Artificial Intelligence (AI) has found its way into general conversation after the emergence of large language models like ChatGPT, capable of generating text on a near-infinite range of topics. However, the discussion is increasingly turning to the search for Artificial General Intelligence (AGI), which, once discovered, could entirely change the game.
AI has become a common topic of conversation after the emergence of large language models like ChatGPT. These models are capable of generating text on a near-infinite range of topics. However, discussions are increasingly shifting to the search for AGI. AGI, once discovered, could completely change the game.
The need for renewable energy
Narrow AI is trained to perform specific tasks, such as Natural Language Understanding (NLU), playing chess or identifying spam. In contrast, AGI will resemble a genius polymath capable of tackling challenges across various fields. AGI is expected to revolutionise sectors like healthcare, scientific research, and automation.
Increased productivity and efficiency are anticipated, but substantial job displacement may also occur. Naturally, AGI’s potential applications come with Environmental, Social and Governance (ESG) considerations. These considerations will be addressed in a separate article.
The discovery of AGI is currently constrained by three factors: processing power, access to information, and energy supply. AGI requires vast computational resources, such as chips, to mimic human-level intelligence. These resources far surpass the capabilities of today’s AI models. Additionally, training AGI requires access to extensive, high-quality datasets.
Human-created data is finite, so AIs and AGIs may autonomously create new knowledge for self-teaching. This raises significant questions about privacy, control, and the implications of AI generating its own data. None of this is possible without sufficient energy.
Training AI demands for energy
Statistics from OpenAI indicate that training ChatGPT-3 took 34 days and consumed approximately 1,248 MWh of electricity. By comparison, ChatGPT-4 required 100 days of training, consuming an estimated 50 GWh of electricity. This energy is roughly enough to power 17,000 South African homes for one year.
Typically, current AI models are trained, reviewed, and then published for public use, resulting in a knowledge cut-off date. When given an instruction, an AI generates responses based on its training.
The largest computational and energy demands occur during training. However, processing power and energy are still required when instructions are fed to the AI. AI systems are now evolving towards an “always learning” phase, continuously updating their knowledge in real-time. This shift increases and sustains energy demand, requiring constant computational processing.
Continuous learning involves ongoing data analysis, model adjustments, and real-time decision-making, all of which consume significant energy. When AGI emerges, this effect will be amplified. AGI’s human-like abilities will demand complex computations and perpetual data processing, further escalating energy needs.
Regions or nations unable to meet these energy demands may struggle to access AI and AGI technology. This could restrict participation in the AI revolution to areas with stable, affordable, and increasingly green energy supplies. Such limitations could reinforce global and regional inequalities if not addressed equitably.
South Africa’s role
As AI becomes more widespread, more data centres will likely be built in South Africa to train and host AI software locally. These data centres must be robust, secure, and always operational to reduce latency and ensure competitiveness. To benefit from the AI revolution, South Africa must adopt a proactive approach.
The first critical step is prioritising investment in renewable energy to prevent stalling AI development. Powering data centres with renewable energy and storage solutions like green hydrogen fuel cells or batteries has many benefits. These include reducing greenhouse gas emissions and lowering operational costs, especially given Eskom’s proposed tariff increases.
Renewable energy also alleviates strain on the national grid and reduces energy losses when facilities are co-located with data centres. Additionally, renewable energy enhances competitiveness in the global economy, where energy sources influence foreign investment. Excess power can be fed back into the grid, contributing to energy security and earning revenue for data centres.
South Africa is already encouraging energy self-sufficiency at data centres. The National Data and Cloud Policy, published on 31 May 2024, prioritizes data centres with self-sufficient energy and water sources.
Driving renewable energy investment
To drive renewable energy investment, South Africa has two options: the carrot (incentives) or the stick (penalties). Incentives, such as tax breaks and reduced tariffs on renewable energy hardware, encourage adoption and innovation. These measures foster growth in the renewable energy sector, creating green jobs and sustainable data centres.
However, recent government actions may discourage renewable energy investment. In 2023, the government chose not to renew solar panel tax incentives and imposed a 10% tariff on imported solar panels. South Africa should consider examples from other African 
nations that successfully incentivize renewable energy. Kenya exempts solar and wind energy equipment from VAT, while Ghana and Botswana offer similar benefits.
Embracing renewable energy investment is essential for South Africa to capitalize on AI and AGI opportunities. Sustainable power for data centres can meet the energy demands of advancing AI technologies. Proactive action today will enable South Africa to harness AGI for economic growth and social development. This approach paves the way for a more equitable and sustainable future.
Mandy Hattingh | Legal Practitioner | Mining and Environmental Law | NSDV | mail me |
Related FAQs: Powering the AGI revolution
Q: What is the role of power in powering the AGI revolution?
A: Power is crucial in powering the AGI revolution as it requires vast amounts of computational power and energy to function effectively and perform a wide range of tasks.
Q: How does the computational power and energy demand evolve with AGI?
A: The computational power and energy demand therefore increase significantly as AGI systems become more advanced and capable of handling even more complex tasks.
Q: What implications does the AGI revolution have for energy production?
A: The AGI revolution could lead to increased energy production capabilities, allowing systems to power back into the grid and utilise excess power back to meet its growing energy requirements.
Q: Why is it important to ensure that AGI is developed sustainably?
A: It is important to ensure that AGI is developed sustainably to manage its vast energy needs and to avoid overwhelming the existing power infrastructure, which may not be prepared for such high demands.
Q: How can South Africa contribute to powering the AGI revolution?
A: South Africa could play a significant role in the AGI revolution by leveraging its resources and infrastructure to support the development of AI, especially if AGI systems are to be constructed in South Africa.
Q: What are the potential economic impacts of automation through AGI?
A: Automation through AGI could lead to increased productivity and efficiency in various sectors, but it also raises concerns about job displacement and the need for workforce retraining.
Q: What strategies should South Africa consider in adopting AGI technologies?
A: The South African government recently elected should consider adopting policies that promote research and development in AI, as well as investing in the necessary power and energy infrastructure to support these technologies.
Q: What are the challenges associated with the computational power and energy demands of AGI?
A: The challenges include meeting the largest computational power and energy demands while ensuring sustainability and minimising environmental impact during the development of AGI systems.
