Why we need Taiwan

Taiwan is a shining example of the value of a relentless drive for finding innovative ways of improving people's lives, while preserving their dignity and privacy.

I recently had the opportunity to spend a week on a research trip to Taipei. I attended a ‘smart cities’ conference and met with many Taiwanese city managers, business leaders, government officials, and several other wonderful locals with whom I hope to stay in touch. I returned home more excited about the future than I have been in years.

Infrastructure

Right from exiting the plane, I was impressed by the pragmatism on display; their arrivals custom hall looks a little dated, but is clearly designed and staffed to move people through quickly and efficiently. It stood in stark contrast to my departure experience at Auckland Airport, which is undergoing a seemingly continuous process of refurbishment. It’s a place with frequent additions of beautiful wood panelling and artwork, but also one where structural inefficiencies are left unaddressed while new ones are added¹.

While form too often trumps function in the West, pragmatism is on clear display in Taiwan. Driving from the airport into town, one sees an extensive system of above-ground expressways. These were built in the 80s - early 2000s while the national economy was still lagging far behind the West. The structures are all simple reinforced concrete designs without ornamentation, lengthy ramps, wide shoulders, or any other modern features. Some of the older structures show signs of spalling, particularly on the underside of the box girders. This is almost certainly indicative of lower-cost materials & techniques being used during construction. Condition assessments are clearly regularly conducted, as I failed to spot any significant issues that had not already been the subject of life-extending remedial work.

The heroic shaping of steel rebar. Xiangshan Park, Taipei.

Wherever form and function are closely aligned, such as at a temple or important state building the results are often spectacular. One such site is the ornate Longshan Temple, where every carefully crafted detail in the rafters draws the eye in to reveal further and further subtleties, inducing an almost meditative state. Other highlights include the enormous Grand Hotel, Liberty Square, and the imperious National Palace Museum (which itself contains some of the most intricate pieces of artwork ever made).

I was only in the country for less than a week, but my impression is that the implementation of the ‘function first’ philosophy is one of the key drivers behind the enormous disconnect between Taiwan’s GDP and PPP figures². The cost of transport, utilities, and restaurants are all significantly lower than in comparably sized cities like Auckland, Seattle, or Rome (even after adjusting for local average salaries). Taiwanese generally get a lot of bang for their New Taiwanese buck. The value of the goods and services they produce is almost twice as high as one might expect given the cost of their inputs. Put another way, far less money is wasted across the Taiwanese economy on processes that would otherwise be captured in GDP figures (e.g. an inefficient public sector, excessive legal and compliance costs, and other ‘rent-seeking’ activities), but ultimately don’t add value to the terminal good or service being provided.

One of the best examples of the high value for money the Taiwanese receive is the case of the wonderful Taipei Metro. Trains arrive every few minutes throughout most of the day, such that few bother to hurry to catch a departing train. The automated trains arrive on time to the second, flush to the platform and perfectly aligned with double doors separating waiting passengers from the track. Acceleration is rapid and can make you stumble if you’re unprepared. The tradeoff is that you’re whisked across town to within a few hundred metres of most destinations within minutes. The system is easily navigated even for somebody foreign and inexperienced. The stations are immaculately clean, the design is ergonomic, and signage is frequent, consistent, and clearly read at a distance. What must surely be one of the world’s best metro systems is provided at what I assumed was a mostly nominal fee of less than 1 USD/trip. Staggeringly, I later learned the system is so well run and utilised that it actually provides an annual operational profit!³

The net result of this infrastructure and public services (including busses and bike hire) are city streets almost entirely empty of vehicles outside of peak times. Of course, the city’s high density lends itself to high public transport use, and the relative lack of parking favours scooters over larger vehicles. Even so, the traffic in Taipei clearly very low compared with most other cities of comparable size.

