세사르 A. 이달고
@cesifoti · 남아메리카
이스라엘국은 남아메리카 섹션의 칠레 관련 연구·분야별 전문지식 보도를 위해 세사르 A. 이달고를 팔로우합니다.
What happens when AI suddenly makes your obscure capability valuable? A few years ago, I invested in a small Swiss ed-tech company called Ha Hu. Ha Hu was created during Covid by an entrepreneurial mom, Nigist Goytom, who was frustrated by the lack of digital tools she could use to teach her children Amharic. Ha Hu eventually expanded into Tigrinya and other underserved languages. But still struggled to grow. Then, when generative AI arrived, Nigist and her team realized that Ha Hu had a capability with a different value: collecting text from languages with little AI training data. So they pivoted to Bee Educate. https://t.co/pV8WGAtM3x, a startup focusing on rare language preservation (or on preserving languages that are rare for AIs). Today Bee Educate is working on a project to collect and digitize Makonde, one of the languages spoken in Mozambique. This is a good example of how a rare capability can find new uses as technology changes. What started as an ed-tech company focused on basic literacy for underserved languages has become an effort to ensure that those languages are not left behind in the era of LLMs.
Many replies to this tweet treat rules as a rounding error. But that tendency to dismiss each criticism as an exaggeration misses the point. Bureaucracy is like a dumbbell. Each 1 kg disc is easy to lift. Put 100 of them together and your arms start feeling like spaghetti. Unfortunately, this is a surprisingly difficult conversation to have in a bar here in Europe. If you question labor regulations or consumer protection you’ll be treated as someone who wants companies to abuse workers and customers. The bureaucracy draws plenty of support from these reactions. In part it is understandable. We all know there are bad employers, fraud, pollution, etc. But there are also powerful incentives that align companies with their employees and customers. Think of restaurants. They are in a hyper-competitive sector and must live on repeat customers. That requires building a reputation, an incentive that goes a long way towards aligning owners with the needs of their customers and workers. But many people in Europe have a strange way to think about business incentives. Ask someone if they care about their own customers at work and they will say: of course. Ask them if other companies care and they won’t trust them a bit. Somehow, people fail to extrapolate their own behavior to others. They think they are alone when it comes to responding to reputation, repeat business, etc. In Europe we want innovation and entrepreneurship, but our system is deeply suspicious of companies. It is a system in which companies are guilty by default and must certify their innocence through bureaucracy. The idea of being born guilty matters for fragile and cash-strapped startups in sectors where Europeans are particularly distrustful (anything that involves computers or data). Many Europeans clearly know how to build good companies and care for their customers. Europe has talent. But we are often ready to make the life of the 99% that behaves well impossible, because of our fear of the 1% that does not. So don’t be surprised if stories of entrepreneurial success by Europeans often take place outside of Europe. Many of the ones who have tried here probably didn’t grow large enough to become a story. Instead of debating whether each individual regulation is reasonable, we should spend more time asking: How much weight can we keep adding to the dumbbell? Soon, we will not be able to flex anymore. 💪
We do not claim these results to be causal. But the findings are consistent with a hypothesis that sheds light on the black box of agglomeration: the idea that agglomeration operates through workplace mixing as a mechanism that enables learning and promotes better wages.
This explains much of the north–south divide. Cities in the south of Brazil specialize in more complex industries that mix the skilled and the less skilled at the workplace. Northern cities cluster skilled workers in the public sector, generating much less integration.
In fact, workplace mixing is particularly strong. For initially poor workers, models with workplace mixing explain about 77% of the variation in initial city wage premiums and 83% of the variation in medium-term premiums. https://t.co/aSUhnesB7c
The point is that the structure of a city, and its ability to mix people, explain wage premiums much better than its size. https://t.co/CptK33uk9O
As expected, larger cities pay better, especially in the south. But the effect of size becomes tiny or even changes sign once we add economic complexity (ECI) or social mixing to the model.
Our new paper with Radu Barza, Ed Glaeser & Martina Viarengo uses 900M+ employer–employee records to explore this question. Bigger cities pay an urban wage premium. But city size is a black box that we must illuminate.
In the cities of northern Brazil, more than 50% of the workers who start poor* are still poor 7 years later. But in southern Brazil, only a third are. Why? 👇🧵 *<1.5x minimum wage https://t.co/GcD5U7aw4N
This morning my wife & I watched the live stream of the humanoid robot games. Most of the live stream is quite boring, with a camera wandering around the arena looking for action. But while many demos feel unimpressive, what is interesting is not the level, but the change. Bad robots can still represent an extraordinary trajectory. The trajectory seems impressive in scale and skill. Watching the opening ceremony and seeing the robots march in unison sends an eerie military message. The images are ominous. You don’t want to mess with a country that can ship endless waves of mechanized soldiers. In the 2010s, demos used to involve a robot or two. The scale of the parade alone is a meaningful uptick in a key derivative. But some demos are also impressive in skill. The robots that played tennis or ping pong are reacting in fractions of a second while adjusting their balance and position while targeting a ball. This seems simple to us, but two decades ago, having a humanoid walk up or down a few steps was considered a good demo. So the slope in skill is there as well. So how should we interpret this? On the one hand, as a scholar, I totally respect the effort. It is a competition filled with startups, grad students, and lab directors who have poured their heart and soul into making these robots. At the end of the day, the event is like an academic conference serving also as a public display of how much the field of robotics has advanced. It is a testament to a growing industry. On the other hand, I also feel the clear military implications and undertones. It is mostly a local event. There are only a few teams not from China, which feels lopsided for a technology with strong geopolitical implications at a time when global cooperation is weak. The key question to focus on is where these humanoids be in 5 or 10 years. Robotics has been around the corner since before I was born. But this time it feels like the technology and the market are coming together within an economy that is vast & booming. Remember that when it comes to scale, China is not a country, but a “planet.” Its domestic market alone might be enough to scale a strategic sector. If the trajectory continues, maybe China will become the first “planet” on Earth to mass-produce artificial people.