Published on the 22/03/2023 | Written by Heather Wright
But knowledge workers, creatives face disruption…
Generative AI is turning old job displacement models, based on technologies such as machine learning and robotic process automation, on their head – now ChatGPT-style technologies are coming for the knowledge workers and creatives.
New reports suggest that, assuming companies can overcome some of the current concerns, including their tendency to ‘hallucinate’, the impacts of GPT could be ‘pervasive’, with knowledge and creative workers – spared during earlier automation rounds – among the most impacted.
“Karen in the call centre is now the escalation expert and AI is the frontline.”
At a recent Brookings Center on Regulation and Markets webinar on ChatGPT and the future of work, Sanjay Patnaik, director of the Center on Regulation and Markets noted: “When we looked back a few years, it looked like automation in blue-collar jobs like self-driving cars, self-driving trucks would accelerate very rapidly, but to some degree the large language models (LLMs) now have overtaken those advances in automation in their pace and are really coming upon us very fast.
“And interestingly, a lot of those tasks that ChatGPT and large language models can actually replace or can help augment are tasks in the white-collar working space.”
That’s a view shared by the founder and CEO of Australian strategic research and consulting agency, Fifth Dimension, Lyndall Spooner, who says generative AI could replace knowledge and creative experts faster than customer service personnel, who already been disrupted.
Spooner says call centre workers’ jobs are ‘fairly safe’ right now, but those doing more creative knowledge-based internal functions in businesses and industry sectors should be concerned.
“Conversational AI models disrupted business models more than 30 years ago, allowing companies to scale the delivery of basic customer service,” she says. “Chatbots can answer simple customer queries 24×7 and as they have improved in their capabilities, they are fast becoming a preferred first point of contact.”
She notes that customers who call contact centres or go into a store or branch nowadays are customers who need advice or have a complicated issue to be resolved or don’t know what they should ask a chatbot.
“Karen in the call centre is now the escalation expert and AI is the frontline,” Spooner says.
Unlike traditional chatbots, the latest generation of conversational AI creates original combinations of text rather than simply retrieving a consistent response from a pre-defined programmed response.
That’s left Spooner suggesting Karen’s job is ‘fairly safe’.
She points to the desire to provide customers with consistent and accurate answers, rather than creative ones, generative chatbot’s inability to give advice, especially financial advice, or discuss options and its inability to help customers who ask incorrect questions, or to identify what the customer has misunderstood about a product or service.
“If customers are asking incorrect questions [generative] AI will respond regardless, whereas a customer service agent can interrogate the customer to ensure their information and questions are accurate,” Spooner says.
“We know from previous research that consumers place a lot of trust in what technology tells them to do because it generally looks and sounds accurate. Hence businesses need to consider their liabilities in terms of what generative AI could instruct customers to do.
“AI is a fast-evolving sector and there are many out of the box solutions being sold and utilised by businesses which range in quality and sophistication, but no AI model can currently replicate the intelligence and reasoning of Karen,” Spooner says.
It’s a different story, however, for the more creative, knowledge-based internal functions of businesses and industry sectors.
That’s where Spooner, and others, believe generative AI will likely replace simple creative functions which could benefit from speed of response, whether summarising disparate information into a simple narrative, generating basic code or generating creative outputs that fit a desired tone and call to action, such as a sales email.
“If AI is the new creative frontline then like Karen, creatives need to ensure they bring the high level expertise that can work with the generative outputs to deliver truly impactful and differentiated ideas that drive sales and strengthen brand reputation,” Spooner says.Spooner’s views echo those of several recent studies, including one by OpenAI, Open Research and the University of Pennsylvania, which says higher paying jobs are more at risk than lower-paying ones.
The report, GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models, investigates the potential implication of GPT models and related technologies on the US market.
It says around 19 percent of workers have at least 50 percent of their tasks exposed to GPTs if you take into account both current, and anticipated, GPT-powered software.
The report says people with professional degrees and higher incomes are more at risk, including those in sectors such as finance, education, journalism, engineering and graphic design. Those heavily reliant on scientific and critical thinking skills are less likely to face automation.
“Despite their limitations, LLMs are becoming increasingly integrated into specialised applications in areas such as writing assistance, coding and legal research, paving the way for businesses and individuals to adopt GPTs more widely.
“We emphasis the significance of complementary technologies, partly because out of the box, general purpose GPTs may continue to be unreliable for various tasks due to issues such as factual inaccuracies, inherent biases, privacy concerns and disinformation risks.
“However, specialised workflows – including tooling, software or human-in-the-loop systems – can help address these shortcomings by incorporating domain-specific expertise.”
Meanwhile, over at Princeton University a similar report, How will Language Modelers like ChatGPT Affect Occupations and Industries?, found the top industries likely exposed to advances in language modelling are legal services and securities, commodities and investments. Telemarketers and a range of post-secondary teachers, including English language and literature, foreign language and history teachers are also likely to be affected.
It’s not all bad news, however. Another study, this one at at MIT, saw more than 400 professionals completing ‘mid-level professional writing tasks’ such as writing short reports, drafting emails and creating analysis plans, half using ChatGPT and half without. It found a significant productivity boost for those using ChatGPT, who completed their tasks in 17 minutes on average compared to 27 minutes with improved results, with those using ChatGPT also reporting higher satisfaction.
“ChatGPT mostly substitutes for worker effort rather than complementing worker skills, and restructures tasks towards idea-generation and editing and away from rough drafting,” the non-peer reviewed report says.
“Exposure to ChatGPT increases job satisfaction and self-efficacy and heightens both concern and excitement about automation technologies.”
The reports come as OpenAI CEO Sam Altman unveiled the GPT-4 model, amid claims it can ‘pass a bar exam’.
Altman says the new model is more creative and less biased than earlier versions and can understand and talk about images and create eight times more text than GPT-3.5.
Despite that, the company admits it is ‘less capable than humans in many real-world scenarios’, though it exhibits ‘human-level performance’ on various professional and academic benchmarks.
It might also pay to remember that back in 2013, a University of Oxford study found 47 percent of US jobs could be eliminated by AI over the next 20 years – something the Information Technology and Innovation Foundation (ITIF) trounced in late 2022, noting that the occupation which had the highest risk of going the way of the buggy whip manufacturer – insurance underwriters – had actually seen employment growth of 16.4 percent in recent years.
“Overall… occupations with higher computerisation risk scores were only slightly likely to see job loss,” ITIF concluded.