Row of scooters parked alongside a typically nearly empty street near the Main Train Station

If I were to criticise the infrastructure, it would be to note that the standard of footpaths in Taipei is generally terrible. Any city that provides such excellent public transit creates a lot of pedestrian traffic as people walk the last mile from their local MRT station, bus stop, or U-Bike station. Unfortunately, older footpaths are mostly an afterthought, if space is available. Even modern paths tend to have an uneven camber, variable texture, and trees or other obstacles blocking people from walking in a straight line. Signalled street crossings lack any sound or haptic feedback. ‘X-crossings’ are rare, and pedestrians will often have to wait for multiple cycles in order to cross even relatively small intersections in several stages. In all, Taipei is more comparable to Bangkok than Amsterdam in terms of walkability; it’s a difficult city to traverse for mobility or vision-impaired people.

Typical footpath quality in Dazhi District, Northern Taipei

Smart Taipei

In ‘At the Dawn of the Smart City Age’, I wrote about some of the opportunities and difficulties inherent in utilising cheap sensors and AI to generate actionable information about a city. Very briefly: the falling costs and enhanced capabilities of sensors and processing software is enabling cities to replace costly analog monitoring devices and/or manual collection with inexpensive cameras⁴ and audio recorders that can be deployed at scale. Image recognition software, or other machine-learning based tools, can be used to generate real-time information with unprecedented scope, scale, and cost. Such capabilities create enormous opportunities for efficiencies in fields like planning and automation but carry downside risks through the loss of privacy.

It’s hard to get a sense of which cities are most advanced in this work. Most of the ‘smart city’ indexes are laughable, usually containing a variety of qualitative assessments of self-reported sustainability goals or twilight zone methodologies that allow the cities with the worst traffic congestion in the world to somehow score highly on mobility and congestion metrics for example. Regardless, as far as progress on taking advantage of the opportunity I’ve described above, and working to mitigate the downsides, Taipei is clearly decades ahead of most.

Taiwanese companies and municipal governments have been deeply involved in this work for such a long time that it’s taken for granted. All of the presentations I attended assumed that conference attendees already understood and were deeply involved in this work. Presentations from managing directors of major tech companies were frequently pitched at the level of promoting a particular piece of software or protocol for addressing arcane problems, and left most of the audience bewildered⁵.

Taipei has sensors everywhere. From what little I was allowed to see, I’d suggest they’re involved in ‘big data’ on a scale similar to that of a major Silicon Valley tech company. Cameras on every street corner count vehicles, pedestrians, bikes; they flag potential maintenance issues, crimes in progress, capture licence plate information, and monitor carparking availability. This data is uploaded to their cloud, then used to inform decisions across different timescales. Applications include dynamically managing traffic lights to improve traffic flow or allow free passage of emergency vehicles, the management of large crowds during public events, or identifying long-term transport and housing capacity constraints.

A limited overview of the sort of systems Taiwan makes use of

What impressed me most about their system was how much effort went into protecting the privacy of individuals. An entire team at City Hall was devoted to anonymising the data from these streams, even going so far as to do their best to ensure identity couldn’t be reverse-engineered. Protecting the integrity of these processes was the primary reason why unfortunately so little of what they were doing was able to be formally shared. I was impressed by the fact that where the CCP devotes considerable resources to the creation of its panopticon, Taiwan instead devotes those resources to preventing it.

These technologies are not only being used by local governments but are being increasingly applied in Taiwan by the medical sector as well. Traditional monitoring sensors are being supplemented with cameras and ML models trained to detect signs of deterioration, assist with diagnosis, and assess patient risks. Patient mortality has subsequently dropped by a staggering 25% in less than a year, while antibiotic costs have fallen by 30%. Emergency patient wait times have halved. Taiwan is also experimenting with ‘smart’ operating theatres and telemedicine, allowing remote and mobile centres, as well as emergency vehicles, live access to central AI models and specialist services.

Challenges

While it does possess a stunningly beautiful landscape, Taiwan is not blessed geopolitically by its geography. A small nation of steep mountains, it is up against the limits of the freshwater supply it needs to support both its agricultural and manufacturing industries. It has few natural resources. All of its significant harbours are on its western coast, a mere 100km from a major geopolitical adversary that believes Taiwan to be a province in rebellion whose independence leaders are tumours that need to be cut out.

