Conduit Collects 10,000-Hour Neuro-Language Dataset For Thought-To-Text Models
San Francisco startup Conduit says it has collected roughly 10,000 hours of non-invasive neural recordings from thousands of participants over the past six months to build a neuro-language dataset. The company captures two-hour conversational sessions with tight alignment of text, audio, and neural signals, prioritizing engagement to maximize usable natural language. Conduit plans to train thought-to-text models to decode semantic content from brain activity seconds before speech or typing.
Key Points
- 1Collected roughly 10,000 hours of non-invasive neural recordings from thousands of unique participants.
- 2Prioritized conversational engagement to increase natural-language yield and ensure tight multimodal time alignment.
- 3Enable training thought-to-text models decoding semantic brain activity seconds before speech or typing.
Scoring Rationale
Large, targeted neuro-language dataset suggests notable research progress, but claims are company-reported and require external validation.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
