By Paul Sandle
LONDON (Reuters) -Education company Pearson (LON:PSON) said generative AI was helping its students tackle complex problems, but it was not infallible and its veracity depended on the quality of its training data.
Chief Executive Andy Bird pointed to promising feedback on tests of AI tools in its learning programmes as he delivered an upgrade in his final trading update before handing over to Microsoft (NASDAQ:MSFT) executive Omar Abbosh early next year.
"We've received positive initial feedback from our Generative AI tools and are evolving our AI capabilities to create further opportunities to maximise the potential of our trusted, proprietary content and data sets," he said on Monday.
Strong demand for English language learning and assessment and qualifications enabled an upgrade in its adjusted operating profit guidance by around 20 million pounds ($24.3 million) to up to 575 million pounds.
The group reported 3% total revenue growth in the third quarter, excluding its business under strategic review and the online OPM unit it has sold.
Bird said the company was testing AI in areas including language-learning speaking tools and its higher education courses.
But he said it was early days. For example several generative AI models had confidently come up with the same answer in a Pearson biology course, but it was wrong.
"In some regards generative AI can appear a bit like a bratty teenager who's always right, even when they're wrong," he told reporters.
"We have to make sure that any output from Pearson's use of generative AI is accurate so that it can be trusted. And that's where we see the value in our high quality datasets in terms of high quality data inputs lead to much higher quality data outputs."
Shares in Pearson, which have risen 15% in the last six months, climbed 1.7%.
($1 = 0.8245 pounds)