engine part. Synthetic hydrocarbons and alcohols don't contribute to climate change at all. And alcohol fuels are both easy to make and about as good as hydrocarbons.
> Synthetic hydrocarbons and alcohols don't contribute to climate change at all.
If you are able to manufacture them in a completely emissionless way. That requires huge energy input though. Certainly has its uses, but too expensive for personal cars I'd think.
What I think would make sense is to use methanol as a seasonal energy store in temperate climates. Manufacture from PV in summer, burn in winter.
To be fair, DIY natural gas production from kitchen waste is a thing some people have done. It’s surprisingly doable and produces surprisingly large amount of usable energy.
Methanol is the obvious choice, because there are so many ways to make it. Shipping certainly looks to be moving in that direction, though ammonia is also being trialled. Ethanol via fermentation is also doable at scale, but most effective pathways rely on using arable land.
If we're talking synthetic fuels, you'd probably run cars on methanol rather than a hydrocarbon. Planes absolutely need kerosene, but cars can claw back most of the performance of petrol with alcohol fuels, and methanol is the simplest and cheapest to make. Certainly shipping seems to be going this way.
By the time you get to variable valve lift and timing, the complexity of hybrids is either a wash or in favor of the hybrid when compared to a traditional engine. Electronics have been integrated in vehicles for generations now, and going hybrid lets you get rid of components and reduce overall mechanical complexity. It's not "two separate systems", it's just the obvious endpoint for internal combustion engines (and a process that is both a century old at this point, and already mature in other industries), because you get a strictly better machine from both a performance standpoint, and a design and engineering standpoint.
Not to mention, much of the complexity in ICEs comes back to cooling and lubrication, both of which are also concerns in BEVs.
This is what the person that I was responding to wrote, and it's objectively wrong:
>Hybrids are the worst of both worlds when it comes to maintenance. They have their benefits and are key to fighting climate change, but mechanically you inherit all the problems from both sides.
Also, much more complex is a value judgement. Complex in what way? High voltage electronics isn't exactly "not complex", and the coolant loops in BEVs are more complex than those in regular cars, because the tolerances are tighter.
I'm going to go out on a limb and assume that you mean mechanical complexity, but even then, that's not a given for ICEs. Gas turbines can have as little as one moving part, and there are piston engines that can get away with as few as four (the piston-conrod-crankshaft assembly and a compressor wheel).
At the end of the day, a good design has the minimum amount of complexity necessary to perform a given task, and no more. Given that BEVs are fundamentally unable to replace ICEVs in several niches (e.g. those relying on on-street parking, off-roading), maybe a little more complexity is warranted? As is, they aren't a 1:1 replacement for a standard car, and that's okay, but they're being marketed as such which is a recipe for failure.
I don’t think that a single-moving-piece gas turbine powered cars are currently on the market? And how many are single piston? Go carts and motorcycles maybe, but cars?
And I was using complex as a shorthand for „more things to maintain“.
A few companies have tried small APU range extenders, but there wasn't really a market for it at the time. The point was more that complexity is relative and not necessarily a given, and it's not necessarily a bad thing if it's required. A door is simpler than a bed, but I'd still rather sleep on a bed.
Maintenance is a function of usage. BEVs still have coolant and lubricants, which means they also have pumps. They have a HVAC system which also means filters and refrigerant. Yes, you don't have to replace spark plugs and a timing belt, but both of those are infrequent and generally easy. You might have to replace brake pads less, but hybrids share that same feature. Hell, hybrids don't even have fan belts, alternators, ignition motors, and transmissions these days. They share nearly all of their underlying foundation with BEVs, for most of the same reasons you would want to go BEV in the first place.
Even our large language models have an implicit embodiment in the domain of text (and more recently, multimodal inputs). That seems sufficient for certain things, and insufficient for others. I suspect that AGI that does everything a human can do eventually turns out to be fairly analogous to humans in terms of sensory input and domain output, even if the scale is radically different (e.g. thousands of robots uploading (touch, sight, audio, smell, etc.) sensory data to a single model, and each being actuated individually).
reply