Unfortunately, the situation seems to be escalating by the day as Xi looks to shore up support at home, weakened by his ultimately disastrous Covid response, by acting tough. With Japan and the US increasingly strongly signalling their support for Taiwan, it is almost certain that any Chinese amphibious invasion of the island would likely end in complete disaster. However, the US Navy War College’s recent report on the potential of a naval blockade makes for sober reading. Not only would such a move be much more difficult to respond to diplomatically, it seems that a blockade is potentially beyond the capacity of Taiwan and its allies to prevent militarily.

As the Taiwanese themselves recognise, the parallel with Ukraine could not be more clear. Again we hear talk of supposed ‘Western aggression’, such as the absurd claims that receiving a democratically elected leader ‘undermines China’s sovereignty and territorial integrity’, and that the US would bear responsibility for the ‘consequences’. It’s clear to all paying attention that this is nothing more than pretext and propaganda. One would hope that by now Western democratic leaders would have learned that making concessions to autocrats merely emboldens them. Shame on Macron for thinking otherwise.

Needless to say, I can only hope that Taiwan can remain a free and independent nation without a violent conflict that would likely set the entire global economy in turmoil (besides the obvious disruptions to global trade, Taiwan is also a centre of the world economy, as TSMC’s facilities in Taiwan produce virtually all cutting edge computer chips such as those used in modern AI).

Assuming the situation with the Mainland can be safely navigated, Taiwan still faces the serious long-term challenge of its rapidly aging population. Taiwan is at the forefront of the global trend of declining fertility. Today there are a third as many Taiwanese aged 10 as age 40. With just 1.2 children per woman (the second lowest in the world above South Korea) for every 100 young Taiwanese adults today, there will be just 10 great-grandchildren. On current trends, by 2050 the proportion of elderly will have doubled, the working-age population will have declined by 25%, and fewer and fewer young people will be available to fill the gaps. With such enormous demographic challenges, Taiwan will struggle to maintain the human capital required to sustain its high-tech economy and continue to make such outsized contributions to humanity.

‘Taipei Skyline, Synthwave Digital Art’. Created using MidjourneyAI running on A100 GPUs built in Taiwan by TSMC.

Why we need Taiwan

‘The West’, and particularly the English-speaking part of it, has become a supernational culture. Increasingly, we all watch and listen to the same Californian content. We follow the American elections with almost as much interest as our own. Truth is increasingly defined by engineers in Silicon Valley, adjusting content promotion algorithms and moulding tools like GPT to align with their values. In a certain sense, we all live in America now. My point here is not to criticise American culture but rather to note the loss of antifragility this trend has caused. When we look around at each other, we see more of the same old politics, policy, art, and ideas. Who then can we look to for inspiration or course correction?

Once, we measured up with the USSR. Sometimes in fear, sometimes in awe, and eventually with pity. Soviet output may never have received a mass market audience in the West, but their cinema, science, and literature were the product of a totally independent intellectual tradition. Their work inspired our filmmakers, rocket scientists, and authors. You might be surprised to learn that Russia’s Yandex and Sberbank were at the cutting edge of fields like natural language processing, computer vision, and cloud computing. I’m told the UX of Russian online services is decades ahead of our own. Unfortunately, any chance those projects may have had to blossom into something the rest of us could admire and learn from passed with the 2022 invasion of Ukraine, as their best and brightest either fled or faced conscription. One more tragedy of that stupid war.

I want to see a cure for cancer and aging. I want us to be so wealthy and powerful that we can afford to continue to scale our investments in regenerating our natural environment. I dream of humans on Mars, probes visiting planets in other solar systems, and fusion-powered electrical grids finally providing energy too cheap to meter. Unfortunately, at some point in the past decade or so, our vision of the future has grown dim. Despite wealth beyond the imagination of our grandparents, we’ve allowed ourselves to become a culture where the default is ‘building nothing’, rent-seeking, wildly counterproductive regulation, and zero-sum redistribution. Advances in rocketry, biotech, and even AI are increasingly looked upon with fear or distrust. We’ve even managed to convince ourselves that building new housing will somehow exacerbate a housing crisis. When our cultural sphere is all caught in the same conceptual traps, it becomes difficult to recognise and acknowledge the error.

