AI replaces expensive jobs first

Expensive Jobs?

The obvious two:

• Doctors 🧑⚕
• Lawyers 🧑⚖

Why are expensive jobs expensive 💰?

Labor gets expensive when demand is high and supply can't grow. Health and legal trouble show up all the time (someone gets hurt, someone gets sued), and it's hard to find a person who never needs either in their whole life. Demand never goes away.

Supply barely moves. Why?

• It grows slowly, because:
  • Good doctors and lawyers take years and years of training.
  • The "prestigious" schools that train them only turn out dozens of people a year.
• And we can't just pump out more, because:
  • Not many people can study for ten years in the first place.
  • Forcing the numbers up causes a backlash (read: underqualified professionals).
• The few we do get work slowly.
  • How many patients can one doctor see in a day?
    • People already complain about factory-style clinics that move patients along like a conveyor belt.
  • How long does one lawsuit take to resolve?

So supply always falls behind demand.

Then why can AI replace expensive jobs?

Two reasons. Economics, and performance.

Economics

Making AI is expensive, because:

1. Good AI needs a really clean dataset.
2. A clean dataset needs a lot of data with clear patterns.

And it's hard in any field. Exaggerating a little, a cleaning AI is about as hard to build as a medical AI.

• To build a perfect cleaning AI...
  • To judge how dirty a room is, you need
    • millions of room photos labeled with how dirty they are.
    • millions of examples of each kind of mess and how to deal with it.
      • Water spill → Dishcloth
      • Garbage → Trash can
      • Dust → Vacuum
  • Then each fix has to be trained until it works.
    • Millions of examples of wiping with a dishcloth
    • Millions of examples of picking up and throwing out garbage
    • Millions of examples of using a vacuum well

So even a cleaning AI costs a fortune. Which means if it's going to be hard to build anyway, you want a model that pays off big and adapts to a lot of cases. That's why it makes sense to train AI for expensive labor first. That's where the return on investment is.

Performance

AI also never forgets, and it can copy itself. Imagine:

• A doctor who never forgets a thing from med school. A lawyer who remembers every case perfectly.
• The best doctors and lawyers cloned into thousands of AIs, each taking clients at the same time.
• Anything newly discovered, shared with all of them instantly.
• A doctor who remembers every detail about you and catches problems before they happen.
• Your family doctor, available whenever, wherever.

Industry Resistance

SEOUL (Reuters) - Korea's parliament on late Friday passed a controversial bill to limit ride-hailing service Tada, dealing a blow to a company that has been a smash hit since its launch in late 2018 but faced a backlash from taxi drivers angry over new mobility services. - Korea passes bill limiting Softbank-backed ride-hailing service Tada

The recent TADA war was a classic alliance-versus-megacorp fight. The taxi drivers won in the end, but it was a hollow win. The real winner was another conglomerate, Kakao Mobility, which ended up taking over the market.

Doctors and lawyers will push back hard too, and they hold immense social power. Easy to picture how that goes.

Scenarios

• Medical AI kills a patient! Can we really bet our lives on a half-baked machine?
  • Whatever the context, that kind of fear can turn people into tech Luddites.
• Lawyer AI caught discriminating? Can we let a biased machine run our country?
  • Precedents can look biased depending on how you cut the statistics. If someone cuts them in bad faith and frames the AI as biased toward some group, it can snowball into a movement against AI as a whole, whatever the context.

Potential Strategy

Start with a platform that helps doctors and lawyers.

• DOCTORS: Start with a medical CRM. When a patient comes in, their information is pulled together before the visit starts. During the visit, what the patient says is parsed automatically and the AI pulls out keywords. The medical staff checks them. Similar cases and suggested treatments and prescriptions show up on one side of the screen. The doctor picks the right one and goes ahead, or adds a new treatment. All of this is saved on the server and used to train on later.
• LAWYERS: Start with a case analyzer. Begin with small local cases (e.g., traffic tickets), like DoNotPay - The World's First Robot Lawyer. As more and more cases go into the database, lawyers can search for similar ones, kind of like "Google Search." Say a fraud case comes in. The lawyer types in the details, dozens of past precedents come up, and the AI breaks down how they're alike and how they're different.
• Basically GitHub Copilot, but for medicine and law.

Hand out sneakers. Tools doctors and lawyers find genuinely useful and can't work without. Make them your main customers and your data pipeline. In a market as entrenched as medicine or law, never open with an all-out war. Build a friendly relationship first, defang them, and only then go to war.

Articles

AI will replace middle management before robots replace hourly workers

• robots that must be built by robots to perform a specific task once done by a person, or a program capable of making the best possible business decision given historical and currently available data
• If you said the latter, you're correct. The "robot revolution" is not coming first; the AI decision-making paradigm shift will, and it is going to save corporations billions of dollars in salary costs per year as middle management disappears
• Legacy businesses will have to rely on retail and hourly support staff to be able to reduce management headcount as a means of freeing up money for implementing automation
• AI managers are the missing link between the when and how of implementing retail-level automation. Reducing middle management, freeing up cash flow, incentivizing hourly workers, and implementing automation will all depend on the fecundity and vision of senior leadership