The Business of Artificial Intelligence: Part I

Originally written on March 4, 2021 ad published via Hey

Artificial Intelligence has been around since approximately 1956, when,?John McCarthy?coined the term during a workshop at Dartmouth College. I’m not a big fan of it today. I think it should be called?Automated Intelligence… at least for now.

Today, compute power and data are more available and more cost effective than in the 1950’s so the ability to scale AI is much faster and much more economical. It is such a buzzword today that it is in every news stream you can probably think of on a daily basis. Everyone is talking about ideas, technology, processes, methods, and data around AI. It seems, in some ways, that it is still very much in the research phase.

There are a lot of companies talking about using AI, and they’re likely working with the only people they think of when it comes to the buzzword, Google, Microsoft, IBM, and so on. They think up use cases and use terms like?pilot project?or?proof-of-concept?in their bidding for business and attention. Everyone speaking about it on a grand stage is, of course, an expert in it and has the best solution(s). However, if you listen carefully they’re using words like,?“what we’re working on is…”, or?“what we can do is…”, and you never hear anyone really say, “what we’ve done”, or “what it does”.

Hear the difference? One is very much a working model and one is very much expressing that it already works and produces results. When it comes to building a business model around AI, this is a BIG difference.

Why?

Well because companies and their respective divisions have budgets. Budgets are lock-tight and when used for a project there is an expected return; a “what do we get for the money we spent”… and rightly so. But when it comes to AI, we know that AI models require data. Today, data is like currency. So there’s often a tendency that if data is used for an AI model, then the person with the data now owns the model or some part of it. While I can certainly see how people come to this conclusion I think this is where a deeper discussion into?how AI actually works?is needed. And, furthermore, why I prefer to call it?automated intelligence.

To explain this I’m going to use the graphic above and the analogy of car racing. Know this ahead of time… I have zero interest in racing. I don’t watch it. I’m not a fan of it. I know very little about it. But cars and technology have always had these similarities and little idiosyncrasies that compare well in a story like this one. So here goes.

The Engine

The engine in the car above represents the AI and the models. The engine was built by Company A. It works and it works well. At?my company?we call this our CORE, or?Cloud Optimized Recognition Engine?because we work primarily in vision computing.

The Driver and Crew

See all the people around the car including the driver? Those represent developers and data scientists because it takes a team of people to build proper AI that delivers accurate results, continuously train it, and maintain the infrastructure supporting it.

The Rest of The Car

The rest of the car, frame, body, tires, steering wheel, etc is the supporting hardware. This is unique to this situation, and different to other cars, but still just hardware nonetheless. There are a lot of options around this.

The Fuel

The fuel in the engine represents the data. The engine/AI model does not run without fuel/data. It’s what makes it go.

Now, most companies bring the data (fuel) and then once they prove it goes around the track want to then own the engine(AI). Here’s why this does not work.

The fuel is cheaper than the engine itself, it’s easier to find than the engine itself, (especially if you have a winning car), and others have similar fuel. The engine, it’s more expensive, took much longer to build, and there’s not many of them just like it, (if any), out there. Therefore, you’re free to to build your own engine, or have someone build one for you at any time but if you want this car, with this engine, there needs to be a recognition that someone else spent the time, energy, and capital to build this engine for this car.

“So if I bring the fuel and use your car what do I get in return?”. Well, you get a couple things:

1. You now know that with your fuel and this engine you can win a race.
2. You know it works, how well it works, and you can either then decide to then build your own engine, or continue using this engine.
3. You know the quality of your data should you decide to build your own engine.
4. You have a strategic use case and model to work with in the marketplace

If I bring the engine to you and I provide / pay for the fuel, that’s called a?demo.

If you want to test the engine one time with your fuel, that’s called a?pilotor?proof-of-concept.

If you then decide you want to use this engine the entire season that’s called a?contract.

If you decide you don’t want anyone else to use this engine, (or others Company A may have built like it), that’s called?exclusivity.

If you want to buy the engine so on one else can have it, that’s called an?acquisition.

Stay with me… we’re going deeper.

A good engine (AI) with a good driver (lead data scientist)and good pit-crew (developers and data scientists)and the rest of the car have likely run several tracks, (AI models), during trials, testing, development, and demonstrations — so they know these tracks (models). But you want them to run on a different track where your race is being held.

This is a new model.

This means you’re going to need to bring some fuel and run the engine around this track a few times so the driver, and crew, can understand it and master this track before they run a race if you want to win. Now, if you don’t want to pay for the fuel for this, then you are back to building your own engine or finding someone else to build if for you. However, you might want to be careful about the time this is going to take. You may not be ready to race this season or next. Of course, during that time someone else might come along and provide them the fuel to run that track and master it. Or, they may find/buy the fuel themselves to learn and master that track while you’re off building your own engine.

After all, you don’t want to bring someone to a Formula One track that’s only been in NASCAR races and throw them right into the race. As soon as they have to turn right they’ll crash into a wall. So, just as you would expect with a race team, this doesn’t happen overnight. But the team that takes the time to perfect the little things and learn the track will win the race; every time.

A note to AI, (engine), builders.

You’ve spent a lot of time, energy, and capital to build your race engine and that should be acknowledged. At times, you spent your own capital on fuel so you could prove what your engine can do — to show it’s a real contender. You know that fuel costs go up and it can add-up having to continuously buy fuel.

So do yourself a favor and stop thinking your engine can run without good quality fuel and stop wasting the fuel of those who are paying for it!

I’m hearing, on a regular basis, about some company who slapped together a go-cart and promised they can win the Daytona 500. You’re lying to yourself, you’re making the rest of us look bad, and ultimately you’re wasting someone’s, (the clients), fuel. Also, this crap about borrowing fuel from one person to run another race, (using ImageNet to prove your model(s) but ultimately needing your own data), is a load of shit too. This is dirt track stuff and the races being run out there where good fuel is being used is pro series. If you’re going to show up with you engine, driver, and team, you better act like a professional and know the race you can win because you never know when you’re going to be called the starting line to run a race… and when you are… you better be ready to win… because while it’s your engine, it’s someone else’s fuel. That’s why it’s called a race?team.

When we bring together client data and tech companies AI models, we can add tremendous and powerful value to an organization. We’ve all heard the old proverb, “If you want to go fast, go alone. But if you want to go far, go together”.

Well, how many races do you know are done just once or in one lap by one person?