Artificial Intelligence is no longer a technology that is on the way. It has arrived.
Within two months of the release of ChatGPT in November it had over 100 million active users, making it the fastest growing online consumer application ever.
The capabilities of it and tools like it are remarkable. They can summarise books and generate images in mere seconds - but what effect might they have on how we work? And which jobs are they likely to impact most?
An academic team in the United States have been trying to answer those questions.
The researchers from Princeton, New York and Pennsylvania universities analysed which workplace abilities were mostly closely related to those exhibited by AI systems like ChatGPT - called large language models - and image generating AI tools.
They then calculated which of 800 specific occupations are most exposed by evaluating the abilities most important for those roles.
Their recently released paper ranks those occupations from most to least exposed, concluding that "highly educated, highly paid, white-collar occupations may be most exposed."
Speaking to Prime Time, one of the paper's authors, Professor Manav Raj of the University of Pennsylvania, said the legal and education sectors scored highest in their analysis.
"I think that reflects that in all those occupations, communication becomes really important and processing information becomes really important. We know language modelling tools are really good at those kinds of things."
"Some of the occupations that are highly exposed to AI will disappear," his co-author, Professor Robert Seamans of New York University added. "But I think that's probably unlikely in most cases."
"What we have in mind by saying 'high exposure to AI’ is that the job will change in some way to take advantage of the things that AI can do well - it does not mean that the occupation is disappearing."
ChatGPT was developed by the Microsoft-backed company, OpenAI. Its capabilities have stunned users and surprised experts, triggering a global wave of interest in artificial intelligence.
Since its November launch, Google, Meta (owner of Facebook) and a variety of other companies have released or updated their own large language models and tools.
Large language models are typically programmed to make connections between words, datasets and within digitised networks using a process called deep learning. Other similar models can be trained on pictures and art to generate imagery.
They can respond to users’ requests with written answers, output and fix computer code, analyse and reorganise data, generate near-photorealistic images and art, recognise patterns, and complete a variety of other computer-based tasks.
As the responses are directly generated in reply to each user request, together image generators and those which mainly output text are categorised as generative AI systems.
The recent improvement in their capabilities has led to widespread discussion about their potential to disrupt economies and displace workers.
Earlier this month, IBM CEO Arvind Krishna told Bloomberg News he could "easily see" 30% of tasks done by back-office staff like those in human resources "getting replaced by AI and automation over a five-year period."
In March, analysts from Goldman Sachs said two-thirds of current jobs are exposed to some degree of AI automation, and "generative AI could substitute up to one-fourth of current work."
While last week, a report from Dublin-headquartered professional services firm, Accenture, said four in ten work hours "will be impacted by large language models."
Beyond occupations within education and the legal sector, the researchers also ranked roles within the media, banking and finance, software and web development, engineering, design, and publishing as among the most exposed.
In the paper, two separate rankings are published. One considers exposure to language generating AI models (like ChatGPT) and another considers exposure to image generating AI models (like MidJourney and Dall-E).
Most exposed to AI language generators are telemarketers, English and literature teachers, and foreign language teachers. Interior designers, architects and chemical engineers are ranked most exposed to image generating models.
Least exposed are dancers, textile workers, bricklayers, fitness trainers and massage therapists.
The paper says: "the effects of generative AI technologies are likely to manifest relatively quickly."
Dr Seán Ó hÉigeartaigh, the director of the Centre for the Study of Existential Risk at the University of Cambridge, agrees.
"I think that this will go from being something of a tech curiosity to something that impacts millions of people very quickly."
"It's going to reduce a lot of the drudgerous work that we take time on, but I also think there are whole swathes of our global society who really don’t have much of a voice in what's happening," he said.
"Remote workers in India who are doing white collar work, annotating data, or doing call centre work - they could easily be completely out of a job very soon."
"No doubt new jobs will be created with the help of AI, but those jobs will not necessarily go to the people who have been losing those jobs. People who for the first time have had an opportunity to pull themselves up to a higher quality of life and are suddenly just cut out of that."
Dr Ó hÉigeartaigh is a signatory to a recent open letter published by Oxford University’s Future of Life Institute. It called for a six month pause to the training of certain AI systems.
"It would make sense to move more slowly to ensure we have time to involve voices from across society, from the industries that are going to be affected and the areas of employment that are going to be affected," he told Prime Time.
"People should have a voice and some democratic input into how AI is developed and put out into the world."
Professor Raj believes policy makers and politicians need to pay attention to the potential impact of AI on the economy.
"Even if this is an aggregate benefit, there will be some people that are going to experience turmoil and see some kind of pain points in their job," he said.
"Some jobs will probably get wiped out and there will be some need for retraining and adjustment, but I'm optimistic."
"I think there will be new jobs and new kind of roles that emerge as we learn how to use these technologies, and as they open up different [employment] pathways."
Professor Seamans has a similar view.
"Don't worry about AI taking your job. That's very, very unlikely to happen."
"However, there's a good chance AI will change your job. So, in order to prepare for that, start adopting and using these technologies yourself."
Many researchers, industry leaders and analysts who spoke to Prime Time over recent months broadly agree that artificial intelligence tools now exhibit traits of what they call general purpose technologies.
General purpose technologies are tools or systems that can change not just industries but economies and societies.
Previous examples in human history of general purpose technologies include the steam engine, the printing press, and the personal computer.
"It's people who are using AI will replace people who are not," said Professor Patricia Maguire of University College Dublin.
"We cannot leave anybody behind here. So, we need to educate, I think that education needs to start even at a basic primary school level."
Barry O’Sullivan, professor of Computer Science in University College Cork said it is impossible to imagine what jobs will emerge from the mass adoption of artificial intelligence tools and large language modelling by industries.
"We can guess at some of them, but we have no idea what the world is going to be like in five years' time, never mind in ten- or 50-years' time."
A Prime Time special programme on artificial intelligence broadcasts Tuesday, May 16 at 9.35pm on RTÉ One.
