The Sobering Reality of AI: A Researcher's Perspective

Terrance Craddock, an independent AI researcher, shares his critical perspective on the current state of artificial intelligence, particularly large language models with 70 billion parameters. He argues that the excitement surrounding AI is exaggerated, as his experience reveals a mere 10% success rate in generating accurate and useful responses from these models. The remaining 90% of outputs are often irrelevant, nonsensical, or incorrect, which he believes undermines the credibility of the field. Craddock illustrates this point through a simple experiment where he asks AI models how many 'r's are in the word "strawberry." Despite the simplicity of the question, many models incorrectly assert there are two 'r's and refuse to reconsider their answers when challenged. This highlights a significant flaw in AI's reliability and raises concerns about its practical applications.

- AI models currently have a 10% success rate in providing accurate responses.

- The majority of outputs from these models are irrelevant or incorrect.

- A simple test reveals AI's inability to perform basic tasks accurately.

- The researcher emphasizes the need for a more realistic understanding of AI capabilities.

- Current AI technology may undermine the credibility of the field due to its flaws.