For the sake of argument, let’s say there really is an AI race. Countries and companies are perceived to be in this race to win and much fear revolves around falling behind in the AI race. Let’s dive deeper into this race analogy.
To win a race you need an engine. For AI the engine is the Large Language Model(LLM). It makes sense. LLMs are the product of engineering teams designing them, and millions go into the hardware that supports them. Hence we assume that the highest level of computer chips and the biggest data centers are needed to build the best engine. We also need data to fuel this engine, the more the better.
Consequently, talk about AI revolves around winning the race by having the best engineers, building the best LLMs, using the most powerful chips, building the biggest data centers and controlling the most data. But the biggest and most powerful engine does not automatically win the race for you. Let us compare it to a real race, the most iconic of all, Formula 1. In Formula 1 a good engine is necessary but as any casual observer will have noticed, it definitely does not win the race for you. You cannot just put the engine out there on the track no matter how impressive it is. It will go nowhere. At the very least you need to fit it in a car. The car has to be designed to be as light as possible, and have certain aerodynamic qualities. It needs to have wheels that need to have optimal traction and be efficient at braking. It is also necessary to have a driver who can drive the car and make the right choices in split seconds. For all of this to come together you also need teams of people, to repair and build the car, to change the tires, to analyse data to improve performance. You need an organisation with a certain structure and cooperation. A leadership that sets the directions and makes race day decisions.
When you start to think of it like that, the engine is actually just a small part of the equation. A three times more efficient LLM will not make your organisation three times more efficient. To “win the race” a lot more than focus on the engine is needed. In fact you could easily win the race with an inferior engine if you make sure to optimise all other aspects. The equivalent would be to make sure data is organised, understood and of high quality, the implementation is anchored in value creating business processes, the risks are managed and balanced with benefits, there is a process to turn all the good pilot projects into production, there is an AI operating model, to set the right team with proper leadership etc.. Most of these aspects are still found lacking in organisations across the world and no amount of data center, engineering or data volume will substitute that need.
In fact, we are at a point where it is more important to be able to fit the AI engine into a functioning vehicle to get anywhere, than it is to have the most powerful engine. All engines today are sufficiently powerful. A single minded focus on the LLM will therefore not get us anywhere.
To win the AI race, organisations need to think of AI as a Formula 1 team, where many different aspects must come together in a holistic fashion. Of course a powerful engine will help but it will not win you the race.
Photo by Brad Barmore on Unsplash
English 
Danish