Limitations of robots and AI

Here is an interesting read from the BBC on 'four things to help us understand our AI colleagues'.

The article summarises the current state of play with regards to AI and how it impacts on the future of work. Of note is the summary of findings from the 2017 McKinsey Report stating only 5% of jobs would be eventually fully automated but 60% of occupations could see 1/3 of their roles be undertaken by robots.

We see the second scenario panning out now in many ways. For example, almost all the 'finding our more information' component of my research, is undertaken through access to databases which filter my searches. The results are collated into another database which is my bibliographical Endnotes. Instead of using manual index cards, the searching is done through electronic means, with its inherent biases and challenges.

The article provides for four rules which support the argument that robots are not quite ready to afford us of total leisure. These are:
- Robots don't think like humans
- Robots are not infallible - they make mistakes.
- Robots are not able to explain why they made a decision
- Robots may be biases.

All of the above can be circumvented with sufficient resourcing, but for the moment, there is some importance in ensuring we humans understand the limitations of robots and AI. It is especially important to work on the ethical issues around how robots and AI are governed as it is how these entities are 'programmed' with their inherent prejudices, which will dictate how they react and make decisions.

As per the book 'Smarter than you think' - see overview - we all need to learn now to work with 'smart machines' and one aspect of working with these 'tools' is to acknowledge their strengths and weaknesses. Robots and AI are powerful tools to augment human work and ensuring everyone understands how to best work with these tools, is one important aspect for the future of work and education.