The rapid advancement in artificial intelligence (AI) and neural networks has raised concerns about its potential misuse. Geoffrey Hinton, a pioneer in this field and a former Google employee, has voiced his apprehension about the serious societal implications of AI. The evolution of AI chatbots like ChatGPT and Google’s Bard, which were once thought to be decades away from surpassing human intelligence, is a testament to the speed of AI development.

The fear of malevolent AI does not primarily lie in the science fiction trope of robots turning against humanity due to a programming error. The real threat lies in the misuse of AI by humans, who may exploit this powerful tool for harmful purposes. As Hinton expressed in an interview with The New York Times, “It is hard to see how you can prevent the bad actors from using it for bad things.”

AI systems are, as of now, devoid of personal desires. They merely execute the commands of their human operators. The immense knowledge they possess and their potential to manipulate, misinform, and surveil is where the danger lies. Governments across the globe are already employing facial recognition technology to monitor dissidents, a capability that AI could enhance, allowing for comprehensive tracking of an individual’s every move and digital trail. AI’s absence of moral constraints could also be harnessed by governments and political groups to disseminate misinformation and propaganda at an unprecedented scale.

ChatGPT and similar systems endeavour to implement safety standards on top of their algorithms. However, malevolent actors could potentially design their own versions of these systems to perform harmful tasks, such as automating malware scams and phishing attacks. All potential harms, seemingly infinite, are ultimately born from human intent.

Hinton’s warnings are not unfounded. Even OpenAI, the creator of ChatGPT, expressed reservations about releasing its language models. Google’s hesitation to release a similar product until compelled by Microsoft could also be interpreted as concerns about the implications of generative AI. Despite Google’s responsible approach so far, Hinton has expressed unease about the company’s rapid plunge into its AI battle with Bing.

The question of how to control AI remains unanswered. Should we halt AI development, as proposed by Elon Musk in a recent open letter? Or could Nvidia’s AI guardrails be the solution? This significant question must be addressed by those who possess the requisite knowledge and insight. If not, we might find ourselves subjugated to our AI creations.