Rajeev Chandrasekhar on why Google Gemini, ChatGPT occasionally give ‘garbage’ responses

Generative AI

Generative AI has gone beyond being the buzzword and is used by millions across the world on a daily basis. The technology, however, still remains a hit and miss when it comes to certain questions or areas.

Former IT minister Rajeev Chandrasekhar, in a post on X, suggested why sometimes ChatGPT or Google Gemini gives bizarre responses. “LLMs "bullshit content" comes from most models being trained on content/datasets that are  - to politely use the phrase - NOT quality assured. Thats why you have the embarrassing sight of billion dollar Gemini/ChatGPT many occasions spewing nonsense.”

The former minister further used a programming adage to describe why AI chatbots deliver inaccurate responses. “Garbage in, Garbage out is an old adage in programming especially if ur depending on scraping the internet,” he further said in the post.

There have been many cases of AI chatbots giving incorrect or distorted responses — a phenomenon known as AI hallucinations. Be it ChatGPT or Google Gemini, they have a tendency of providing inaccurate and irrelevant information. These chatbots generate responses that may be nonsensical, misleading, or even offensive. It depends, as Chandrasekhar, said on how it scrapes the internet. These misinterpretations occur due to factors like overfitting, training data bias, and high model complexity.

It’s still early days and companies like Google, OpenAI, Microsoft are training their models that power generative AI chatbots. All companies do give a disclaimer about the accuracy of responses generated by their chatbots.

Also read: Google rolls out Gemini Nano AI for Pixel 8 and Pixel 8a: How to download