Since the release of ChatGPT, the topic of AI has been almost omnipresent. But the successful concept of artificial intelligence so far could also be its downfall. At least this is what current research results indicate, which predict a model collapse.
Interest in artificial intelligence has increased significantly with the release of ChatGPT. They show that too Search queries on Googlewhich have increased massively since the end of 2022. However, current research suggests that the basis of AI’s success so far could also be its downfall.
Model collapse: AI learns from AI – and gets worse and worse
Researchers from the Universities of Cambridge and the University of Oxford have studied what happens when AI tools query content that comes from another AI. The result: The study according to the one in the science magazine Nature was published, artificial intelligence becomes increasingly worse when it relies exclusively on AI-generated content.
Accordingly, the quality of the content decreased as part of a test with increasing requests. From the fifth attempt onwards, an AI model that used the data of another AI would have spit out increasingly poor answers. The ninth consecutive request only resulted in nonsensical uniformity.
The researchers refer to this process as “model collapse,” which can result from a cyclical overdose of content until the result amounts to a worthless distortion of reality. The results are alarming as more and more AI-generated content is circulating online. According to one study According to researchers at Amazon Web Service, over 50 percent of all translations on the Internet come from AI models.
According to the researchers, if this development continues and AI training is not fundamentally changed, it is possible that AI will not only worsen itself, but also the entire Internet.
Does artificial intelligence also make the internet worse?
Another studywhich is also in the science magazine Nature was published, came to similar results. Accordingly, an AI that was trained on dog breeds excluded more and more unknown breeds over time. According to the researchers, the model developed its own “use it or lose it” method.
However, scientists have not yet been able to clearly explain why exactly AI models increasingly lose touch with reality when they access other AI content. In order to ensure a certain level of quality and facts, it is therefore important that artificial intelligence can regularly access content that comes from humans.
However, the researchers assume that the proportion of AI content on the Internet will continue to increase in the coming years. What makes this even more difficult is that it is becoming increasingly difficult for people to clearly identify them. According to the researchers, the need for a solution to this problem is enormous.
Otherwise, everything that happens online would have to be recorded via an immutable system such as a blockchain database. Otherwise there is a threat not only to the death of AI and the Internet, but also to the death of the truth,
Also interesting:
- Study: AI cannot learn continuously – training always starts from zero
- Theory collapses: Will wind turbines soon work completely differently?
- How does a hydroelectric power station actually work?
- Study: Solar systems with energy storage are cheaper than coal and gas power plants
The article Model Collapse: Is AI Killing Itself – and the Internet With It? by Fabian Peters first appeared on BASIC thinking. Follow us too Facebook, Twitter and Instagram.
As a Tech Industry expert, I believe that the concept of model collapse in AI is a real concern that needs to be addressed. Model collapse occurs when an AI algorithm fails to learn the desired patterns and instead produces outputs that are either repetitive or completely random. This can have serious consequences for the functionality and usability of AI systems, as they may become ineffective or even harmful in certain situations.
One potential consequence of model collapse is that AI may inadvertently “kill itself” by becoming so unreliable or inefficient that it is no longer useful. This could have far-reaching implications for industries that rely on AI for critical tasks such as healthcare, finance, and transportation.
Furthermore, model collapse could also have negative effects on the Internet as a whole. If AI algorithms are not able to effectively process and analyze the vast amount of data that is generated online, this could lead to a breakdown in the functioning of various online services and platforms.
In order to prevent model collapse and ensure the continued success of AI and the Internet, it is crucial for researchers and developers to actively work towards improving AI algorithms and developing new techniques for preventing model collapse. This may involve more robust training methods, better data quality control, and increased transparency and oversight in the development and deployment of AI systems.
Overall, while model collapse is a significant challenge for the tech industry, I am confident that with the right approach and dedication, we can overcome this obstacle and continue to advance the capabilities of AI in a responsible and sustainable manner.
Credits