Since its November 2022 release, Microsoft-backed ChatGPT has become a household name. The free software developed by OpenAI has registered more than a billion subscribers in under six months, making it one if the fastest adopted technologies in history.
Attracted by the platform’s clean interface and instantaneous results, users have grown accustomed to the technology behind generative artificial intelligence (AI). Requiring only a few key input commands and minimal instruction, AI has shown its ability to create sophisticated graphic designs or write college-level papers, delivering output that is ordinarily achieved by humans in hours, in just a matter of seconds.
But despite the benefits on offer, several aspects of generative AI’s potential have come under the spotlight. When asked about the hype surrounding it, 99 year-old Berkshire Hathaway vice chairman, Charlie Munger, expressed initial scepticism, telling media, “I think old-fashioned intelligence works pretty well”.
Others, including Apple company co-founder, Steve Wozniak and Tesla CEO, Elon Musk, signed an open letter published by the Future of Life Institute, urging AI labs to pause training of those AI systems more powerful than GPT-4, for at least six months.
Value disruption
While businesses remain cautious, markets appear bullish. AI-related stocks have outperformed broader equities this year, with major tech groups leading S&P 500 gains. Microsoft has rallied by a third since the release of ChatGPT, while Alphabet has returned more than a fifth amid elevated expectations for its Google chatbot, Bard.
The potential for value disruption is evident. Media and education firms have already cited interruption to traditional business lines on account of chatbots. Online tutoring group, Chegg Education, for example, saw more than half of its market valuation erased after citing concerns that AI could impact future customer growth rates.
Companies are transitioning to what US tech company, Nvidia's chief executive, Jensen Huang, describes as a “tipping point” ahead of a new computer era. The chipmaker behind the H100 processor, which supports generative AI functions, breached the trillion-dollar market valuation level after the group reported better-than-expected sales for its fiscal quarter ending in May.
The firm’s upbeat revenue guidance coincides with the race by Asia-based computing firms to upgrade their AI ambitions. In May, Tokyo-headquartered SoftBank announced its collaboration with Huang’s firm to roll out new distributed AI data centres across Japan.
“Businesses in Asia recognise that generative AI is potentially a game-changer for their operations,” Amita Haylock, Hong Kong-based intellectual property (IP) and technology, media and telecommunications (TMT) partner at Mayer Brown, told FinanceAsia.
“There is a consensus that it could transform the way organisations operate – from HR to IT, risk and legal, to creations.… However, there is a lot more work to be done to see if generative AI will work at desired levels and we also have to acknowledge that generative AI has its challenges,” she said, underscoring the many questions that clients ask around ownership of ideas and plagiarism.
Shannon Ker, associate within Simmons & Simmons’ Singapore Corporate and Commercial team, echoed this call for caution.
She told FA that organisations should take utmost care to comply with personal data protection regulation when harvesting or processing AI-derived data. “Any extraterritorial effects can attract hefty fines because of non-compliance,” she said, adding that the area is difficult to navigate, “especially where the works cannot be easily attributed to a human author.”
“Pick-and-shovel” investments
Ultimately, how companies will respond to disruptive AI technology remains uncertain at this point in the cycle, industry experts suggest.
In a research note, Jefferies global head of Equity Strategy, Christopher Wood, wrote that in the meantime, fund managers are investing in related, backbone equipment. He drew a parallel with the “pick-and-shovel” proxies of the 19th century gold rush era, which saw businesses invest in the tools related to gold mining in order to participate in the trend while removing direct involvement in the speculative risks of the mining process.
Daniel Maier, head of Thematic Investing at Switzerland-based Vontobel Asset Management told FA that the enthusiasm surrounding AI is well-founded. While AI technology is not entirely new – in that it builds on current advancements – its capacity to impact a wide range of industries at scale, provides structural demand for the equipment that supports it.
“To handle the demanding nature of training large AI models, advanced chip architectures and next-generation data centres are required to address the significant energy consumption involved. While a large model in 2018 consisted of 1 billion parameters, today's large language models often exceed 1 trillion,” he said.
