Chapter 3: The Importance of Ideas

As I've worked in the field of AI over the years, one thing has become increasingly clear to me: with AI, so much of the execution can happen quickly and at zero cost. In other words, the time and cost it takes to implement an idea have become negligible, whereas the value is created in understanding the tools and the problems we want to solve.

This is a major shift from how things used to be done. In the past, when executing an idea was a slow and costly process, the idea itself had less value. But with AI, the idea becomes more important than the execution.

Don't get me wrong – execution is still important. But with AI, it's easier and more affordable than ever. So the real value is created in the ideation stage.

To illustrate this point, let me expand on a few stories from my own experience.

A few years ago, when I worked with the team to develop AI to detect fraud in payments, we had become frustrated as we tried several conventional approaches. Every time some fraud network would try to break the system we would do our best to figure it out and write new rules. At first the rules we developed were very basic. If fraud was concentrated on internet protocol addresses from Argentina we would try to see if there was another indication. Did the fraud come from the same website? Were the criminals using cards that had something in common? Did they use a similar device like a Windows PC? Once we identified a pattern we would block transactions that matched that pattern. Unfortunately, many normal customers could be from Argentina, using a Windows PC and with a card issued by the same bank. The goal was to only block the fraudsters and let good customers buy what they wanted. It was hard to do with the limits of human pattern recognition. 

We had a massive amount of data – millions of transactions – but we struggled to make sense of the patterns that indicated fraud.

We saw this as an opportunity to apply AI to the problem. We started by gathering as much data as possible and then training a machine learning model to identify fraudulent transactions.

But it wasn't as simple as it sounds. The data was noisy and inconsistent, and there were a lot of false positives. It took years and massive investments to get good results, but eventually, we were able to detect fraud with a high degree of accuracy.

This experience taught me a valuable lesson: the value isn't in the data itself, but in how you use it. It's in the ideas you come up with to make sense of the data and solve problems.

Now, let me expand on the controlled environment agriculture example from earlier.

As I mentioned earlier, this team uses AI to create systems that can learn what plants need to reach their full potential. The idea is to create an environment where plants can thrive and produce more, without using excessive resources.

To do this, the team is using a combination of sensors, machine learning algorithms, and robotics to monitor the plants and adjust the environment as needed. The system can detect changes in the plants' behavior and make adjustments in real-time, without any human intervention.

Without an AI system we were reliant on humans, and we’re all limited as humans. We can’t easily see the earliest signs of a suffering plant. There are just too many plants for agricultural experts to review. The plants are also changing constantly. Using our human intelligence led to constant stress on the team. What if they caught a change in the health of the plant too late in the process to fix it? What if they missed something completely and the plants died? 

The ability of the AI to see what is happening to the plant and adjust the complex inputs reduces this stress dramatically while ensuring that the plants will reach their full potential. The AI can also go further by finding combinations of ideal inputs at each stage of growth that a human would never have thought of using. 

Again, the idea behind this project is what makes it valuable. The technology itself is impressive, but the real breakthrough is in how we're using it to solve a problem and create value.

In both of these examples, the value wasn't in the technology itself, but in how it was applied to solve a problem. The technology was just a tool to achieve a goal.

So if you're working in the field of AI, my advice to you is this: focus on the ideas. Spend time understanding the problems you want to solve and coming up with creative solutions. Don't get bogged down in the execution – that's the easy part. The hard part is coming up with the ideas that will drive real value and impact.

And if you're not working in the field of AI, don't be intimidated by it. AI is just a tool, like any other technology. It's not a silver bullet, and it won't solve all our problems overnight. But if we use it wisely and creatively, it can help us achieve great things.