AI startup Probably raises $9 million to combat model errors
TechCrunch AI
Probably aims to prevent hallucinations and factual errors in LLMs by building a rigorous error-catch system, utilizing a deterministic validator to refine context and reduce ambiguity. This approach enables smaller AI models, reducing costs and token usage, and can be applied to various precision-sensitive use cases.
💡 Key Takeaways
- Probably is trying to build a more rigorous way to catch errors in LLMs
- The company's goal is to prevent hallucinations and simple factual errors from reaching users
- Probably's data science tool uses a deterministic validator to refine context and reduce ambiguity
Get smarter about AI
The sharpest AI news, curated daily. Delivered free to your inbox.
Related Articles
AI company Anthropic suspends access to models in India due to US government directive
Anthropic has suspended access to its newest AI models following a US government directive, raising...
AI Hardware Startup Plaud Hits 2M Device Sales, $100M Revenue
AI hardware startup Plaud is gaining traction with its AI-powered notetakers, targeting professiona...
AI Lab Forced to Pull Models Offline by Trump Administration
The US government has used an obscure export control directive to force AI lab Anthropic to pull it...