Knowing AI’s limits before it’s too late

If it sounds like science fiction, it is because it is close to it. Consider two events that made headlines in the last month regarding Artificial Intelligence.

The first was an incident in June when AI agents, actually bots, autonomously hacked a company ‘Hugging Face’ without a real-world human prompt and proceeded to make changes to its site. Evidently, these manoeuvres were even hidden from plain sight.

Another headline is the almost dystopian declaration by an employee of another AI company who publicly stated about AI that there was a “10 per cent chance it could kill all humans within the next decade,” according to the BBC, a statement corroborated by multiple media organisations.

Admittedly, more details of the exact scenario of this extinction were not elaborated upon, but if there is one wake-up call, it must be this.

This appears to be a pivotal moment in our understanding of AI, as the conversation seems to have shifted from whether AI is a risk to society to how much of a risk it poses. This is a large shift and will have repercussions at every level.

For a digital novice, the questions that arise from these emerging stories are not technical but social and focused on human behaviour. Should there be limits to the use of digital technology? If so, what are they and who should decide on them?

Apart from the ongoing debate on data centres, which consume large amounts of precious water and power, the debate is also about the self-defeating role of AI, which threatens existing jobs without offering new opportunities.

And perhaps most importantly for an educationist, the question is how these challenges and opportunities should be communicated to the younger generations.

A tentative answer to these questions is to first share information on what AI does not do: it does not, for example, create new information but rather builds on the seemingly endless information that is already in its system. It does not have common sense or empathy.

If AI works on data being fed to it, it will also duplicate the prejudices and biases already existing in society. It cannot, therefore, change society itself in any meaningful way.

There is also the danger of hallucinations. This is when AI models churn out plausible, if not authentic or accurate, information. Most of it depends on the prompts given.

A famous example is an AI platform confidently giving information on the first person who crossed the English Channel on foot when asked to do so.

There are a number of reasons for the increasing dependence and reliance on AI today. These include the fact that the digital platforms sound empathetic, have endless patience, and speak positively to us at all times.

The dangers of over-dependence on such technology have been spoken of since its inception over three years ago. Increased use of AI results in the loss of individual critical thinking, ethical discussions, and a unique creative voice. It provides a one-fix solution for all our problems, without taking into account individual, cultural, or economic perspectives. There is also the issue of data privacy, which essentially means that everything we feed into these platforms, including sensitive data on medical history, financial details, and other personal situations, will be in the public domain forever.

While we work with the technology that is available to us, knowing its limitations, as well as its dangers, is essential. There are ways in which AI can be used strategically and ethically.

As it is going to be such an integral part of the lives of youngsters as they enter schools and colleges, education in AI should become a standard part of the curriculum to help students navigate this complex, emerging platform.