Although there are already plenty of examples of success stories of Artificial Intelligence, there still are a lot of challenges that this new technology is facing or, in a near future, will have to face. For this blog post in our series 'A future with AI' we talked to experts with different perspectives to get a clear view of what these challenges are and will be. We discovered that the challenges could be categorized into two groups. Some of these challenges have to do with the optimization of the technology itself, while others relate to the actual implementation. Let’s dive into what bumps we can expect along the future road of Artificial Intelligence.
A biased character
We all know that humans are biased. We have programmed preconceptions which we are not always aware of, but they do influence our daily behaviour. Many say that a benefit of machines could be unbiased decision-making. While this may be the case when it comes to simple hardcoded computers, this unfortunately does not apply for Artificial Intelligence. Artificial Intelligence learns from the structured information that we humans provide it with. If this information is biased due to the human aspect, algorithms with the same human bias will be the outcome. This danger is something which you could call ‘the feedback loop’. Artificial Intelligence constantly updates it algorithms to be more accurate by looking at feedback. If new information confirms its former - in this case biased - conclusions, the algorithms will therefore become even more biased. This process of ‘the feedback loop’ could be compared to a downwards spiral, which is never ending.
When it comes to Deep Learning, a component of Artificial Intelligence where learning is done from big streams of structured data, we can see yet another big challenge. The algorithms that Deep Learning uses are written in a machine language which we as humans can not understand. For us, that is a tricky position to be in because there are cases in which algorithms take elements in account in decision-making which do not matter at all. For instance, when it comes to identifying if an animal on a picture is actually a wolf or a husky, the algorithm could potentially only look at the environment and call an animal surrounded by snow a husky. This example shows how algorithms could be smart but stupid at the same time and why we should not blindly trust their judgement.
One of the conditions for Artificial Intelligence to work is to beforehand gather a collection of data and structure it carefully. Luckily collecting data is not a problem in this digital age. Many companies could have an advantage when it comes to implementing Artificial Intelligence because of the information that they hold. However, only having big numbers of files is not enough. All of this information needs to be structured before it is useful. This is a task that takes many many hours of human labour. The challenge for the future would therefore be to invent a more efficient way to structure this data, or even carry out the process of structuring data while simultaneously collecting it.
Interaction between man and machine
For now, many experts agree that it is unlikely that technology will completely take over our jobs. Even though machines could bring a lot of efficiency to the workplace, it is questionable if they will ever have the same emotional and creative capabilities as humans. Therefore, instead of focussing on worrying about job replacement we could concentrate on getting more comfortable interacting with machines. A scenario in which you work together with a robot to complete tasks is almost inevitable, so why not start getting used to this.
It is very tempting for companies to jump on the innovation train as soon as it takes off, but in essence technologies such as Artificial Intelligence have not been developed to be used as a way to be cool. They offer serious opportunities to improve human experiences which should be the focus when working with the technology. As a result, organizations are facing the challenge of first identifying problems which ask for meaningful solutions. If these solutions could be born out of Artificial Intelligence implementation, go for it! Yet, if this technology is not the answer, don’t rush into it.
This blog is the second one in our series 'A future with AI' by shareNL. We are a global agency specialized in the sharing and platform economy acting at the interface of technology and society. This means that the topics we discuss are not limited to the sharing and platform economy only, they also include new technologies such as Blockchain, IoT, Robotics and Artificial Intelligence. Together with our ecosystem we create impact through insights (publications), inspiration (consulting and presentations), intelligence (strategy development and concept development) and interaction (pilot projects and experiences).
Are you interested in learning more about Artificial Intelligence? Stay on the look out for next week's blog post which will be focussed on societal trends that might influence technology in the near future. If you would like to discover the opportunities of working together on the topic of Artificial Intelligence, don't hesitate to get in touch with us!
Special thanks from shareNL to Jasper Wognum, Peter Blomsma and Daniel Gebler for taking their time to answer questions about their personal experience with Artificial Intelligence.