Why chipmakers’ stock surge is no guarantee of AI revolution’s future

While generative AI holds transformative promise for some, for others it heralds large-scale job losses and widespread ethical and security challenges
Even if this technological revolution lives up to its potential, it is clear its benefits will take years to reach consumers and firms in poorer countries

David Dodwell

Published: 8:15pm, 25 Feb 2024

Why you can trust SCMP

As the stock market value of Nvidia surges past US$2 trillion, making it the third-most valuable company in the United States, so in Europe the Netherlands-based ASML is moving towards US$400 billion, making it one of Europe’s most valuable companies. The equity markets have decided that the next technology revolution is under way, with Nvidia’s graphics processing units (GPUs) and ASML’s chipmaking tools providing the foundations for the newest and slickest digital kit as well as an artificial intelligence (AI) revolution that has Silicon Valley and the tech world transfixed.

For some, the powers of the likes of OpenAI’s ChatGPT and other forms of generative AI have inspired visions of massive productivity boosts. Transformative applications are appearing in areas such as medical treatment and healthcare.

For others, the breakthroughs spell large-scale job losses, an array of unresolved ethical challenges and a grave threat to the future of humankind. In the middle are sceptics who see a great deal of hype and some salutary stock market corrections not far over the horizon.

Some of the controversies are easier to manage. The machine learning driven by generative AI has already shown its value in speedy and reliable early indication of diseases, for example. We are at the earliest stages of discovering how these new technologies can enhance productivity in many fields.

It is also becoming clear that while AI is likely to lead to job losses in certain areas, it will also be at the heart of job creation in others. A PwC survey revealed that 32 per cent of executives had already adopted AI across their companies, with 58 per cent expecting it to improve product quality and services and 69 per cent predicting that staff would need to learn new skills.

Headaches are becoming clear, too. Among the most discussed is the capacity of AI algorithms to “hallucinate” – producing flagrantly false results and even inventing false but plausible evidence. Their capacity to create “deepfake” images and videos has generated serious privacy concerns if the technology falls into malicious hands.

A green wireframe model covers an actor’s lower face during the creation of a synthetic facial reanimation video, known alternatively as a deepfake, in London, Britain, on February 12, 2019. Photo: Reuters TV via Reuters

These concerns have triggered intense debate on how best to regulate these new technologies and keep them under trusted human control. AI safety institutes are being established, and the European Union is in the process of drawing up an Artificial Intelligence Act.

Opinions range widely on the extent of risk, though, and agreements on how best to regulate are not likely to be reached any time soon.

Many are concerned by the extraordinary vulnerabilities evident across the new tech sector. Foremost is concern about extraordinary market concentration. Nvidia, for example, accounts for an estimated 82 per cent of the world’s supply of the GPUs at the heart of AI innovation. Taiwan Semiconductor Manufacturing Company (TSMC) accounts for around 58 per cent of the world’s semiconductors.

ASML in the Netherlands has a virtual monopoly on the extreme ultraviolet lithography tools used by the world’s leading chip producers. Its newest machine will cost around US$380 million. It weighs the same as two Airbus aircraft and will require 250 engineers for installation alone.

The microscopic precision of the equipment and the chips it produces requires a cleanliness of operation that would make a hospital operating theatre look grubby. Chip makers are vulnerable not just because they are so heavily dependent on a single supplier worldwide but because of the exacting operating environment needed to avoid contamination of their chips.

02:38

Apple supplier Foxconn to build ‘AI factories’ using US hardware leader Nvidia’s chips and software

Vulnerabilities have been augmented by the weaponisation of the hi-tech sector by those who fear the semiconductors that come from the likes of Nvidia and TSMC might be used by unfriendly competitors to undermine national security. Many users of semiconductors would like to reduce their vulnerabilities by seeing more producers competing as chip suppliers. Instead, this weaponisation is increasing their vulnerability and reducing the number of chip makers that are regarded as “trusted”.

Even if this latest technological revolution is real, it is also clear that it will take years for many companies across the world to experience any beneficial impact. Companies and consumers in poorer countries will be left behind. The machine learning driven by the new AI technologies needs prodigious computing power and access to reliable electricity – not a small matter for the 775 million people worldwide who lack access to electricity.

AI regulation will help drive growth – the biggest tech companies know this

Even the current fractious state of international trade could create obstacles that limit the spread of the AI revolution. As we head into this week’s World Trade Organization ministerial meeting in Abu Dhabi, leaders are struggling to reach a deal to retain a 26-year moratorium on tariffs on digital trade and the cross-border movement of data. Failure to reach agreement will have a serious impact on all digital services and almost certainly frustrate the potential roll-out of AI-based innovation.

All this leaves unanswered the existential questions over whether the AI revolution represents a threat to the future of humankind. Some might find such fears paranoid, but I’m not sure. Geoffrey Hinton, who until May last year sat at the heart of Google Brain, provided a blunt warning in Oxford University last week: “If a digital superintelligence ever wanted to take control, it is unlikely that we could stop it.”

A sobering thought, but for the time being I think we can borrow a well-known phrase and say reports of humankind’s death have been greatly exaggerated.

David Dodwell is CEO of the trade policy and international relations consultancy Strategic Access, focused on developments and challenges facing the Asia-Pacific over the past four decades