Dana Blankenhorn

Open Source in an AI World

The history of generative AI begins with a betrayal. 

OpenAI, meant to give developers a high and rising model for AI development, went closed source and got into bed with Microsoft. Now Microsoft is leading the charge to regulate AI, raising costs for competitors and creating closed source lock-in. The company wants licensing for all AI, bureaucratic delays before anyone or anything can compete with its proprietary offerings.

The current listing for the Google AI page claims it offers an “ecosystem of open-source tools, datasets, APIs and more.” But open source is never mentioned on the page itself. Google, like Microsoft, now thinks AI is too dangerous to be put into the community’s hands.  

If it’s that dangerous, guys, then ban it. Forbid machine learning, forbid progress. If code exists, someone will try to misuse it. That’s part of the law of code. The only answer is for code not to exist.

The Wealth of AI

The battle over who gets what when productivity blossoms has been going on since the Renaissance made learning cheap 500 years ago. It accelerated with the Industrial Revolution, accelerated again with the Computer Age, and again with the Internet. It’s accelerating again.

The wealth of AI is the rocket fuel of Moore’s Law.

The late, great Gordon Moore was an engineer, who saw his theory as a self-limiting phenomenon. In various editions of my book on the subject I have tried to explain how this constant, exponential improvement works outward, from processors, to everything silicon touches.

Like software.

Software is still mostly written by hand. In the 20^th century productivity improved only in line with the tools used to write it. The 21^st century brought Moore’s Law to software. Once software goes online, in a managed cloud, it’s instantly accessible to everyone. All the barriers created over a century to cordon off markets disappear. Some can be rebuilt, but at an enormous cost to efficiency.

Open source is a big part of this new Moore’s Law of Software. The more hands on the code, the faster the code improves. Software communities mean everyone is building, together, on a constantly rising platform. There’s no more reinventing the wheel.

A Minor Announcement

Over the years my blog, and now my Substack, has covered everything from sports to history, from technology to urban planning.

Now, as I enter the last lap of my career, it’s time to focus.

From here forward, my main subject will be the business of technology. Other topics will intrude, they always do in my mind, but I have spent my entire life fascinated by the intersection of these subjects. I’ve learned a lot about them. And that’s what I want to share.

The first point I will make is, don’t get carried away by the AI hype. It’s important, but it’s not a destination. AI isn’t the Web, or the Cloud. It’s software, which can take huge inputs and deliver what appear to be rational outputs. Its key outputs won’t be music, or art, or stories like this one. It will be insights drawn from networks of sensors, on millions of devices, measuring systems as large as cities and hospitals. AI is a productivity tool, like a SaaS database, and its evolution is built on that base.

The AI Silly Season

We’re currently in the “silly season” of hype over “Generative AI,” machine learning programs whose output can be text, code, pictures, audio, or video.

There’s a lot of experimentation and there are a ton of bad takes.

Here’s how it’s going to go.

There are two things to know.

First, any software requires training before it becomes useful. Right now programs like ChatGPT are creating garbage. Interesting garbage, but garbage. As people learn how to get value from these tools, however, there will be huge jumps in productivity. The timing and size will differ from industry to industry, and from organization to organization. But it will happen. People will be able to do more, high-quality work in less time. This won’t reduce the need for people. It will give people more things to do.

Beating Moore’s Law

I spent 20 years, off and on, researching Moore’s Law, its history, and how it worked. 

As I wrote my book, and rewrote it twice more, most industries succumbed to its charms, its promise of productivity. Journalism succumbed. Retailing succumbed. So did distribution and management.

Three industries have not yet succumbed, and I have a chapter in the book devoted to them.

These are health care, the law, and education.

It’s not that there aren’t solutions. Law libraries are obsolete. So, too, college libraries. Health care has finally surrendered to electronic records, lured by sweet, sweet stimulus cash.

But final costs, and prices, have gone up in all these areas.

It’s politics.

States refuse to define what graduates need, and unions insist teachers must give the lesson, not just test and tutor. Courts have used technology to pile more work on lawyers and judges. Time is money. Doctors have used their gatekeeping function to stop standardization of treatment, while drug companies have used regulatory capture to create pricing based on the value of your life.

That 2002 Feeling

The last tech crash began in 2000 and I was in the center of it, a mid-career writer with a specialty called Internet Commerce.

