AI Chatbots Are About to Get Personalities Soon
A key challenge for leading AI companies like OpenAI, Google, and Anthropic is imbuing their AI models with personality. These companies have dedicated teams working on “model behavior,” a growing field focused on shaping how AI systems respond and interact with users.
The distinct approaches each company takes to model behavior will likely play a significant role in determining who leads the rapidly expanding AI market. Making these models more responsive and useful to a global audience of millions of individuals and businesses is critical.
These AI teams are working to introduce qualities like “kindness” and “fun” in their models while simultaneously implementing safeguards to prevent harm and ensure nuanced communication.
Google, for instance, aims for its Gemini model to offer diverse perspectives only when specifically requested, while OpenAI’s ChatGPT is programmed to maintain an objective stance. “It is a slippery slope to let a model try to actively change a user’s mind,” Joanne Jang, OpenAI’s head of product model behavior, explained to the Financial Times. She further added, “How we define objective is just a really hard problem on its own. The model should not have opinions, but it is an ongoing science as to how it manifests.”
Anthropic takes a different approach, arguing that, like humans, AI models cannot be entirely objective. “I would rather be very clear that these models aren’t neutral arbiters,” said Amanda Askell, head of character training at Anthropic. Instead, their model, Claude, is designed to express its beliefs transparently while remaining open to other viewpoints.
Since Claude 3’s release in March, Anthropic has implemented specific “character training.” This process, which follows the initial training phase involving human labeling, “transforms the model from a predictive text tool into an AI assistant,” according to the company.
However, we still have a long way to go before we see any real personality in these models. It should be interesting to see how model behavior turns out for these AI chatbots in the future
ALSO READ: