The Brazilian government requires large firms to hire workers with disabilities. Before writing a paper about this quota for disabled workers, I was exposed to it in several circumstances in my life. I want to take you on a journey on how I transformed these events into a paper

If you found this useful, share and follow for more evidence-based takes on AI & economics. Full paper with more examples + hard data:.🔗 I will soon share more info on it. #FutureOfWork.#AI #econtwitter.

Blazing-fast image creation – using just your voice. Try Grok Imagine.

This is what others have predicted:.@professor_ajay @davidautor @joshgans @avicgoldfarb @DAcemogluMIT @MITshapingwork. AI gives low-skilled workers expert knowledge, reducing inequality within & across occupations, just as @davidautor suggested.

This is not something that happens only in Brazil. US factories are seeing similar transformation. See this paper by @erikbryn and coauthors.

That’s why, in Brazil, AI is increasing jobs for low-skilled machine operators. 📈 After the boom in AI software worldwide, employment rose most in occupations more exposed to AI. Why low-skilled? Because AI gives them what they lacked: knowledge.

⚡ Copel crews know what to fix and where before the line goes down. 🐕‍🦺 Operators in the dog food factory adjust quality in real time. 🤖 Factory workers can ask a chatbot for expert guidance. ➡️ Knowledge turns less-skilled operators into near-experts.

Did you notice the pattern? . AI puts the right knowledge in the right hands at the right time.

Example 3 — Tbit (computer vision QC):.On a dog food line, Tbit’s AI scans pellets in seconds, catching defects and impurities. It instantly feeds results to operators, so they can recalibrate machines, adjust inputs, and boost yield.

Example 2 — Copel (electric utility):.Copel’s AI monitors the power grid from plant to home. When it detects a potential failure, it sends a maintenance team with:.• What’s broken.• How to fix it.• Where to go. The result: fewer outages.

Example 1 - LLM for the shop floor:.Techplus built an industrial copilot. • Trained on machine’s manuals.• Connected to real-time factory data.• Workers can ask it anything:. “How do I recalibrate this?”. “What’s wrong with that motor?”. It gives workers information quickly.

In production, two use cases dominate:. • Organizing production (MES) — scheduling, routing, maintenance.• Quality control — computer vision to spot defects in real time. Let me show you some real cases of it, starting with a Brazilian LLM for the shop floor.

Brazilian national registry covers nearly every AI program ever built in the country. One thing jumps out: AI is used as often in production as in admin work.

When people think of #AI, they picture #ChatGPT. But the quiet revolution is happening on factory floors, where AI is raising employment by making workers more productive. I’ll show you 3 live examples from the Brazilian national software registry. Source:

This is one of the few things that most economists are going to agree on: the government should support research. The reason is simple, and we can use Terence as an example. When Terence Tao writes a new paper, his personal gains are small:.– a line on his CV.– a few citations.

Scared AI will take your job?. Bro, chill. AI has actually increased employment. Check the figure: occupations more exposed to AI saw more job growth, not less. Source:

If I managed to get your interest, I will present this paper tomorrow at the @nberpubs Summer Institute. Thanks to @TrebAllen and Cécile Gaubert for including it in the program.

You can find the paper here:

The main policy takeaway:. To get the full benefits from trade liberalization, governments should also support technology transfers—so local firms keep learning as tariffs fall.

This figure shows the gains from trade liberalization in Brazil—with and without tech transfers. If you ignore the learning f/ tech transfers channel, you’d think trade boosts growth and welfare much more than it actually does.

This matters because it suggests the gains from cutting tariffs may be smaller than we thought. Lower tariffs → fewer tech transfers → fewer chances for local firms to learn.

Here are the main results:. A 100pp tariff increases citations from Brazilian to foreign firms by 58%, and tech transfers received by 18%. These effects reflect reduced exposure to goods from a country—controlling for other trade effects like competition or input access.