Introduction
Is it possible for AI systems to collapse and turn into “nothing”?
After all the hype around this technology, is this a possibility?
This is because many websites are filled with content created by artificial intelligence itself. So we will try to understand what is happening, hoping to get the answer.
In recent years, there has been a surge in excitement around text-generating systems like OpenAI’s ChatGPT, Gemini, Co-pilot and others. This excitement has led to blog posts and posts on platforms and platforms that have been created by these systems, as more and more content is produced online by AI.
Sometimes this is good as it makes publishers' jobs easier, but relying entirely or even almost entirely on AI could be devastating in the coming years if not addressed.
In addition, making such things easier will lead to an increase in misinformation or inaccurate results and thus lead to real disasters. So the development of artificial intelligence, if it is wrong, will be destructive. Just like raising a small child, it can be good and it can also be evil, the difference here is that artificial intelligence has greater destructive power.
However, many companies that produce artificial intelligence systems use text taken from the internet to train them. This can lead to a cycle in which the AI systems used to generate these texts are trained on the same content.
This is likely to lead to these AI tools becoming weak, uncreative, or not delivering the desired results. This will naturally increase the concern about the “dead internet theory,” which suggests that more of the internet is becoming automated and bot activity will increase in what could be a vicious cycle, in which will also lead to negative results related to different aspects in our lives.
Noting that, it only takes few cycles of generating contents and then training on it for these systems to produce what can be called the “nothing.”
Taking for example, if we make tests using a specific text, with repetition, the result can be disappointing and this can be on any category not only in a specific field or major, such as Engineering, healthcare, even on finance and so on.
Conclusion
The concept of AI being trained on datasets that were also generated by AI is referred to as “model collapse.” This could become increasingly prevalent as AI systems are used more online.
In which this model refers to when the machine learning models (MLM) gradually degrade due to errors coming from uncurated training on synthetic data.
After all these expectations and hopes on artificial intelligence, and in case this really happened, this can cause a big harm on different fields in our lives. Yet the artificial intelligence is such a critical technology that can improve our way of living, thus a lot of ethical studies along with hard trainings are needed to reach the expected goals.
Once again the question here would be is the AI systems will collapse? Or its danger will be more than its benefit?
*Image designed using Canva