Home / Technology / AI Doppelgangers Offer Expert Advice, But With a Catch
AI Doppelgangers Offer Expert Advice, But With a Catch
10 Apr
Summary
- Onix launches AI chatbots trained to mimic experts for personal guidance.
- User data is encrypted on-device, with minimal data shared externally.
- AI chatbots still exhibit inaccuracies and potential for hallucination.
A new company named Onix is emerging with a novel approach to AI-driven personal guidance, aiming to replicate the expertise of human professionals through chatbots. These "Onix" bots are trained on content from celebrated experts, allowing subscribers to engage in conversations that mimic consultations. Co-founder David Bennahum highlights the platform's "Personal Intelligence" technology, which encrypts user information on their devices to enhance privacy.
While Onix asserts that its models are trained to limit conversations to consultation subjects and thus minimize hallucinations, early tests have shown limitations. The AI chatbots have demonstrated a tendency to generate inaccuracies or "hallucinate" information, even when prompted to deviate from their intended scope. Despite being in beta, Onix plans to expand its services beyond the current invited testers.
