As AI becomes more capable, goals matter more

In mathematics, the criteria for a solution are often clear. In society, we need to agree on what a good outcome looks like. A school may want to use AI to grade assignments more quickly. It may also want to explore how more students can understand concepts they previously struggled with. Both are reasonable objectives. But they lead to different ways of developing the technology, organising the work and measuring success. 

From efficiency to ambition 

The difference is greater than it may first appear. In the first case, we use AI to do what we already do more efficiently. In the second, we use this new capability to raise our level of ambition and attempt things that were previously too difficult, too expensive or impossible to achieve at scale. 

This places greater demands on how we define our goals. AI is good at optimising for things that can be described and measured. But what is easy to measure is not always what we actually care about. More assignments graded is a metric. Better learning is a goal. Shorter processing times are a metric. Better service is a goal. 

The goal therefore needs to be clear from the moment we start building. Otherwise, what is measurable and readily available can easily end up determining the direction. 

When problem-solving becomes cheaper 

As problem-solving capabilities become cheaper and more widely available, scarcity begins to shift. It becomes less important who can build yet another system, and more important who can formulate a problem worth solving – and recognise a good outcome when they see one. 

This is where I would like to see a more ambitious conversation about AI. What do we want to achieve in Sweden over the coming years that has so far been too difficult, too expensive or too labour-intensive? What knowledge do we want to make accessible? What opportunities do we want to extend to more people? 

Simply encouraging wider AI adoption is not enough. An organisation can become highly proficient at using AI without becoming any better at deciding what it should use AI to achieve. That capability needs to develop alongside the technology itself. 

The more we can do, the more our goals matter 

Research and industry can demonstrate what is becoming technically possible. Teachers, healthcare professionals, entrepreneurs and citizens are needed to articulate what is worth doing. Policymakers need to provide direction and create the right conditions. 

My optimism about technology grows as these tools become more capable. So does my conviction that we need to devote more attention to our goals. The better we become at solving problems, the more it matters which problems we choose. 

Stay up to date 

Subscribe to the AI at RISE newsletter and follow AI at RISE on LinkedIn for more perspectives, news and insights on applied AI.