The Journey from Automation to Autonomy: A Deep Dive into Digitalization, AI, and Beyond

As I sat down at a recent round table discussion about the role of AI in procurement, one question that piqued my interest was the difference between digitalization and artificial intelligence.

To clarify, digitalization involves moving data and processes into the cloud, speeding up business operations, automating tasks that algorithms can handle, improving global collaboration among stakeholders, and enabling insights through structured data.

AI, on the other hand, aims to imitate cognitive capabilities like perception, reasoning, prediction, and classification—tasks that are difficult for static algorithms to perform. However, AI relies on data made available by digitalization, and in turn, it makes software applications smarter and easier to use.

The current trend is to incorporate AI everywhere to automate tasks, whether they are repetitive and low-value for employees or high-value tasks that require data analysis and speed beyond human capabilities.

Let's take a look at some of the key trends in the world of AI, particularly those that leverage the new multi-modal generative AI models:

1. Making AI More Human-Like The aim here is to drive adoption and improve how we interact with AI systems. This involves enabling AI to converse with us using voice and natural language, as well as to view and understand what we produce, such as documents or images. OpenAI's GPT4o, Google's Project Astra, and Microsoft's Copilot + PC are all great examples of this trend.

2. Building Smarter Foundational Models The objective here is to create smarter foundational models capable of solving complex problems that require multiple steps and tools. This necessitates substantial investments in research and training capacity, which only a select few companies globally can bear. This situation raises concerns about sovereignty, cultural disparities, and the overall societal impact.

3. Deploying Efficient AI Models Everywhere The aim is to deploy millions, if not billions, of highly efficient AI models that can operate in the cloud and on local computers (edge), executing a vast array of specialized actions such as reading and generating text, analyzing images, producing and executing code, making API calls, or searching data.

4. Combining Perception, Reasoning, and Actions Lastly, the goal is to create AI systems that combine perception, reasoning, and actions to perform tasks requiring interaction with the physical world. This is perhaps the most daunting aspect, as it suggests that no domain is safe from potential machine replacement.

Now, the million-dollar question: Will we stay in control, or will AI make decisions without supervision?

In my opinion, we'll stay in control, but not in the way most people think. Today, in the enterprise world, AI is primarily used as a tool, with users controlling and making decisions based on what the AI produces. However, as AI becomes smarter, we're automating full processes, such as the invoicing process, and letting AI make most of the decisions.

For instance, consider the autopilot feature in a Tesla vehicle. Instead of following the instructions of Google Maps or Waze, you simply enter the destination, and the car makes all the decisions to get from point A to point B. While you're currently required to keep your hands on the wheel due to regulations and the system not being mature enough, it's not hard to imagine a future where we're content to let the car do its job.

Another example is in military equipment, particularly drones and missiles. This is a more contentious area, but allowing drones or missiles to autonomously execute a pre-determined mission is likely inevitable.

In conclusion, as AI becomes capable of automating more complex tasks involving multi-step perception, reasoning, decisions, and actions, it will become increasingly autonomous. 

The role of humans in the loop will continually diminish, especially when it slows down the process and degrades the overall performance. This, of course, raises a plethora of questions about responsibility and accountability.