Have We Misunderstood AI?

The advent of artificial intelligence (AI) has pushed the boundaries of what mankind can achieve using the powers of technology. And we are quite impressed. The victory of Google DeepMind’s AlphaGo over the world Go champion was a landmark moment in the history of AI.

The game of Go is incredibly complex, with a vast number of possible moves, far exceeding that of chess. In 2016, AlphaGo defeated Lee Sedol, a world-renowned Go champion, in a series of matches. AlphaGo’s victory demonstrated the potential of AI to tackle complex problems and inspired further research in the field.

The tipping point for AI came with the introduction of ChatGPT, a powerful chatbot that can communicate in plain English. It can produce paragraphs of written content with the same ease as writing blocks and blocks of workable code. One code programmer can now do the work of what used to be the work of many programmers, with the help of generative AI technology such as ChatGPT.

AI has also made possible several breakthroughs in science and innovation. Using Google DeepMind’s AlphaFold 2 model, scientists were able to predict structures of 200 million proteins. To put it into perspective, determining the 3D structure of a single protein, takes typically a year or more. Using this vast data, scientists around the world are able to work on making newer discoveries in developing new medicines, fighting antibiotic resistance, and tackling plastic pollution.

Concerns Around AI

Though AI has made leaps and bounds of progress in the past few years, it has limitations and concerns. On the outset, AI’s increased adoption could replace the equivalent of 300 million full-time jobs, revealed a report by investment bank Goldman Sachs. Another concern of increased use of AI is around wages. Due to the introduction of GPS technology and ride-hailing platforms, experienced drivers had to undergo large wage cuts. A similar pattern may emerge from excessive use of generative AI on a wide set of creative tasks.

AI also suffers from a bias problem. Some AI models have been known to show biases against people of color, women, people with disabilities, the LGBTQ community or other marginalized groups of people. This not only hinders people’s ability to participate in the economy and society, but also fosters mistrust around AI’s potential. Biases found in applicant tracking systems were found to favor men over women when analyzing their resumes. This biased approach, led to Amazon discontinuing using a hiring algorithm. AI algorithm biases can turn ugly when similar faults are found in AI models used in healthcare and predictive policing tools.

A growing body of research also points to mounting negative effects on the environment due to the increased adoption and use of AI. Data centers housing AI servers produce electronic waste, often containing hazardous substances, like mercury and lead. AI servers also consume excessive amounts of water, to cool its electrical components. One estimate finds that globally AI-related infrastructure may soon consume six times more water than Denmark – a country of 6 million. AI models also require a significant amount of energy to operate. IEA reports that a request made through ChatGPT – an AI-based virtual assistant – consumes 10 times the electricity of a Google Search. 

Way Ahead

AI has a multi-faceted role to play in our society and its importance to our future existence and growth cannot be ruled out. But exploring and tapping into AI’s vast potential, will need both an apt leadership and a robust governance system.

In terms of loss of livelihoods, governments need to act now. Ensuring common people understand its benefits and how the technology can be used in easing burdens around everyday lives, should be the first starting point. Several countries such as the US, Canada, South Korea, China, Estonia, and England have actively incorporated AI education into their school curriculum. From 2025, Beijing’s school going children are receiving a minimum of eight hours of AI-focused lessons each academic year. Estonia’s government has partnered with OpenAI to introduce AI-driven education tools to secondary school pupils and teachers.

AI governance systems like human-in-the-loop and counterfactual fairness identify bias, ensures human intervention in an AI model’s decision-making approach. This will add an additional layer of quality assurance, ensuring results are not biased towards sensitive issues such as gender, age or race.

The UN Environment Programme (UNEP) recommends five ways we can rein in the impacts of AI on the environment:

• First, countries should work towards standardizing procedures to measure the environmental impacts of AI.
• Second, governments can work with UNEP to develop regulations around mandatory disclosure of the environmental impacts of a company’s AI products and services.
• Third, companies could also look at ways to reduce the carbon footprint of their AI products and services, either by using renewables for their energy needs or reusing components where feasible.
• Fourth, greening the data centers is another critical area which companies should also ensure.
• Finally, countries should incorporate AI-related policies into their broader environmental goals. 

One single theme of “working together” emerges as the driving factor of both understanding and managing emerging technologies such as AI. Horasis is playing its role in this endeavor by organizing the 10th Horasis Global Meeting, scheduled to take place in São Paulo, Brazil, between 7 to 10 October 2025. The meeting will bring together leaders from the government, businesses, and the civil society, who will discuss sustainable and inclusive pathways to our future.