Cohere Inc., a generative artificial intelligence startup that’s seen as one of the main rivals to ChatGPT creator OpenAI LP, said today it’s partnering with LivePerson Inc. to adapt its large language models for enterprise use cases.

LivePerson is a leader in conversational AI and chatbots, and the partnership will see the two companies combine their technologies to try and overcome problems around AI bias and hallucinations — which refers to incidents where AI simply fabricates a response. The companies say the initiative could have a major impact, resulting in more reliable and responsible AI that’s safe to deploy in enterprise environments.

Toronto-based Cohere stands out from among a crowd of AI startups thanks to its close ties with Alphabet Inc., the parent of Google LLC. The company was co-founded by its Chief Executive Aidan Gomez, who previously worked as a researcher at Google Brain and is one of the co-authors of a landmark 2017 academic paper that first detailed the concept of a transformer model. Transformers are a kind of neural network that have since become the basis for many key AI use cases.

Under the partnership, LivePerson intends to adapt Cohere’s LLMs in such a way that they can safely be used by enterprises to automate more business processes. LivePerson, whose models are already built using various third-party LLMs, is looking to evolve its conversational chatbots so they can decide on the “next steps” to take in any given conversation. For example, the bot could follow up with additional text or recommendations, or take other actions, such as processing a payment, that support a user’s inquiry.

If a conversational AI assistant is going to start taking actions on its own initiative, then it must be more reliable than existing implementations of the technology are. Knowing that, LivePerson said it will work with Cohere to fine-tune its models in such a way that all statements they make are both accurate and supported by facts.

In an interview with VentureBeat, Gomez explained that Cohere handles AI adaptations by combining reinforcement learning and supervised learning to emphasize what’s known as “AI explainability,” which is where the model can always explain its answers.

One of the techniques employed by Cohere to do this is “retrieval augmented generation,” which basically involves asking the model to cite its sources each time it provides a statement. So whenever the model responds to a prompt, it will cite the corpus of knowledge it was trained on, Gomez explained. This AI explainability is intended to prevent hallucinations, which are especially risky when models are deployed in enterprise applications, by enabling humans to verify answers.

Analyst Holger Mueller of Constellation Research Inc. told SiliconANGLE that partnerships of this kind between Cohere and LivePerson are likely to become much more common. “Large language model vendors are looking for fresh outlets, especially other companies that provide viable use cases for their technology,” he said. “There is a need for differentiation with such partnerships and Cohere is approaching this with a higher level of AI explainability. If it works, a lot of enterprises will be seriously interested.”

LivePerson will deploy Cohere’s LLMs internally within its own systems to test and refine the technology before extending the integration to its customer’s deployments. Ultimately, the companies hope to enable enterprises to deploy the most advanced and accurate LLMs to improve customer engagement and business outcomes.

Photo: Tara Winstead/Pexels