The “pick-and-shovel” analogy does not come without a hint of irony, as the surge towards chips and other AI-related assets mirrors elements of gold rush euphoria, drawing comparisons to the dot-com boom of the late 1990s. The Philadelphia Semiconductor Index, which tracks companies primarily involved in the sale of semiconductors is up nearly 40% for the first six months of 2023, closely followed by the ICE FactSet Asia Semiconductor Index, which is up by a third. Both continue to outpace mainstreaming indices thus far.
As Wood explains, the potential for the mass adoption of AI is much higher than anything that has preceded it, bringing investors into somewhat uncharted water. The bubble across equipment proxies appears inevitable as demand for chips exceeds supply. Following general economic principles, prices should normalise when competition attracts new production to meet demand.
Political factors
But the reality is that generative AI is inherently enmeshed with geopolitical pressure, distorting macroeconomic theories as the appetite for semiconductor chips coincides with trade tensions between the US and China. Both Washington and Beijing hold national ambitions to drive the future direction of AI, along with aims to control manufacturing.
The pair continue the fight for technology’s upper hand. In late May, Beijing effectively blocked imports of memory chips from US manufacturer Micron Technology, in a move that came less than a year after Washington issued export controls to cut off China from US semiconductor supply.
“Over the long term, China will attempt to enhance its domestic chip design capabilities. However, this endeavour is expected to be difficult and time-consuming due to the presence of various bottlenecks controlled by the US,” explained Maier.
He added, “Notably, American companies not only excel in AI chip design but also dominate in electronic design automation software and semiconductor manufacturing equipment components.”
The interest in “pick-and-shovel” equipment is not only inevitable, but warranted, suggested Germany-based Felix Roemer, founder of Gamdom, an online crypto gaming platform. Roemer told FA that AI renders the price of memory chips irrelevant, because there is an even more valuable commodity at hand: user data.
“In AI, data is king,” he opined. “OpenAI played it incredibly well, as ChatGPT is rooted in almost everyone’s lives, receiving millions of data streams for free.”
“Essentially, they have bought massive mind shares,” he continued. “People are telling the software their darkest secrets, including their billion-dollar business ideas, their code bases, their love affairs, everything.... It is the sort of data that will be incredibly valuable, not only to companies, but also to state level actors.”
Maier noted that while companies such as Baidu, Tencent and Alibaba possess troves of user data capable of informing their own generative AI systems, these platforms would likely underperform against ChatGPT, since they have less access to the trainable, unstructured data required to compete with it.
China’s generative AI is hampered by the external resources its data can pull from, and it relies on powerful hardware equipment parts, like microprocessors. Maier noted that while more than half of the world’s websites are in English, less than 2% are in Chinese and this limits where the market’s bots can draw from, with a likelihood for more domestically focussed results.
Regulation and growth
How policymakers instill protective guardrails around AI technology while simultaneously fostering innovation, is not a new challenge. But because there are no real precedents in this emerging technology space, and since its impact will cross different industries, uncertainty lingers.
At the balance, vigilance is critical to ensure careful implementation and oversight, Denys Peleshok, head of Asia at CPT Markets, a financial service provider, explained to FA. He motioned that any policy overseeing generative AI would require rigorous testing and security measures installed.
However, regulation rarely moves as fast as technological developments. “The cost of failing is quite high… less well-developed technologies could create biases or errors. In addition, their complexity could make it difficult to maintain,” Peleshok shared, echoing the concerns cited in the open letter signed by Wozniak and Musk.
But these issues have yet to derail enthusiasm for AI, as early estimates of its economic value are already sizable. Consulting firm, PwC, calculates that AI could offer the global economy output-related savings and investments of nearly $16 trillion by 2030, which is almost equivalent to China’s total GDP.
The early euphoria for AI is palpable, but the ultimate winner and regulatory outlook remains uncertain. What is certain – to Maier’s point – is that the days when having a phone or a computer was considered a privilege, are over.
Soon, failure to participate in the emerging world that is brought about AI will be a major handicap not only for companies, but also for the individuals that the technology leaves behind. In the meantime, as new AI relies on old data, it remains unable to predict the future… for now.
