Canadian startup Cohere, which operates in the artificial intelligence industry, focuses on the development of tailored AI models for enterprise users.
The mentioned startup, which was last valued at $5.5 billion, does not intend to guide as a priority on building a larger foundation machine intelligence models.
The development of the strategy laid out in the letter of Cohere to investors is because many companies are still trying to figure out how to incorporate large language models into their daily work two years after the debut of ChatGPT from OpenAI.
Nick Frosst, co-founder of the startup, said during a conversation with media representatives that customers do not just want bigger artificial intelligence models. They need digital products of the appropriate category that are really built for their specific use cases.
At the same time, Cohere intends to continue developing foundation AI models. However, the startup plans to focus on other training techniques to improve the mentioned models, and not just increase their size. Selling of the Application Programming Interface (API) will remain a small part of Cohere’s offering. It is worth noting that in this case, the API for artificial intelligence models developed by the startup is meant. At the same time, Cohere’s main focus will be on customized AI models deployment.
The race to build bigger and better artificial intelligence models has fueled a kind of investment boom from startups to companies in the Big Tech category. Industry players such as OpenAI and Anthropic have raised billions of dollars to fund the development of advanced digital intelligence systems.
Cohere, which has headquarters in Toronto and San Francisco, has received investments worth more than $900 million from several companies, including Nvidia and Cisco.
The startup positions itself as a player in the artificial intelligence industry, which specializes in enterprise AI and is independent of cloud service providers. Cohere works directly with customers such as Oracle and Fujitsu to tailor models for specific needs.
In the artificial intelligence industry, which has made significant progress by increasing computing power and model size, there is currently a decrease in return from bigger functional systems. AI labs are facing delays in training a new generation of large language models. Ilya Sutskever, co-founder of Safe Superintelligence and OpenAI, stated that results from scaling up pre-training have plateaued.
Nick Frosst said that simply increasing the size of an artificial intelligence model does not always provide better results. It was also noted that the focus on customization could allow Cohere to increase capital efficiency and reduce the need for computing power.
The startup does not seek to develop artificial general intelligence (AGI). At the same time, the corresponding goal is typical for many other players in the digital intelligence industry, such as OpenAI.
Nick Frosst stated that the startup is going to collaborate with an enterprise to figure out how to make an artificial intelligence model perfect for a concrete use case, tailor this functional system for specific needs, and get to production, without expecting that the AGI future will come next year.
It’s worth noting that during the period of active development of artificial intelligence, the issue of cybersecurity has become more relevant. Scammers also have access to AI technologies, which is why their activities have become more sophisticated. To counteract the corresponding threat in the cyber environment, personal awareness of users is important. For example, a query in an Internet search engine, such as how to know if my camera is hacked, will allow anyone to get information about signs of unauthorized access to the device.