That year saw denial and bargaining.

The next year saw anger and depression.

It was only in 2002 that acceptance came.

We know what followed. Talented programmers worked on open source projects and standards bodies created what became the Cloud. When proprietary advantage became worthless, and there seemed there was nothing to do, people worked together with technical goals in mind rather than just financial ones.

It was a slow build. In 2002 my gross income was zero dollars. In 2003 it was zero dollars. I got interested in what’s now the Machine Internet, which I christened the World of Always On, and in 2004 I finally made a few dollars. The next year I was asked to cover Open Source, and my career took off. For a time.

I’ve mentioned before how, in every photo of a stock market crash the people in it look similar. Similar age, similar backgrounds. They’re different people. Those who watched the 1987 crash were different from those who saw the 2000 crash, different from those who saw 2008, different from those who saw 2022.

Tech and its Workers

One of the biggest mistakes I made in my journalism career lay in assuming tech companies would have the back of their workers, because they depend on human talent rather than muscle. (The picture is from a book on computer coding.)

The tech oligarchs who made their piles in the last decade have become just like other oligarchs. No worse, but certainly no better.

The latest round of layoffs is especially egregious. It’s not how many were let go. They were mostly the pandemic overhang, an adjustment as life returns to “normal.” It’s how they were done, and who was targeted. Firing via e-mail? No warning for families or teams? It sends the message that people don’t matter.

To the oligarchs they don’t. Elon Musk, Marc Andreessen, Peter Thiel – they’re one-way guys. They care about themselves, no one else. They don’t really care about technology, or what it can do. All they care about is how much money it is making for them right now.

There was, once, a difference between Wall Street and Silicon Valley. The wealth creation of the last decade eliminated the difference. Wall Street doesn’t care about the long-term, about the need to keep investing in the new across the chasm of a recession, or the talent that is tech’s raw material. It cares about this quarter, about today’s trade.

What’s funny is that the horrors now being visited on the Valley are 10 times worse on Wall Street. Layer after layer of management has been eliminated, as the industry has gone online. More and more of its wealth is in fewer and fewer hands, and those hands care nothing about the people who made the wealth. Just as they don’t care about the technologists of the Valley.

Failures of the Cloud Czars

The failures of “Big Tech” aren’t what you think.

Each of the five “Cloud Czars” – Apple, Microsoft, Google, Amazon, and Meta – grew complacent over several years and let a strength become a weakness. They each now face problems that will cost tens of billions of dollars to fix. They’re cutting workers in growth areas and denying the reality.

This will spell more trouble in 2023, and opportunities for smaller companies.

I can summarize each Czar’s big problem in one word:

 

Apple – China

Microsoft – Windows

Google – Gmail

Amazon – Alexa

Meta – Facebook

Let's take 'em one by one

Business Needs Consensus

I have long believed historians misread the Progressive Era.

They see it as a victory for liberal values.

I read it as a victory for business. Which isn’t necessarily a different thing.

Fact is, the business of America is business. America has always been led by a dominant industry. Farming, then cotton, then railroads, then infrastructure, manufacturing, oil, and now technology.

Power flows to the highest-value industry and, when that is supplanted, flows away from it. Just as cotton lost to the railroads in the Civil War, I believe railroads lost to infrastructure in the Progressive Era.

What the great inventions of the late 19^th century – electricity, telephony, and the automobile – required for mass adoption was scaled, organized infrastructure. Having two or more electric utilities, two or more phone companies, or (God forbid) two or more private road networks made no sense. What Wall Street saw in the 1890s was that monopolies created efficiency.

But the monopolies would need the approval of government. The price of government approval, in a democracy, would be regulation, and some means of control over the result.

The Word for 2023 is Efficiency

If you can save costs, in business or consumer markets, you can make money.

This is technology’s big opportunity in 2023.

The cloud has already transformed how people interact. It’s now going to transform how machines interact.

Sensors and network software can be applied everywhere. Thanks to falling chip prices this can now be done cheaply. There is labor available to write the software, build the companies, and reap the rewards.

Consider the water system. Why aren’t sensors monitoring the condition of pipes, warning of potential failures, and identifying leaks the moment they occur? Why are water authorities still depending on people finding cracks in the street with water coming out of them on their own?