What is the significance of AI in corporate strategy? Can AI bring a competitive advantage?
A well-known Silicon Valley investor told me a few days ago: "I would no longer invest in a company that didn't have an explicit AI strategy." But this raises the question of what belongs in an AI strategy? I'll explore that in this and other posts.
At the very beginning, there is the pragmatic question: make or buy? Should you develop your own AI models, or can you buy ready-made models from the market? The answer is both.
In enterprise use, the specific issues are often so individual that transferring a model to another organization shows little success. One example: the Canadian airline Westjet tried to adopt AI models from U.S. airlines. However, it soon turned out that the Canadian network, with many small airports and different climatic influences, has completely different characteristics than the U.S. network. Only with our own models do we now succeed, for example, in predictive capacity planning, which enables shorter processing times for baggage handling. From this, we can deduce that there are three classes of AI models suitable for enterprise use: transferable models, custom models, and large models.
Transferable models are tailored to a very specific use case while being transferable to many other enterprises. To do this, the use case must be generic enough that it is very comparable across these companies. A very simple example is an AI for business intelligence that helps make data in a data warehouse easier for business users to query. Such models increase productivity, much like classic software, and therefore become indispensable.
Customized models are based on a company's specific data. This can involve data that is not available to other companies or that requires special protection. With such models, a company can gain a special market advantage. Obtaining suitable data is one of the biggest hurdles in developing AI models today. Companies that recognize this and base their business strategy on it can build an unbeatable competitive advantage.
The third class of models are so-called Foundation Models. These are very large AI models that have been trained by their developers on a massive amount of data. Large Language Models (LLM) are the best known representatives of this class (e.g. Chat GPT), Diffusion Models - more rarely called Large Image Models - are another representative of this class (e.g. Midjourney), which are based on the generation or transformation of images. Due to the large amounts of data with which these models have been trained, they have - so to speak - "seen it all before". This means that these models are able to learn from the examples they have seen during their training and apply that experience to new data they have not yet seen. This allows these models to achieve valuable results with new data with very little additional training. This is true at least when the data is reasonably comparable to the data used for training. Thus, a Large Language Model trained with huge amounts of text in English and German will produce good results in processing English and German text. However, it is almost worthless when processing Chinese or Arabic texts. Foundation Models are thus suitable to be used in very different contexts with little additional adaptation. They have a very large knowledge base and receive the specific adaptation to the actual task within the user company through so-called finetuning or prompt engineering. We'll take a look at exactly how that works in a separate post.
So what does this mean for our strategy? A good strategy takes advantage of opportunities in all three classes of models and applies each class where it provides value. In practice, every company should buy transferable models, develop its own models, and use Foundation Models. But there is no USP to be gained by doing so. So this behavior can only be the basis. A strategic competitive advantage can only be achieved by creating or adapting your own AI models based on your own proprietary data. And for this, the handling of data must be completely rethought. I would like to give two examples of this:
OpenAI has created a really interesting product with ChatGPT. The Large Language Model behind it is market leading and the strategy to implement a combination of open API, plugins and integration in Microsoft Azure is smart. But APIs and integration with a cloud provider are also offered by the competition and plugins will easily copy the competition as well. So what happens when the competition's next model has a better architecture? It is a competition of the best minds. OpenAI's strategy is therefore now far more sophisticated: ChatGPT has an outstanding user interface and is accessible directly and free of charge to completely normal users. Thus, OpenAI collects a huge amount of training data for the upcoming generation of AIs at low cost. According to the current state of the art, this so-called Reinforcement Learning with Human Feedback (RLHF) is the best method to generate a very good AI model. AI researchers also refer to this as "the human in the loop." Through interactions with human users, the model learns and improves noticeably. And OpenAI has secured a unique dataset with this tool, free to users, that maps the feedback and ideas of more than 100 million people. No competitor, not even Google and Facebook, can currently keep up.
Tesla is pursuing a similar consistent strategy. Tesla is the market leader with its electric vehicles and, with the Model Y, leads the passenger car registration statistics of all drive types in Europe. At the same time, Tesla has by far the highest margin of all mass manufacturers and the only global charging network. All signs of an extremely good strategy. But the main attraction of Tesla's strategy lies elsewhere. Tesla has been equipping all of its vehicles with the full hardware for autonomous driving for years, and regularly updates the software on all of them through over-the-air software updates. In every vehicle in the fleet - now more than four million vehicles in major vehicle markets worldwide - Autopilot is constantly running in the background, even when the driver is not actively engaging it at all. Autopilot constantly compares the driver's own (planned) driving behavior with the driver's actual driving behavior. As soon as the software detects important deviations, relevant data is saved and transmitted to Tesla at night. Tesla thus has feedback on its own software from 100 billion actual kilometers driven by day and night and in every weather situation. No other manufacturer has even remotely comparable data. It is estimated that Waymo (Google) has less than 1% of this amount of data and other manufacturers are much further behind. In particular, no fleet of test vehicles, no matter how large, has data on real accident situations, because test drivers must avoid accidents at all costs, if only for regulatory reasons. In practice, however, critical situations and accidents do occur, and Tesla receives detailed data on each.
What do companies do that don't have their own data? There are basically three alternatives: they buy data from third parties or enter into partnerships for it, they build test environments where the data is collected in a controlled environment, or they build a simulation where the data is artificially generated. Often, a combination of these is used. However, simulation in particular places considerable demands on their quality. You can observe this in the example of autonomous driving: today's simulators used by companies like Waymo or Tesla exceed the quality of video games by orders of magnitude - and that for a "player" who pays nothing at all for the game experience.
So focus your attention on exactly one question: how can you gain a data advantage?
What questions do you have about artificial intelligence? Are you already using AI in your company or are you still looking for suitable strategies? I look forward to the dialog.