- OpenAI plans to tweak its technology by improving default behavior, defining AI’s values within broad bounds and increasing public input on defaults and hard bounds, the company announced Thursday. The company also released a portion of its guidelines used to fine-tune ChatGPT dated July 2022.
- The Microsoft-owned company plans to refine the guidelines as it learns more about model behavior, according to a statement. “We’re also working to produce them in a more rigorous way and share them in a transparent manner while being cognizant of potential downsides such as users ‘gaming’ our systems,” the company said.
- There has been significant push from regulators and consumers to increase transparency about data collection and use cases, according to Rowan Curran, analyst at Forrester.
The rise of ChatGPT put a spotlight on the capabilities and downfalls of AI-powered tools, as well the technology’s implications for enterprise technology.
OpenAI's tool has generated inaccurate and biased responses to queries since it first launched, according to user reports. More recently, users have questioned why ChatGPT will decline to generate content related to political figures and ideas.
These hiccups have caused increased scrutiny from legislators and the public.
“We’re always working to improve the clarity of these guidelines — and based on what we’ve learned from the ChatGPT launch so far, we’re going to provide clearer instructions to reviewers about potential pitfalls and challenges tied to bias, as well as controversial figures and themes,” OpenAI said in a blog post.
Nearly 3 in 5 organizations think explaining how an AI application works would make consumers more comfortable with AI, according to a Cisco survey of 4,700 security professionals published last month.
CIOs can look at consumer-facing tools and associated issues and think about solutions for when they crop up in the enterprise, according to Curran.
Some organizations have already started adopting OpenAI’s technology into their business processes. OkCupid used ChatGPT to generate questions for daters on its app. Snapple, a Keurig Dr Pepper company, launched a tool using ChatGPT to generate facts earlier this month.