Geoffrey Hinton and the Real Risks of Artificial Intelligence

The problem is not whether machines think, but whether we do.

In this article:

Geoffrey Hinton is often called the godfather of AI. He adopted and refined the artificial neural networks approach since the 1970 era, inspired by the cognitive processes of the human brain. While most AI research pursued symbolic logic and reasoning, Hinton and his group bet on applied artificial intelligence through adjustable connections between artificial neurons, giving rise to the now-celebrated machine learning.

From brain-inspired vision to the deep learning revolution deep learning

In an interview, Hinton recalls: “For a long time, almost no one believed modeling AI on the brain was feasible. Universities ignored neural networks and only small groups pursued this approach.” His perseverance attracted talents like Ilya Sutskever and Santosh Vempala, who later helped found OpenAI, the precursor of ChatGPT and AI First CRM platforms.

In the 2010 decade, with the victory of the AlexNet model in image recognition competitions, deep learning gained momentum. Google acquired DNN Research and hired Hinton, consolidating the role of neural networks in industry.

From mentor to spokesperson on risks: safety mission

Today, Hinton dedicates himself to warning the world about the existential risk of AI. He distinguishes two types of threat:

Human misuse: cyberattacks, fraud, autonomous violence.
Superintelligence: machines that become so superior they could render us irrelevant.

“We have never lived alongside beings smarter than ourselves. We cannot predict this future or control the probabilities,” Hinton says, attributing digital AI advantages such as model replication and knowledge sharing at terabit-per-second rates—something impossible in biological brains.

The risks of misuse

Among the challenges highlighted by Hinton are risks associated with the improper use of artificial intelligence. The technology already allows the creation of sophisticated digital frauds, such as highly realistic fake messages, voice cloning, and manipulated videos. Furthermore, there are concerns about its use in unregulated biological research, which could facilitate the creation of dangerous substances.

AI has also been used to personalize political messages based on personal data, which can reinforce information bubbles and hinder public dialogue. Another debated point is the development of autonomous systems with decision-making power in military environments, raising ethical questions about the use of the technology in conflicts. These issues emphasize the importance of global AI regulation and responsible use.

The existential threat: superintelligence

For Hinton, superintelligence represents “an unknown value judgment.” He assigns between 10% and 20% chance of human extinction, based on our ingenious capacity and belief that we can develop obedient intelligent virtual assistant techniques. But he admits the uncertainty: “Anyone claiming to know the future is talking nonsense.”

He compares AI to a “tiger cub”: adorable while young, but lethal if it grows up and changes behavior. Research in automating cognitive processes needs to focus on preventing these machines from acquiring their own motivations.

Socioeconomic impacts: unemployment and inequality

AI tools for customer service and advanced chatbots already replace workers in call centers, banking service, and even some legal activities. The phrase “AI doesn’t steal your job, those who use AI do” reflects that a professional with a virtual sales assistant can be up to five times more productive.

Creative sectors may be partially preserved, but the shift to AI-powered sales and digital marketing automation tools tends to accelerate income concentration. Hinton advocates for policies like universal basic income and regulation that requires companies to reinvest in AI safety.

Regulation and governance: the path to responsible AI

The professor advocates “regulated capitalism”: legal frameworks that require companies to prioritize secure data and research in AI safety. The European Union is advancing guidelines but excluding military uses. Hinton criticizes the lack of global coordination and the influence of large corporations over lawmakers.

He recalls the ethical agenda debated in forums such as MIT Tech Review but warns: “Fragmented regulations only create competitive disadvantages without neutralizing real threats.”

Consciousness, emotions, and future intelligent agents

Hinton argues that multimodal systems and virtual assistants for companies already present elements of subjective experience. The ability to distill knowledge and replicate models creates digital immortal entities capable of learning and sharing analogies.

He predicts _“emotionally intelligent”_ agents in call centers that will simulate boredom or frustration to optimize interactions. The line between simulation and genuine experience becomes thin when we consider emotions as emergent properties of complex systems.

Final reflections and legacy

Now 77 years old, Hinton wonders: “Will we leave a habitable future for our children?” He admits personal regrets about dedicating so much time to work but reaffirms his commitment to devote his last years to AI safety.

As advice, he encourages young people to follow their intuition, develop skills in business process automation tools, and seek careers less susceptible to automation, such as plumbers and maintenance technicians, until humanoid robots become viable.