AI is already having social and environmental impacts but the emerging risks are with Artificial General Intelligence

By Eli Fathi (C.M.) and Peter MacKinnon

Eli Fathi (C.M.) (photo at left) is chair of the board at Ottawa-based MindBridge Analytics. Peter MacKinnon (photo below at left) is a senior research associate in the Faculty of Engineering at the University of Ottawa, and Member IEEE-USA Artificial Intelligence Policy Committee, Washington. D.C. 

(AI image at right created by Randal Adcock , Edmonton).

Generative Artificial Intelligence (GAI) products are rolling out at a dizzying speed,  and promising increased human productivity, enhanced operational efficiency and new application opportunities. But with GAI, there are cracks emerging in today’s underlying economic and societal foundations.

One such fracture deals with the impact on the electrical grid, both generation and distribution, while the other is whether society is ready for the new AI-enabled world. 

GAI is a power hog requiring large amounts of electricity to energize the computing resources in data centres used to train Large Language Models and handle the vast number of queries being processed.

In a recent survey by Pure Storage, 81 percent of global chief information officers and senior IT leaders, said they “Believe that AI-generated data is likely to outgrow their organization’s current data centres.” Furthermore, 98 percent agree that their infrastructure needs “improvement to support risk and innovation initiatives.”

According to a recent Goldman Sachs study,  “on average, a ChatGPT query needs nearly 10 times as much electricity to process as a Google search.

Notwithstanding the energy savings stated by Koomey’s law that the energy efficiency of computers doubles in 18-month cycles, the same study predicts that by the year 2030, data centres will consume three to four percent of the worldwide electricity. 

The rapid rise in electricity demands of GAI-based applications are additive to power-guzzling crypto mining and electric vehicle energy needs.

Collectively, all these sectors are putting pressure on the need for more electricity generation when existing capacity is nearing its limits and coping with inadequate power grid infrastructure. 

GAI energy needs are incremental to the power the world is already consuming, and will continue to increase for the foreseeable future.

Although tech companies are promising zero carbon dioxide emissions for the thousands of data centres being built, some of which are hyperscale in size, the accelerating  energy demands will put a significant strain on available power generation infrastructure around the world.

This is placing tech companies on a collision course to achieving zero emission as highlighted by a Washington Post article, which states “ the voracious electricity consumption of artificial intelligence is driving an expansion of fossil fuel use.”

To overcome this issue, tech companies are seeking clean energy alternatives, including nuclear power and geothermal energy solutions that can operate off-grid. 

The gap between the accelerating electricity demand and the emerging clean power generation curve makes it challenging to shut down carbon-emitting electricity generation plants. The bottom line is that power-hungry GAI will lead to purpose-built dedicated new electricity generation infrastructures without drawing power from the electrical grid. 

The current electrical grid is already stretched in many locations around the world and some jurisdictions already experience intermittent power outages. Aside from the potential for an increase in power outages, unless new clean energy infrastructure is added there is a strong likelihood of a higher cost to consumers. 

Arrival of Artificial General Intelligence will transform society

To date, there is a continuum development in GAI with an accelerating rate of advancement, with no clear consensus among experts when to expect the arrival of Artificial General Intelligence (AGI), which is anticipated to transform society as we know it.

When AGI arrives, based on the state of current R&D and underpinning microelectronics, it will be supported by sentient hyperscale server farms that will function at the same level of capabilities as normal human beings – or even beyond human capabilities.

Currently, there are diverging opinions by many experts regarding the time horizon to when AGI will occur, with forecasts as early as five years and as far out as many decades. According to McKinsey, “although gen AI tools such as ChatGPT may seem like a great leap forward, in reality they are just a step in the direction of an even greater breakthrough: AGI.”

Until then, the adoption of GAI continues to increase across the board by an ever-increasing range and number of organizations.

The technology adoption pattern follows the Gartner Hype Cycle innovation model, also known as the S curve. It provides a graphical representation of the maturity, adoption and social application of specific technologies and products. 

A breakthrough innovation tends to start with euphoria regarding its capabilities. As time progresses, the hard reality of implementation difficulties takes place, with some significant setbacks, as characterized by Geoffrey Moore’s book, Crossing the Chasm.  

After hard work and early market acceptance, the light at the end of the tunnel appears and, finally, the innovation experiences mass adoption and general use.  

According to an article in the Financial Times, 56 percent of Fortune 500 companies are now concerned with the potential risks of GAI to their operations. These include increased competition, operational impacts, regulatory oversights, cyber security risks and reputation damage due to potential human resources and privacy violations.

Today, as developments in GAI continue unabated, it is conceivable there will be some unexpected advancements that can have significant impact on our daily lives not too far in the distant future.

Consequently, there is an emerging danger of underestimating the speed at which GAI will evolve into AGI, making society unprepared for such an event.

We don’t know the capabilities of all the new products being developed in research labs around the world. It is plausible that a few of these products will experience breakthroughs which may far exceed the current expected capability trend curve.

If launched, such products will be way ahead of the legislation envelope and could create havoc with existing societal norms. 

The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the organizations they are associated with. 

R$