In the era of rapidly advancing AI, one challenge that requires attention is the transparency and trustworthiness in generative AI. IBM researchers are working on tools to make generative AI more explainable and reliable. The complexity lies in the fact that language models (LLMs) are not very good at detecting the content they generate or tracing the source of a tuned model.

To address this challenge, IBM and Harvard researchers have developed the GLTR AI-text detector, which analyzes the statistical relationships among words or looks for tell-tale signs of generated text. Additionally, IBM has created a tool called RADAR that can identify AI-generated text that has been paraphrased to deceive the detectors. Measures have also been implemented to restrict employee access to third-party models, preventing leaks of client data.

Another challenge in the world of generative AI is identifying the origin of models that produce the text, a field known as attribution. IBM researchers have developed a matching pairs classifier that compares responses and reveals related models. This automated AI attribution using machine learning helps researchers trace the base of a model and understand its behavior.

IBM is committed to creating explainable and trustworthy AI. They have introduced the AI Fairness 360 toolkit, which incorporates bias mitigation and explainability in their products. With the release of Watsonx.governance, IBM is enhancing transparency in AI workflows and making transparency tools accessible to everyone.

Check out the IBM Article. All Credit For This Research Goes To the Researchers on This Project. Also, don’t forget to join our 27k+ ML SubReddit, 40k+ Facebook Community, Discord Channel, and Email Newsletter, where we share the latest AI research news, cool AI projects, and more.

Astha Kumari is a consulting intern at MarktechPost. She is currently pursuing a dual degree course in the Department of Chemical Engineering from the Indian Institute of Technology (IIT), Kharagpur. She is an enthusiast in machine learning and artificial intelligence and is keen on exploring their real-life applications in various fields.