Bain’s capital ventures invests from its $1.9 billion tenth flagship fund, announced in February 2023 following launch ChatGPTtechnology that has taken generative artificial intelligence into the mainstream.
Despite the company’s decades-long presence in the investment market and significant funds – it manages assets price $10 billion – investing in artificial intelligence corporations involves a recent set of risks and a highly competitive financing environment.
“This market is unlike any other. The field is moving very fast.” Cancer Gargpartner at BCV, who splits his time between AI infrastructure and cybersecurity investments, said in an interview.
To keep pace, BCV established a lab in Palo Alto, California, over a yr ago to support emerging researchers and incubate corporations in the artificial intelligence sector. The laboratory also hosts article readings and debates, and eminent researchers in the field are invited to present.
First, research
Garg said the only solution to stay ahead of development is to adopt a research-centric approach.
“Two years ago, we doubted whether large models could even exist and whether they would be operated reliably,” he said.
A yr ago, vector databases were seen as the answer to enterprise artificial intelligence.
And now the answer is “some combination of alignment, tuning, and vector databases.”
Small groups of individuals have to determine, “How do we make these models work efficiently? How do I prompt them in the right way?” he said.
Since launching lower than a yr ago, BCV has made 4 seed investments in the lab.
The company was the first company to launch publicly from the lab Contextual artificial intelligence, a startup that builds models that enable enterprises to access proprietary data to extend worker productivity. In mid-2023, the company raised $20 million in seed financing led by BCV.
Context co-founders Dear Keeli AND Amanpreet Singh met in Facebook Artificial Intelligence Research (FAIR). There, Kiela helped develop RAG, or search augmented generation, for large language models.
If you feed the model the right data at the right time, and then go back and look at how a human interacts with that data, you get a type of closed-loop system where it takes the right static information and then dynamically enriches it with data regarding interactions,” Garg said.
Contextual has quite a few use cases in finance, healthcare, legal, and other industries. It may also help employees answer questions corresponding to: “Who should I prioritize this week?” or “what should the message be?” Garg said.
Compared to technology like ChatGPT, context models are smaller and require less computation.
Contextual built a common system of “two small models with appropriate embeddings,” he said. “We can continually fine-tune them based on the interaction data, and then we can develop a solution that is truly accurate for that enterprise.”
“You would never use a context-aware system to generate a Shakespeare sonnet about employee perceptions,” he said.
The second company to come back out of the lab in hiding is The Prophet’s Safetyof which he is a co-founder Kamal Shah AND Vibhav Sreekanti. Last month, it raised $11 million in seed funding led by BCV. The company says it uses artificial intelligence to investigate and triage security alerts in seconds, saving security teams time.
The lab’s other investments that are still hidden are in data science and analytics.
Outside the lab
Garg said competing with established software vendors with an additional summarization layer won’t cut it in this market. “Co-pilot Atlassian it will be built by Atlassian,” he said.
That’s why BCV is focusing on investing in hyper-specialized models on the infrastructure side and inefficient processes for which there are no incumbents, Garg said.
BCV is an investor in Unstructured technologies, a company-funded company before ChatGPT became mainstream. The company recently acquired a Series B z Menlo Ventures.
Unstructured has a proprietary model that extracts text information from PDF files, emails or SaaS applications to acquire clean data into a database or data before training or tuning. It boasts 50,000 customer deployments, Garg said.
Other notable BCV portfolio corporations in AI include Move the gunsthat uses artificial intelligence to assist internal teams solve problems in a world of distant work; By the pool a core model company that goals to generate natural language-hinted code; AND Clean laboratorydata cleansing company.
Garg said that in software, more dollars are spent on services that integrate technology solutions than on products. Generative AI creates the opportunity to rework service revenues into product revenues.
“These services take data from one place, shape it in different ways and put it in another place,” he said, citing Unstructured as an example. “Models are really good at this.”