Many people still see Artificial Intelligence (AI) as a science fiction dystopia. However, as artificial intelligence has improved at a fast pace in recent years, many organizations across different industries are already widely embedding AI within their business processes (Uzialko, 2019). AI gives us great potential to tackle major challenging problems we face nowadays, for example, it could optimize our electrical grids to reduce our increasing energy and it can help us to improve accuracy of medical diagnoses (Lemon, 2019). Even though this prospective might sounds promising for our society, AI may also involve some serious unexpected risks which are not considered thus far. Hence, AI is proving to be a double-edged sword, as it comes with great power but also great responsibility (Cheatham, Javanmardian, Samandari, 2019).
Yet, before organizations are able to bear the responsibilities that come with AI, they have to understand where potential risks may be hidden. Especially, because business leaders, according to Cheatham et al. (2019) “tend to overlook these pitfalls and overestimate their capability of mitigating these pitfall”. So, it is crucial to point out some of these pitfalls that organization, as it could help them to recognize the risks that come with AI before they might prey fall to. This blog highlights five potential risks, whereas the first three relate to the empowerment of AI, and the final two refer to the interaction of humans and machines driven by AI algorithms.
Risk 1 – Data difficulties. Using, sorting, connecting of data has become more complicated as the amount of unstructured data from mobile devise, social media or the Internet of Things has increased tremendously over in recent years. Consequently, companies, often unknowingly, trap into pitfalls such as unintentionally using or exposing highly private information hidden among anonymous data. It is crucial for companies to be aware of this, as they have to comply with the General Data Protection Rules (GDPR) to avoid reputation risk (Cheatham et al., 2019).
Risk 2 – Technology Trouble. Technology and process complications in the companies’ operating system can have a negative effect on the performance of the AI system (Cheatham et al., 2019).
Risk 3 – Security snags. Furthermore, a rising concern is the possibility of unauthorized parties to have access to data, such as marketing, financial or health data, which is collected by companies. Especially, if the company itself does not consider this data as sensitive at first sight, they might prey fall to unauthorized parties taking advantage of the data that fuels their AI systems. As a result, companies could experience consumer distrust leading to reputation damage, as well as regulatory consequences (Cheatham et al., 2019).
Risk 4 – Models misbehaving. Also, AI models themselves may form a potential risk for companies. AI powered models collect, track and analyse huge amounts of data, as a result they can deliver biased outcomes, draw conclusions of which the actions make no common-sense in the real-world (Uzailko, 2019) or become unstable (Cheatham et al., 2019). Bad data used to train AI can cause models to misbehave (IBM, n.d.) but AI models can also accidently misbehave. This can be exemplified by AI models accidently discriminating, like gathering zip codes and income data to create targeted advertisements for people with a only an above average amount of income (Cheatham et al., 2019).
Risk 5 – Interaction issues. The misalignment between humans and machines could be another critical pitfall, as humans often set the goals for AI machines (Cheatham et al., 2019) If we are not clear with the goals we set for AI machines or if we make script errors when developing algorithms, the AI system can become dangerous because it is not armed with the same goals that we aimed for. For instance, when traffic rules are not defined well enough in the AI system by humans, automated transportation can lead to terrible accidents (Marr, 2018).
Since AI models will be embedded within more and more business operations in the future, it is crucial for companies to recognize the potential risks of AI in order to bear the responsibilities and to deal with its consequences. If companies have a better understanding of what AI risks might drive, they will have a greater chance of catching the risks before the risks catches them (Cheatham et al, 2019). Furthermore, it enables companies to deal with the double-edged sword due to developing AI models in its best way, while anticipating on potential danger and consequences for the society as a whole.
Sources:
Burkhardt, B. Hohn, N. & Wigley, C. (2019). Leading your organization to responsible AI. McKinsey. [online] Retrieved from https://www.mckinsey.com/business-functions/mckinsey-analytics/our-insights/leading-your-organization-to-responsible-ai
Cheatham, B. Javanmardian, K. & Samandari, H. (2019). Confronting the risks of artificial intelligence. McKinsey Quarterly. [online] Retrieved from https://www.mckinsey.com/business-functions/mckinsey-analytics/our-insights/confronting-the-risks-of-artificial-intelligence
IBM. (n.d.). 5 in 5: AI and bias. IBM. [online] Retrieved from https://www.research.ibm.com/5-in-5/ai-and-bias/
Lemon, M. (2019). 5 Weird Problems AI Could Solve. Postfunnel. [online] Retrieved from https://postfunnel.com/5-weird-problems-ai-could-soon-solve/
Marr, B. (2018). Is Artificial Intelligence Dangerous? 6 AI Risks Everyone Should Know About. Forbes. [online] Retrieved from https://www.forbes.com/sites/bernardmarr/2018/11/19/is-artificial-intelligence-dangerous-6-ai-risks-everyone-should-know-about/#46d1f1542404
Uzailko, A. (2019). How Artificial Intelligence will transform business. Business News Daily. [online] Retrieved from https://www.businessnewsdaily.com/9402-artificial-intelligence-business-trends.html
Hi Annick,
What an interesting article on this popular topic. I agree on these risks, however when I think of risks concerning AI, I mostly think of problems that will arise due to AI becoming too powerful.
Although this is maybe a more long-time risk, I think the potential effect of this risk is larger than the ones named above.
Therefore I would like to ask you: what is your opinion about AI taking over jobs that require basic human skills and thereby replacing humas?