Testing LLMs for trust and safety


This post is by Georgian Team from Georgian


We all get a few chuckles when autocorrect gets something wrong, but there’s a lot of time-saving and face-saving value with autocorrect. But do we trust autocorrect? Yeah. We do, even with its errors. Maybe you can use ChatGPT to improve your productivity. Ask it to a cool question and maybe get a decent answer. That’s fine. After all, it’s just between you and ChatGPT. But, what if you’re a software company and you’re leveraging these technologies? You could be putting generative AI output in front of your users. On this episode of the Georgian Impact Podcast, it is time to talk about GenAI and trust. Angeline Yasodhara, an Applied Research Scientist at Georgian, is here to discuss the new world of GenAI. You’ll Hear About:
  • Differences between closed and open-source large language models (LLMs), advantages and disadvantages of each.
  • Limitations and biases inherent in LLMs due to their training on Internet data.
  • Treating LLMs as untrusted users and the need to restrict data access to minimize potential risks.
  • The continuous learning process of LLMs through reinforcement learning from human feedback.
  • Ethical issues and biases associated with LLMs, and the challenges of fostering creativity while avoiding misinformation.
  • Collaboration between AI and security teams to identify and mitigate potential risks associated with LLM applications.
The post Testing LLMs for trust and safety appeared first on Georgian.