Our definition of artificial intelligence
Artificial intelligence is the use of digital technologies to extract actionable knowledge from a dataset that humans can barely make sense of.
At least, that’s how we see it. Because for us, AI is not primarily about robots but about data. And not just big, cloud-based data either. In a world where AI entities such as self-driving cars or industrial robots need to make snap decisions, the local, energy-efficient processing of small data is at least as important.
If you harness that power of data, you open the door to an abundance of AI applications that add value to just about any process.
4 stages of complexity
Some AI solutions are hardly more than automation projects. Others come close to mimicking the human mind. Generally, there are four levels of increasingly complex AI uses.
Stage 1: helping to make complex decisions
In its humblest form, artificial intelligence limits itself to supporting the human decision-making process. For instance: helping doctors with their diagnoses by analyzing a large dataset such as the human genome.
Stage 2: taking autonomous decisions at the edge of the IoT
AI applications such as genomic analysis require a massive amount of computing power. And are therefore usually handled in the cloud. But going back and forth to remote servers is not an option for AI systems that make on-the-spot decisions – such as autonomous cars or factory robots.
The answer lies in IoT sensors equipped with optimized hardware and algorithms. They pair computing performance with energy efficiency, and work together to weave an intelligent web throughout our environment.
Imec is leading the research on edge AI, focusing on three proof of concepts.
Stage 3: interacting autonomously with other decision-making entities
Once AI applications start to appear in the world around us, they’re bound to meet. They need to ‘talk’ to each other and work together – without the help of a central entity.
A typical example of such a multi-agent system is a fleet of autonomous vehicles that continuously pass on information to each other and to the road infrastructure.
Stage 4: intuitively interacting with humans
Computers talking with computers is a challenge. But still relatively simple compared to computers fluently communicating and collaborating with people. Not only because we humans are notoriously unpredictable, but also because computers need to be able to detect and interpret contextual information like social behavior and cultural background.
Since Alan Turing established his famous test in 1950, this level of AI is the one that speaks most to our imagination. And although a lot of progress has been made since then, there’s still a long way to go.
Imec is currently working on three proof of concepts for human-machine interaction.
What’s your definition of AI?
Are you looking for quick ways to optimize your production process? Or do you want to create a care robot that understands the needs of patients? It doesn’t matter which AI application you envision – you always start at the data level. And imec is ready to help you out with algorithm development, prototype creation and other forms of support.