The discourse surrounding Artificial Intelligence (AI) is no longer confined to the corridors of the technology industry.
From offices and educational institutions to businesses and newsrooms, its influence is pervasive. Tasks that previously required hours of human effort are now being completed by AI in a matter of seconds.
However, as AI grows increasingly powerful, a critical question emerges – if most of the new information on internet is generated by AI in the future, where will the next generation of models find new and reliable data to learn from?
How AI truly learns
Current AI models do not possess innate knowledge; they learn exclusively from information created by humans. By analysing years of books, research papers, news reports, blogs, online discussions, and programming code, AI learns the nuances of language, information, and logic.
In essence, today’s AI stands on the foundation of human-created knowledge. The more diverse and high-quality the information available on the internet, the greater the learning opportunities for AI.
Consequently, human-generated data is not merely a resource for AI; it is the very basis of its existence.
Concern of “model collapse”
In recent years, researchers have identified a potential risk known as “model collapse.”
This theory suggests that if AI consistently learns from AI-generated content rather than human-created data, the diversity of information will gradually diminish.
This creates a risk where misinformation or biases could be passed down and amplified from one generation of AI to the next.
According to researchers, the real world is immensely complex, containing subtle aspects of human experience, culture, language, and behaviour. If AI primarily learns from its own output, many of these real-world nuances could vanish from its training datasets.
Experts note, however, that the reality is more complex than laboratory experiments. Major technology organisations typically collect data from various sources and regularly update their training methods.
Therefore, while the concern is significant, it is not yet viewed as an inevitable outcome.
Changing face of the internet
The volume of AI-driven content has increased significantly over the last few years. Many organisations now utilise AI to generate advertising copy, product descriptions, documentation, summaries, and even software code.
The presence of AI-generated content is also growing across social media platforms, blogs, and websites.
This shift is creating a new reality where the dominance of human-written information on the internet is fading. Technology analysts believe that the future of AI development will depend heavily on the balance between human-created and AI-generated content.
Rising value of the human touch
As a result of this shift, major technology companies are becoming increasingly interested in collecting high-quality human-made data. News reports, research, expert opinions, and discussions based on real-world experiences are now considered extremely valuable for AI training.
In many instances, new datasets are being developed with the assistance of doctors, lawyers, engineers, and subject-matter experts. This is because authentic knowledge, experience, and analysis are still most reliably sourced from humans.
This human contribution remains indispensable, particularly in fields where there is no single “correct” answer.
The Bangladesh context
A vital aspect of this discussion involves local languages and knowledge. While global AI models are trained on vast amounts of English data, there remains a significant shortage of information in the Bengali language.
Digital content regarding Bangladesh’s agriculture, local businesses, law, education, regional culture, and social realities is relatively limited.
The more Bangladeshis publish reliable data, research, and analysis online, the better future AI systems will be able to understand the country’s context.
Preserving knowledge in local languages is therefore not just a cultural duty, but a vital component of the future digital economy.
It was once feared that AI would replace humans, but reality suggests that regardless of how advanced AI becomes, humans remain the primary source of new knowledge. Humans gain new experiences, conduct research, ask new questions, and seek new answers.
AI can analyse and utilise that knowledge, but the original source remains human. In this era of rapid technological change, the value of human writing, research, experience, and creativity is not diminishing; in many respects, it is becoming more vital than ever.
