Topics
Latest
AI
Amazon
Image Credits:Viktor Tanasiichuk / Getty Images
Apps
Biotech & Health
Climate
Image Credits:Viktor Tanasiichuk / Getty Images
Cloud Computing
DoC
Crypto
Image Credits:Voyage AI
initiative
EVs
Fintech
Fundraising
Gadgets
Gaming
Government & Policy
Hardware
layoff
Media & Entertainment
Meta
Microsoft
privateness
Robotics
certificate
Social
outer space
Startups
TikTok
Transportation
Venture
More from TechCrunch
Events
Startup Battlefield
StrictlyVC
Podcasts
Videos
Partner Content
TechCrunch Brand Studio
Crunchboard
get through Us
AI tends to make things up . That ’s unappealing to just about anyone who uses it on a regular foundation , but specially to businesses , for which fallacious outcome could hurt the bottom line . Half of workersrespondingto a recent sight from Salesforce say they interest response from their company ’s generative AI - power organization are inaccurate .
While no technique can solve these “ hallucination , ” some can serve . For example , retrieval - augmented generation , or RAG , pair an AI modelling with a noesis base to provide the model supplemental info before it answers , serving as a sort of fact - checking chemical mechanism .
Entirebusinesseshave been built on RAG , thanks to the sky - in high spirits demand for more honest AI . Voyage AI is one of these . Founded by Stanford prof Tengyu Ma in 2023 , Voyage power tatter system for companies including Harvey , Vanta , Replit , and SK Telecom .
“ Voyage is on a mission to raise search and retrieval accuracy and efficiency in enterprise AI , ” Ma told TechCrunch in an consultation . “ Voyage solvent [ are ] tailor to specific domain , such as coding , finance , legal , and multilingual applications , and tailor to a fellowship ’s data point . ”
To spin up RAG scheme , Voyage trains AI model to convert text , documents , PDFs , and other forms of data into numerical representations foretell transmitter embeddings . Embeddings capture the meaning and relationships between different datum point in a compact data formatting , making them useful for hunting - interrelate applications , like RAG .
Voyage apply a picky type of embedding call contextual embedding , which catch not only the semantic meaning of information but the context in which the data appears . For example , devote the parole “ depository financial institution ” in the sentences “ I sat on the coin bank of the river ” and “ I posit money in the bank , ” Voyage ’s embed example would generate different vectors for each instance of “ bank ” — reflecting the dissimilar meaning implied by the linguistic context .
ocean trip host and certify its example for on - premises , private cloud , or public swarm use , and fine - melodic line its models for clients that opt to pay for this help . The troupe is n’t singular in that regard — OpenAI , too , has a tailorable plant service — but Ma take that Voyage ’s models extradite better performance at abject costs .
Join us at TechCrunch Sessions: AI
Exhibit at TechCrunch Sessions: AI
“ In RAG , give a question or query , we first retrieve relevant info from an unstructured cognition base — like a librarian searching ledger from a program library , ” he explained . “ formal RAG method often struggle with circumstance loss during information encoding , leading to failures in retrieve relevant information . Voyage ’s embedding modelling have best - in - class recovery accuracy , which translates to the terminate - to - end response tone of RAG systems . ”
Lending weight to those bold claim is anendorsementfrom OpenAI chief rival Anthropic ; an Anthropic support doc describes Voyage ’s models as “ state of the art . ”
“ Voyage ’s coming uses vector embeddings trained on the fellowship ’s data to provide context - cognisant retrieval , ” Ma tell , “ which significantly improves retrieval accuracy . ”
Ma says that Palo Alto - based Voyage has just over 250 customer . He decline to respond questions about revenue .
In September , Voyage , which has around a dozen employees , closed a $ 20 million Series A turn go by CRV with involution from Wing VC , Conviction , Snowflake , and Databricks . Ma says that the cash infusion , which brings Voyage ’s aggregate raised to $ 28 million , will support the launching of new embedding models and will let the company double its size of it .