We need peers that can challenge our ways of thinking and demonstrate the value of taking an alternative approach. All the better if they are also allies willing to share their learnings. Taiwan has been at the cutting edge for long enough that they’ve deployed solutions to problems we’re only just starting to grapple with.

The Taiwanese are some of the most wonderful people you’ll ever meet. As a free and open democratic society, they would deserve our support for that alone. If that’s not enough, then we should support them out of pure self-interest, as we’re all better off for having them.

1. Since my last experience at Auckland Airport some years ago, a new system in the departure customs hall places e-passport holders and paper passport holders in the same holding queue. A single customs official slowly clears the queue until you reach the point where you’re able to use the half dozen odd e-gates that are mostly sitting idle. This is after already having waited two hours to check in my baggage through a similarly horrific system of terrible signage, inadequate staffing, and queues that were so long they informally snaked out of the queuing area, into the main concourse, down to the end of the building, then doubled back upon itself with nobody but other passengers to explain to newcomers what was going on. Thinking the chaos may have been partly a lingering consequence of damage during January’s flash floods (which, to be fair, the airport was able to operationally recover from extremely quickly), I was able to ask a staff member if this was the case, only to be told in a friendly and casual manner “oh no, it’s normally like this on weekends”. I can’t fault the staff, but the management has much to answer for.

2. The estimates of Western institutions such as the IMF, World Bank, CIA, CEIC & World Economics Research, vary significantly for reasons I can’t quite get a handle on, but possibly have something to do with the diplomatic status of the country. Averaging these after discarding estimates wildly out of sync with cost-of-living statistics, I’m picking a PPP per capita of somewhere around $65,000 USD, from a nominal GDP of around $36,000 USD in 2022. Taiwan is a major outlier in this regard, with a disparity more commonly seen in developing countries with low labour costs and large informal economies than in advanced market economies. This points to significant value being provided to consumers at very low relative cost, such as in the case of the high-value yet cheaply operated Metro system.

3. I was able to speak at length about the Metro (and a range of other topics) with the Taipei City Finance Commissioner over a few beers one evening. He was understandably proud of the operational profit of the system (Western cities typically deliver between 20-50% operational farebox recovery), but confessed annoyance that it still ran at a loss after interest and depreciation costs (e.g. the costs of replacing the trains at end of life) were properly considered. Of course, considering the direct consumer surplus (we frequently skipped the free shuttle bus to the conference from our hotel in order to save 30min of our time by taking the MRT instead), vastly reduced traffic congestion, and almost negligible carbon & particulate emissions, there’s clearly an absolutely enormous net value to society being provided by the system. However, I think it speaks volumes that despite this, they were still thinking about ways to further improve the efficiency of the system in order to further reduce costs in the hope of eventual breakeven.

4. The price of a camera + edge processor + network device system is now around the same price as the installation of an induction loop sensor that triggers most traditional light phases. The camera thus has the advantage here even without the additional information it generates, as a single camera can cover multiple lanes. The camera can also determine how many cars are queuing in a given lane, and thereby enable a much more precisely timed and efficient light phase compared with loop systems that extend the phase by a set amount of time for each detection, up to a limit. This is the logic mechanism behind the common situation where a single green phase can extend for a long period of time to accommodate a sporadic sequence of cars, while other lanes with large queues sit idle.

5. I ended up developing something of a reputation at the conference for ‘translating’ at these events, providing much-needed context of the core ideas and use cases that the given technical solution was designed to address. Of course, it’s to the Taiwanese presenter’s credit that they assumed that attendees of a smart city conference would understand the core principles of smart cities a little more deeply than the compilers of smart city indexes.