Generative artificial intelligence startup Cohere Inc. today announced that it has raised $270 million in funding from a group of prominent tech investors.

The investment, a Series C round, included the participation of Nvidia Corp., Oracle Corp. and Salesforce.com Inc.’s venture capital arm. Publicly traded cybersecurity provider SentinelOne Inc. contributed as well. The four companies were joined by more than a half-dozen other backers, including Inovia Capital, which led the round.

Toronto-based Cohere provides generative AI models through cloud-hosted application programming interfaces. It’s led by Chief Executive Officer Aidan Gomez, a former AI researcher at Google LLC. During his time at the search giant, Gomez co-authored the academic paper that first presented the concept of Transformers, one of the most significant machine learning breakthroughs in recent memory.

Transformers are a type of neural network optimized to process text. They can understand and generate prose with significantly higher accuracy than earlier models, as well as perform other tasks such as writing code. The technology powers many of the industry’s most advanced large language models, including OpenAI LLC’s GPT-4.

To determine the meaning of a sentence, a neural network analyzes the surrounding text for contextual clues. Prior to the introduction of Transformers, neural networks lost a significant portion of the collected contextual clues due to an issue known as the vanishing gradient problem. That limited their accuracy.

Transformers, in contrast, can accurately retain large amounts of contextual information. Moreover, they prioritize each piece of contextual data based on its significance. That allows them to interpret text and code more reliably than earlier models.

“It’s still super early in the technology’s trajectory,” Gomez said in an interview on SiliconANGLE’s theCUBE a few weeks ago (below). “I think in the next year, we’re gonna see some major changes. You’ve already seen the very first hints of that with stuff like Bing Chat, where you augment these dialog models with an external knowledge base. Now the models can be kept up to the date to the millisecond because they can search the web.”

Gomez launched Cohere with co-founders Nick Frosst and Ivan Zhang in 2019, two years after the publication of the Transformers paper. Prior to today’s funding round, the startup raised about $183 million from investors. 

Cohere offers a collection of generative AI models headlined by a neural network called Command. It’s optimized to generate text based on user-provided prompts. Command can produce marketing copy such as product descriptions, as well as summarize documents and power customer support chatbots.

Cohere also offers a number of more specialized models. One set of models is geared toward text summarization tasks, while another is designed to power website search engines. It also offers neural networks optimized to turn words into embeddings, a data format that AI applications can process more easily than raw text.

For companies with advanced requirements, the startup provides the ability to fine-tune its models by training them on custom datasets. Cohere also enables customers to chose among multiple deployment options. According to the startup, its models can be deployed on all the major public clouds and in companies’ internal technology environments.

“These large language models can be deployed all over the place, into every single industry sector,” Gomez stated during his recent appearance on theCUBE. “Language is everywhere.”

Cohere’s latest funding round reportedly values it at $2.2 billion. The investment comes a few weeks after Anthropic, another startup developing advanced generative AI models, closed a $450 million investment led by Spark Capital. The raise reportedly valued the startup at $4.1 billion.

Image: Unsplash