The pitfalls and opportunities of AI

Aaron Woods detailed the practical applications of artificial intelligence (AI) technology within the optical sector during his workshop presentation at the National Optical Conference (18–19 May, Hilton Birmingham Metropole).

Woods, who has more than 25 years’ experience as a chief information officer in the charity and membership sectors, outlined different categories of AI – including machine learning, agentic, large language models (LLMs) and generative AI.

He explained that machine learning covers developments such as analysing medical scans and detecting fraud by identifying transactions that do not match a pattern.

Another example of machine learning is behaviour prediction – for example, by suggesting products that a person may like to buy.

Woods outlined that agentic AI covers a range of customer service and technical support applications, while LLMs answer questions and summarise data.

AI that provides an image from a written description, or transforms a written article into a podcast, were provided by Woods as examples of generative AI.

Woods highlighted that there is potential for optical professionals to use AI to enhance their personal productivity.

However, he encouraged conference delegates to be mindful of data security and ownership when using AI tools, such as ChatGPT and transcription services.

Woods highlighted that instead of thinking of these tools as a private conversation, it is better to treat information shared with AI in the same way you might a post uploaded to social media.

“That doesn’t mean that these AI tools are taking your data and chucking it out on the internet. But these algorithms are using your data to learn,” Woods explained.

He shared that data protection and commercial sensitives should be kept in mind – for example, a practice director should pause before putting their accounts into ChatGPT and asking for the most productive year.

“You have to be really careful about what you are sharing,” Woods emphasised.

“One of the issues with AI is that you can do some great things but you are potentially giving up your intellectual property,” he said.

Wood highlighted paid AI tools often provide more data protection as part of their service.

Wood emphaised the need to be mindful that some AI tools will pull information from different organisational software.

He shared a cautionary tale of an executive assistant who used AI to create a summary of the minutes from a meeting and was puzzled when the AI repeatedly referred to one staff member using a rude word.

“The theory was that two people were talking to each other directly in what they thought was a private online conversation during the meeting and referred to the person using rude word,” Woods explained.

“AI is listening, it is watching, and it will pull all these things together,” he added.

Woods highlighted that text drafting tools, such as Microsoft Copilot, can be useful in pulling out the salient points of a document and providing prompts on how to reply.

However, this should be balanced against the time it takes to edit the suggested response to ensure that it maintains the right tone of voice.

He observed that while AI can be helpful in taking minutes for straightforward meetings, humans are often still better at capturing the feeling in a room and knowing what to leave in and leave out.

Woods briefly outlined some of the ethical considerations around AI – including the use of other people’s source material as training data, the workforce effect of junior roles being taken by AI, and the environmental impact.

He added that the digital sector as a whole is responsible for more carbon dioxide emissions than aviation. In Devon, the heat coming off a data centre is used to heat a local pool.

Woods emphasised the need to take a considered approach to using AI in practice.

“We are at a gold rush point with AI where lots of people are going to make AI but it’s not necessarily going to be productive,” he said.

“Think of what the question is before you figure out that the answer is AI. Use AI sparingly for the right reasons,” Woods highlighted.