Summary
A recurring theme in this week is that there is an apparent abundance of reasons for us to be concerned about the use of AI technology, to the point that in many cases it is clear that such technology should not be used, because it is both unsafe and prone to error. While humans do not do well with repetitive algorithmic tasks, AI and Data Science technology performs far better on these types of tasks – such as looking at the quickest way to ship something, or the most cost-effective way to organise a manufacturing process, or predicting complex weather patterns – where it is far clearer what counts as a successful outcome, for example when something is measurably faster or more efficient or accurately predicted. On the other hand, AI performs far worse in tasks that involve complex decision making, particularly with respect to social phenomena (e.g. ethics, fairness, transparency, sustainability) that do not have clear outcomes, and there is currently not much prospect for AI to do this autonomously from humans.
Given all of this, might there then be good reason to do a major rethink? Perhaps we need to step back and question whether, in fact, we should be using AI, at least in certain cases?
As a first response to this question, it is important to recognise that the set of technologies often referred to under the label of AI technologies are already under quite a lot of scrutiny by those developing them. However, such scrutiny is more effective in some settings than others, some large companies and even some government agencies do not seem to have carried out the right level of scrutiny, with serious consequences for clients or customers.
A second response to this question is that indiscriminately attempting to put in place serious obstacles to progress and innovation, is neither possible given the increasingly high levels of public and business interest and activities in AI technology, nor is it desirable given the range of potential benefits such technology may well deliver in the coming years.
This OpenLearn course is an adapted extract from the Open University badged course LG003 - An introduction to artificial intelligence [Tip: hold Ctrl and click a link to open it in a new tab. (Hide tip)] .