The topic of AI has exploded this year. It’s impacted all of us in some way, whether it’s the subject du jour at the family dinner table, academic debate, or company boardroom. For Cohere, it provided mainstream market understanding of their work almost overnight and put them at the center of the latest wave of development and conversation.
The AI Talent Web
The story of our initial investment in Cohere, a leading startup in the growing field of generative artificial intelligence, is a story made possible by the deeply interconnected AI network.
At Index, an early curiosity about self-driving cars allowed us to forge some important links here. We initially invested in Chris Urmson’s self-driving car company Aurora in 2018, which introduced us to the likes of Alex Wang at Scale AI, and Pieter Abbeel at Covariant. This allowed us to not only learn a ton about how the technology was evolving but, more importantly, allowed us to connect with the people behind the technology.
Cohere: Top Talent Meets Product Focus
Our first investment in Cohere came in 2021, when we led the company’s Series A after Pieter Abbeel intro’d us to co-founder Aidan Gomez. The landscape at the time was much different – there was no ChatGPT, and no broad understanding of the power of generative AI.
For us, a couple of things made Cohere attractive at that point in time. The first was the background and existing accomplishments of the team. While at Google Brain, Gomez co-authored the famous 2017 paper (titled “Attention Is All You Need”) that introduced “transformers” to the world. Co-founders Nick Frosst and Ivan Zhang added to the team’s strength with a combination of research and applied experience. Nick Frosst was a grad student at the University of Toronto under Geoffrey Hinton – who is sometimes called the godfather of AI – and was the first employee in Hinton’s Google Brain lab. The third co-founder, Ivan Zhang, had also attended the University of Toronto and worked at other startups. The team had both learned from the best and were pioneers in their own right when it came to the power of transformers.
The second enticing element was the team’s focus on building a product company, not a research company. They wanted to get AI models into the hands of customers as soon as possible, and we found that powering belief to be incredibly important.
Other entrepreneurs had seen the broad potential of transformers, especially after OpenAI released GPT-3, in 2020. But at the time, a lot of companies were talking about artificial general intelligence, or AGI, and the idea of creating an all-knowing and all-capable machine that thinks like a human. The Cohere founders were much more product-focused. They wanted to build an application programming interface, or API, that would let their customers use AI models to help businesses build internal knowledge bases, augment the workflow of their employees, or build customer support chatbots. They were very much not trying to boil the ocean, and their focus was on creating something that could be delivered to customers in the near-term.
Building in a Hype Era
The wave of excitement that ChatGPT created when it launched last November had a powerful impact on the space and was a major moment for all of us. It enabled Cohere to raise more money, capture the interest of customers, and recruit top talent. A broader group of people were able to more easily imagine the myriad applications for language models. Cohere, which was a little Canadian company doing some cool stuff in AI, was suddenly centerstage in the technology industry. They now have established products, close to 300 employees, and more processes and structure. They’ve come a long way since our initial investment back in 2021.
"The opportunity to partner with exceptional entrepreneurs as they build their business is the greatest joy of my role as a venture capitalist."
— Mike Volpi, Index Ventures
Helping them grow in this competitive and dynamic environment has been a privilege. The Cohere crew are eager to learn and partner, soaking up information like sponges. I talk to Aidan every day or two about strategic partnerships, management choices, hiring decisions, or some unpredictable hiccup. The opportunity to partner with exceptional entrepreneurs as they build their businesses is the greatest joy of my role as a venture capitalist – and whenever you find amazing founders like Aidan, Nick, and Ivan, you just want to put your arms around them and help out in any way you can.
The Cohere Difference
Today, it’s no secret there are a number of major players and open-source projects building large language models. Cohere differentiates itself on a couple of core dimensions:
- Clear focus on serving businesses: Cohere remains laser-focused on building models for businesses. Instead of making a chatbot, they help others build chatbots, the tools for searching knowledge bases or coding assistants. Along with that comes the need to build proprietary models for each customer, with each model trained on private data for a targeted purpose. If Boeing is your customer, and they want a chatbot to help technicians figure out what’s wrong with an engine, they don’t want the data used to train their model to become public.
- Ease of use driven by GTM approach: Some companies produce open-source models that customers download and then fine-tune. But that process requires customers to have a lot of technical wherewithal. Instead, Cohere hands their customers a pre-trained model with the tools that allow them to fine-tune their models with proprietary data.
AI is on a hype cycle now and everything seems to be going to the moon. It’s exactly this type of market backdrop that separates the winners from the losers. Ultimately, a well-conceived strategy and technological strength will be the critical elements that determine the outcomes in this business. One thing is for sure, it’s an exciting time, and we are privileged to partner with entrepreneurs building technology that will shape how we live and work for the years to come.
Published — Nov. 13, 2023