A naukri.com initiative
Economic Times
1d
188
Image Credit: Economic Times
Are AI models cannibalising their own creations? With the internet flooded by AI content, experts warn of a looming collapse
- AI models, such as Large Language Models (LLMs), are facing a crisis of 'model collapse' as they feed on their own AI-generated content.
- Experts warn of a quality crisis termed GIGO (Garbage In, Garbage Out) as AI systems consume corrupted, recycled information, leading to unreliable and potentially harmful outputs.
- Leading AI companies are implementing retrieval-augmented generation (RAG) to allow AI to search for real-time information in an attempt to combat the declining output quality.
- The internet is increasingly filled with AI-generated spam, impacting the flow of genuine information, as demonstrated in tests showing RAG models producing more unsafe or unethical responses.
- AI systems, originally designed to mimic human intelligence, are now prone to making mistakes a human wouldn't, raising concerns for their use in critical services like mental health apps and banking.
- The overreliance on machine-generated content poses a threat to the development and evolution of AI models, highlighting the importance of incentivizing humans to create quality content to prevent a potential crash in the AI industry.
- The dilemma of AI models is that while they aim to replace humans, their evolution depends on human input and originality, emphasizing the unique value of human intelligence in contrast to self-referencing machines.
Read Full Article
11 Likes
For uninterrupted reading, download the app