AI
GPT-4.5 (OpenAI)
In brief
OpenAI makes GPT-4.5 available, a general-purpose model emphasizing world knowledge, reduced hallucinations and a more natural answering style.
Key points
- OpenAI makes GPT-4.5 available, a generalist model focused on world knowledge.
- The emphasis is on reducing hallucinations and a more natural style.
- Answers gain in factual reliability.
- The accuracy of brand information reused by the assistant improves.
Analysis
GPT-4.5 bets on a broader knowledge base and a lower tendency to invent, which makes answers safer for the end user. For brands, this reduces the risk of erroneous information reused by the assistant, provided that reliable sources exist and are accessible.
A model less prone to hallucinations relies more on verifiable information. The presence of clear brand content, consistent across pages and well referenced, becomes a factor of accuracy in generated answers.
What to do
- Audit the consistency of brand information (address, offering, figures) across all pages and directories.
- Correct obsolete or contradictory content likely to induce erroneous AI answers.
- Test brand queries to verify the accuracy of the answers provided.
Impact
More reliable AI answers and fewer factual errors. The accuracy of brand information relayed by the assistant improves.