Kathy Gibson reports – Artificial intelligence (AI) is shaping up to be as big a world-changer as the general-purpose technologies (GPTs) that triggered the previous Industrial Revolutions.
That’s the word from Brad Smith, co-chair and president of Microsoft, speaking today to a group of South African government and business leaders.
“Over the course of history, technologies have redefined the economy,” he explains.
He says the first Industrial Revolution was driven by iron working, the second by electricity and machine tools, and the third – which we have been living through in the last 50 years – by computer chips and software.
“Whenever there is a new Industrial Revolution, people tend to fixate on who the leader is,” Smith adds. “But the real lesson from history is that the countries that advance the most are those where the technology is widely diffused and adopted.
“Because, if technology changes every part of an economy, the country that uses it most will benefit the most.”
This can be clearly seen in the case of electricity: gross domestic product (GDP) grows as electricity is adopted and used. “There is no country where this is the exception,” Smith states. “And the same is true in South Africa.”
This means the key to future success is in deploying general purpose technology, he adds.
“In the new AI era, we start with the strong conviction that it is the next GPT.”
Nations that plan to succeed in harnessing a GPT need to master four things, says Smith.
The first is the technology itself. “It turns out that every GPT is built on a technology stack. This creates a new economy since every layer of the technology stack matches up with new companies, new types of firms and new types of jobs.
Today’s AI technology stack consists of three layers: infrastructure, platform and applications.
“The IS layer is massive,” Smith says. “It requires the billions of rands that we are investing in data centres. Just as you can’t have electricity without power plants, you can’t have AI without data centres.”
The platform layer consists of the open source or closed source models that are created, to harness data and train models. “This is where a new generation of AI platform and software services are being created.”
At the top of the stack is the application layer, which is where people can get things done.
“With electricity, it was in the appliance layer that the magic happened, and the same is true today with the AI technology stack, where applications are what people see and use,” Smith says.
“To build a new AI economy you need to get all three of these things to get the flywheel turning. The infrastructure allows the models to be created, so people can create the apps – and then you can extend the infrastructure further.”
He singles out the South African Revenue Services (SARS) as a world-class example of an organisation using technology that includes AI to collect revenue and redefine how people interact with government.
A new economic model
When building a new economy, governments and businesses need to do more than master the technology, though. Smith points out that they need to master economics as well.
“The economics of AI comes down to economic structure, financial architecture and business models.”
As an example, he explains that every technology stack has an economic structure, with the bottom of the stack tending to be expensive and the top of the stack relatively cheap.
“It is the same with AI: you see billions of dollars invested at the infrastructure layer; and find students can start to build new companies relatively cheaply at the applications layer.”
Next, it’s necessary to build a financial architecture. Smith explains that companies like Microsoft are working with partners to finance the construction of AI infrastructure around the world.
“This typically starts with private companies and private capital, then sovereign wealth and other public funding. This architecture is not just about unleashing capital, but about spurring demand as well.”
Where countries find themselves unable to muster the type of capital required, Smith says they could look at partnering together in geographic blocs. “Another way to accelerate this is public sector use.
“Government has a critical role to play in getting the flywheel turning, not only by setting policies but by getting every part of the public sector to use the technology.”
Creating the right business models are the final, and critical, step in enjoying sustainable success. “And the insight we can learn from history is that these business models evolve,” Smith points out.
Skills to the fore
The third imperative to taking advantage of the new AI economy is skills development
“This is only common sense,” says Smith. “If it’s key to have something used everywhere in the economy, the skills to do so must be everywhere in the economy.”
The previous Industrial Revolutions addressed this with technical institutes and apprenticeships in the first; land grant universities and industry standards in the second; and employee training coupled with computer science in higher education in the third.
As we look to skilling at scale for the AI era, Smith believes we need three types of skills: Ai fluency, where people can use the tools to do their work; AI engineering, which is the natural progression of computer science; and AI systems design, which equips business people to use AI in designing business cases, workflows and processes.
“Every country needs a national AI talent strategy to assess market needs and broaden AI fluency,” Smith says.
This would encompass not just universities and technikons but include initiatives like South Africa’s Youth Employment Scheme (YES) for broader skills transfer.
Trust is key
For a technology to become truly pervasive, it must have a level of social acceptance, Smith says.
“And for a technology to be socially accepted, it must be useful and it must be trusted.
“The construction of a real strategy and architecture of trust is indispensable for the future of AI.”
To build this trust we need to have privacy, digital safety and responsible AI, he adds.
“This is why we are building AI governance from the ground up, from internal company policy to global policy.”
A key part of the trust equation is sustainability, says Smith. “We can’t have a responsible approach to AI unless we address the need for sustainable AI.”
Making the AI economy a reality
To make the AI economy a reality, companies and countries need to be focusing on all aspects. “And this is amazing, exciting and daunting,” Smith says.
The secret to getting it all done, he adds, is inspiration. “Just think about all the amazing things we can do with AI.
“But, while we should be inspired by everything that can go right, we should draw equal inspiration from everything that can go wrong.”
As an example, he points out that it is now 140 years since Thomas Edison lit up lower Manhattan with electricity. And yet, today, most of the population of Africa has no access to electricity.
“This is a cautionary tale that should inspire is to do better,” Smith says.
“We have the opportunity to dream bigger, be bolder and pursue something that Edison could never have dreamed of doing.”