Although not an avid app user, I do take regular benefit from a select few and less frequent benefit from others. This summer, I found I was turning to one particular app more and more. In a similar vein to my app usage, I intermittently engage in bird watching. As you will have guessed, there is a connection between the app in question and my ability to recognise the different birds that I observe. Visually, I am happy to use old-fashioned print technology: looking the birds up in a guide when I return back indoors. However, in late spring, summer and early autumn, the dense foliage makes spotting the birds a challenge. That is where the app comes in. It provides automated recognition of bird calls, so I can identify birds that I can hear but not see. It is not foolproof: birds will often go silent, fly off or dive below the water. It is as if they know I am trying to record them and take umbrage.
Despite the birds’ best efforts and hindrance from human voices and motor noise, I have managed to identify a variety of species. Dunnock, willow warbler and greenfinch, to name a few. It also allowed me to establish the identity of the birds nesting in the trees, which I suspected were rooks or possibly crows. Turns out they were jackdaws. Each identification is graded certain, almost certain, likely, uncertain and highly uncertain. The app has been developed by an interdisciplinary team of biologists and engineers. The main developer works on artificial intelligence (AI) applications using convolutional neural networks for bioacoustics, environmental monitoring and the design of mobile human-computer interaction. I think I am safe to state that this is an application of AI that has enhanced my enjoyment of the natural world.
However, there are other applications of AI that impact on our lives in a more opaque and insidious manner. Algorithms can direct young people to harmful content with tragic results, as we know from the findings of a recent coroner’s inquest. Luminaries in science and technology have warned that AI has the potential to be humanity’s biggest existential threat. I read in a national newspaper that a third of AI researchers surveyed believe it will result in a catastrophe within the next 80 years, while nearly three quarters expect it to cause large-scale societal changes. Proponents often claim that AI applications and robots will free up humans from laborious tasks to do other more enjoyable things. In another column in the same newspaper, it was reported that an AI-generated paragraph was read out to the American Congress. While AI-generated text and the conversion of text to images may be seen to aid creativity, they can just as easily perpetuate subtle and extreme biases. In addition, outputs based on machine training on content taken from the internet opens an issue of legal rights. As shown in the output from the bird call app, AI algorithms often work on the balance of probabilities and, although machines can calculate these efficiently and eliminate human error, they lack any sense of risk.
Concerns over the potential downsides of the application of the technology have led to what is referred to as ‘human-centric AI’. The EU vision for such an approach is to ensure that there is focus on regulatory and ethical matters to promote the responsible development of trustworthy AI and to ensure that AI works for people and protects fundamental rights. An article by Stanford University provides more details on human-centric AI [1]. It states that: “keeping people in the loop can ensure that AI is working properly and fairly and also provides insights into human factors that machines do not understand.”
Artificial intelligence is already being used in many areas, including in non-destructive testing (NDT). Its ability to automate NDT image analysis and identify patterns indicative of defect signals offers a big improvement over conventional operator analysis. The human-centric approach is being applied, working with the operator to highlight where they may want to investigate further. We have discussed such augmentation previously. It avoids the black box approach, allowing transparency of how the decisions have been derived and hence confidence in the output. The human-centric approach to AI offers many opportunities for non-destructive testing and we should look to promote and continue to take advantage of it.
As we move into winter, I look forward to the bird app confirming that a robin is singing in my garden. On that note, I wish you all a happy and peaceful Christmas.
Reference
Please note that the views expressed in this column are the author’s own personal ramblings for the purpose of encouraging discussion within NDT News. They do not represent the views of Jacobs or BINDT.
Comments by members
This forum post has no comments, be the first to leave a comment.