This blog post summarizes the most important takeaways. Prefer to hear the whole story? Watch the webinar.
Rules of thumb
- If a person can take a decision within a second, so can AI.
- Always start from a challenge that is of current relevance within your organization. Reflect on how technology could help with this challenge, how it could create value. Never start from the technology itself.
- Don't let yourself be put off by the 'black box' aspect that is all part and parcel of AI. We are happy to rely on Excel to perform calculations correctly, without necessarily knowing the complex formulas being used.
Transfer learning is one of the first major trends seen in AI today. You start with a trained model and personalize it as required. Because you are not starting from scratch, much less data is needed to feed into the model, and the algorithm can get up and running more quickly. Explainable AI represents a second trend. Here, the computer shows where or why it recognizes something so the human operator can see that the model is reasoning correctly. An example that demonstrates this is using AI to find dogs in photographs: in this case, explainable AI will draw a box around any dog images found. As a result, you can see that the model is in the process of training itself, and you get to 'look inside the black box'. This is important because it means you know what basis the model is using to take decisions. Serverless scale is another interesting trend. Here the cloud is used, which means there are far fewer limits on the process. You can focus wholly on the problem, because there is no longer any need to spend time on configuring the (on-premise) IT background.
The common thread among these trends is the ability to take tedious, time-consuming tasks off your hands. Technology is becoming much more accessible, opening the way to its wider adoption.
Fact or fiction: AI will take over all human jobs
The technology will replace jobs characterized by a large number of repetitive tasks. It comes down to seeing this as an opportunity; every industrial evolution has resulted in jobs and/or tasks changing. It frees up more time for work that creates greater added value. AI will also generate new jobs, not least in the field of data management.
Three principles to get started right away
1. Start small and scale up quickly
Don't feel it's beyond you. Start small with a limited data project with a specific business purpose. Learn as much as possible from this and bring these lessons to bear on a bigger project. Never lose sight of the overall business objectives and the big picture.
2. There's no need to start from scratch to get to your goal
A proven data architecture means you don't need to go through learning cycles for architecture design. We use a flexible blueprint to build a data architecture that meets your needs.
3. Focus on the platform, the people and the processes
To go from raw data to an optimized decision-making process or leveraging this to bring about new activities and services means change has to happen in your organization. That's why we don't just look at developing the technology – we also consider the processes and the people who are needed to develop a data-driven culture.
Fact or fiction: A self-driving car must know our values and standards in order to be able to take the right decisions (hypothetical situation: driving into a tree or swerving onto a busy sidewalk)
Probability theory can provide a solution to this dilemma. Firstly, it is (fortunately) very rare to be faced with such a choice when driving. Secondly, you need to have an idea of the percentage probability that this situation would occur if only autonomous cars were on the road. By making both as unlikely as possible, a situation as described above can be avoided and the concept of ethics in relation to self-driving cars becomes superfluous.