It is just over two years since the launch of ChatGPT, and while artificial intelligence (AI) has undoubtedly had a massive impact on the stock markets since then, observers agree the technology has largely failed to live up to the early hype that surrounded it.
The truth is AI has been around for 75 years in one form or another, points out Séamas Kelly, who is professor of organisation, technology and society at the UCD School of Business. Many “AI summers” and “AI winters” have played out over that time, he says.
“During the summers, there is a sense of we are going to solve it, we will create a super intelligent machine and there is a huge amount of excitement and investment, but then it hits the barrier, then an AI winter follows that, where there is a real sense of letdown and a lack of investment,” he says. “This is a very familiar type of cycle and the hype around AI in recent times has been outlandish.”
Kelly believes that, as a society, we are gripped by “techno optimism” and as a result we are poor at questioning any technological change, not just AI. “When you have this kind of culture that sees technology as always being progressive when you add that to something like AI, it can lead to hysteria.”
‘A gas emergency would quickly turn into an electricity emergency. It is low-risk, but high-consequence’
The secret to cooking a delicious, fuss free Christmas turkey? You just need a little help
How LEO Digital for Business is helping to boost small business competitiveness
‘I have to believe that this situation is not forever’: stress mounts in homeless parents and children living in claustrophobic one-room accommodation
According to Rory Timlin, data and AI lead with KPMG in Ireland, we are still in a hype bubble surrounding AI, but he notes the focus is now shifting from hype to finding meaningful value from AI – “benefits over buzzwords”.
“Given the noise around AI, it is important to be aware of the different kinds of AI systems and the different ways of integrating AI into an organisation’s digital capability,” he says, noting that, while Generative AI (GenAI) has recently dominated the airwaves, it is only one form of AI being employed in real-world application.
“There are sound proven use cases for AI and where AI is integrated into end-to-end business processes, it has a higher potential to create meaningful value,” he notes.
Machine learning systems have been successfully employed in areas including in supply chain optimisation, manufacturing efficiency, cyber security, digital marketing, media content recommendation and healthcare, while GenAI is now gaining traction in domains including customer service, digital content generation and marketing, he says.
KPMG’s recent Global Tech Report found that, while 70 per cent of organisations say AI investments are delivering business value, only 31 per cent reported delivering value at scale. “We see AI investment decisions moving from fast-following to being more return on investment focused, based on evidence and data,” Timlin says.
He echoes Kelly by noting that it is “vitally important” for an organisation to have the right processes and controls in place to deploy AI in a trustworthy and ethical manner. “A challenge with GenAI solutions is that they are extremely fluent in creating content that reads well. Without sufficient guardrails to guide appropriate use, to specifically train AI on accurate data and to verify the output for accuracy; people can over-trust the outputs GenAI creates.” Timlin also agrees that aspirations of creating AI systems with human-level general intelligence are still a long way off.
Eoin O’Reilly, EY Ireland partner and head of data, analytics and AI, agrees GenAI had its “break out” moment with the launch of ChatGPT. He believes that, while the hype cycle around GenAI is slowly beginning to subside, from a business perspective the surface has only been scratched when it comes to its potential for value creation.
“At EY we fundamentally believe that AI is changing everything, because AI is becoming part of everything,” he states. “The next decade will see a fundamental transformation of organisations, business models and value creation all underpinned by AI in combination with continued advances in digital, cloud and metaverse solutions.”
And despite the hype, O’Reilly warns that “waiting is not a winning strategy”.
“Delaying AI will not stop AI,” he says. “And ‘toe dipping’ experimentalism and incrementalism will only mean competitive advantage is being surrendered.”
The AI and data team at EY Ireland is working with a host of organisations to embed and harness AI in their operations, including partnering with organisations to improve efficiency in both clinical laboratory and pharmaceutical manufacturing settings and working with State agencies to support them to illuminate and improve their efficiency.
“Once you step away from the hype cycle and actually begin working with organisations, it quickly becomes clear that flaws are less around the capability of the technology and more around data quality and implementing a change management process to effectively harness AI,” O’Reilly says. Another concern that organisations may have is trusting the quality and accuracy of the output.
Kelly says there is a disconnect between what large language models (LLMs) can achieve and what people think they can achieve, noting that “it’s a very different type of intelligence”. Those who believe a truly super intelligent machine is around the corner “are way off”, he adds.
“Just because they do things like a human doesn’t mean intelligence – can a submarine swim, for example? Is a calculator super intelligent? A lot of hype has included some of the really important questions like what it would take to produce something really intelligent, but reliability is a huge problem that people are aware of too.” AI “hallucinations”, where it provides false or misleading information, are years away from being solved, he adds.
Yet Kelly is keen to emphasise how “extraordinary” the capabilities of LLMs are from a technical point of view. “They’ve really constituted an exciting leap forward in natural language processing and have surprised a lot of people,” he says. “This doesn’t mean that we shouldn’t treat the very hyperbolic claims made for them, for example, that we are approaching forms of artificial intelligence with the scepticism that they deserve. Technologies always have their politics – they are not neutral, and the implications associated with them are not inevitable – so we need to be discussing these and responding appropriately.”