Will ChatGPT power or inflict pain on the economy?

‘The new electricity’

Last week, the Bank of America equity strategist Haim Israel weighed in on what advancements in AI may signal for economies, reportedly saying in a client note: “We are at a defining moment — like the internet in the 1990s — where artificial intelligence is moving towards mass adoption, with large language models like ChatGPT finally enabling us to fully capitalise on the data revolution.”

To drive home his point, Israel predicted: “If data is the new oil, then AI is the new electricity.”

These are significant comparisons. 

The expansion of electrical technology during the second industrial revolution, led to a huge increase in productivity, as businesses were able to produce goods and provide services at a faster rate. The rapid advancement of mass production, during the late 19th century into the early 20th century, led to the rise of large conglomerates as well as labour unions.

Many decades later, in the 1990s, the rise of the internet had a profound impact on the global economy, lowering barriers to competition and widening the reach of markets. The internet also drove growth in the so-called knowledge economy, a term popularised by Peter Drucker.

In 1969, Drucker, who is considered the father of modern business management, wrote in The Age of Discontinuity: “Knowledge, during the last few decades, has become the central capital, the cost centre and the crucial resource of the economy. This changes labour forces and work; teaching and learning; and the meaning of knowledge and its politics.”

[related_posts_sc article_id=”540283″

ChatGPT — which, after reaching 100 million active users after just two months, set the record for the fastest growing application in history — is the most recent advancement with the potential to profoundly alter the knowledge economy by undermining its value. A disruption to the knowledge economy would strike at the heart of the finance, real estate and business services industry, South Africa’s largest industry.

The knowledge economy

“ChatGPT is promising (or threatening) to reinvent the way people work in a multitude of industries,” ENSAfrica pointed out in a research note outlining the technology’s impact on the legal industry. ChatGPT, the law firm noted, can be used for legal research, drafting emails and legal documents, contracts and document review and answering legal questions or defining legal concepts.

“It changes the way that we engage with information,” Aalia Manie, head of legal and technology solutions at Webber Wentzel, explained.

“So, in many respects, it will have the biggest impact on information and knowledge work, because of the way in which you engage and get a human-like response and the feedback is a broad synthesis of all the information that may have been, for example, on the internet.”

Because of this, Manie said, jobs with a big research element may be affected by the use of ChatGPT.

San Francisco-based OpenAI, ChatGPT’s creator, appears to be alive to the technology’s economy-shaping potential.

Almost exactly a year ago, OpenAI made a call for expressions of interest to study the economic impact of large language models on individuals, firms and society. The call was specifically about the potential impact of Codex — which analyses natural language and generates code — with the view that the research would help shape future assessments of other large language models.

As large language models improve, OpenAI said, “they have the potential to impact many aspects of society, including work, productivity, skill development and other economic outcomes”. 

Codex, for example, was expected to complement or substitute the tasks completed by workers in coding-centric occupations, like engineers, data analysts and software developers.

Last month, OpenAI launched a survey looking for responses from users in different industries about how ChatGPT impacts their work. “This information can help improve OpenAI’s understanding of the real world impact of its AI systems so that they can be safely developed and deployed into the economy moving forward,” the lab said.

Navigating risks and limitations

According to Manie, ChatGPT’s limitations, which often results in various inaccuracies, means its use will not necessarily result in large-scale job losses.

ChatGPT will, however, influence what kind of jobs are coveted, Manie said. People who are skilled at, for example, creating effective prompts for large language models could find themselves in demand. 

Overall, these technologies can have positive economic outcomes — if their deployment is carefully managed and regulated, Manie added. “Definitely, it is disruptive. But disruptive is not necessarily negative and not necessarily positive. I think it can be both, in the same way that the internet was.”

[related_posts_sc article_id=”539138″

South Africa’s regulation framework may, however, not be robust enough to protect against AI’s potential pitfalls, such as copyright infringement. ChatGPT-generated content is also vulnerable to churning misinformation and bias, as the data the model uses is not automatically clean.

Carla Collett, a partner at Webber Wentzel whose expertise is in intellectual property law, noted that in South Africa, content generated by ChatGPT can be owned by an individual, despite it being a synthesis of other people’s work. This is because of provision in the country’s copyright law for computer-generated works.

Africa’s catch up

Countries ought to regulate how AI models are built and used, Manie said. She cited the European Union’s Artificial Intelligence Act, which classifies technology applications based on risk, banning those that present an unacceptable risk. EU lawmakers have reportedly considered redrafting the regulations in ChatGPT’s wake.

Shamira Ahmed, the executive director of the Data Economy Policy Hub, said there should be a balance between “the hype, the reality and the risks” associated with AI models, noting that there is a dearth of African-led research on the subject, which tends to come from jurisdictions with greater AI-related capabilities. 

“In South Africa it is kind of weird for us to be talking about AI disruptions, when we do not necessarily have the infrastructure and the access to electricity needed for AI … There are a lot of challenges in the African context,” Ahmed said.

“So we need more African-led research to understand the impacts and opportunities in our ecosystems, which have their own systemic problems.”

Ahmed noted that standards and guidelines relating to generative AI are formulated at a global level, meaning that African countries become standard-takers next to economies that have more geopolitical heft. Africa’s public sector officials, she added, need to be equipped to lead the charge on how AI technologies are deployed in their economies.

Countries that already have a competitive edge, human capital and more AI maturity will reap the technology’s economic benefits quicker, Ahmed said. African countries, she added, need to both understand the risks associated with AI and invest in electricity and ICT infrastructure to avoid being left in the dark.

“In many African countries there is a duality where we also need to think about electrification and connectivity at the same time as thinking about using AI. So there is that challenge. We need agile governance. And we need holistic approaches to work with the dualities of 20th century problems and the 21st century’s advancements in technology.”