Receive Free Artificial Intelligence Updates
I will send myFT Daily Digest E-mail summarizing the latest information artificial intelligence News every morning.
Buy, partner, or build? That is the perpetual question when any significant new technology emerges, and one that nearly every company in the tech industry grapples with when it comes to generative artificial intelligence. is.
The focus shifted to acquisitions this week. Databricks, a private company that stores and manages large amounts of data for its customers, has paid generative AI startup MosaicML his $1.3 billion. In addition to developing open-source versions of the kinds of large-scale language models made famous by OpenAI, Mosaic is also helping companies leverage the technology within their own businesses to build and refine models using their internal data. provides tools for
News also came in this week that Thomson Reuters will pay $650 million for Casetext. Casetext, a 10-year-old AI company specializing in legal services, expanded into generative AI this year. This suddenly reveals the importance of knowing how to use large language models to power generative AI.
While any company can take advantage of OpenAI’s underlying model, Casetext got ahead of the opposition by forging relationships with AI startups early in the development of its latest system, GPT-4. According to Thomson Reuters executives, if he could learn how to operate GPT-4 to solve problems of particular value to lawyers and know the best prompts to get good results from the system. That was enough to justify the acquisition, he said.
This series of acquisitions demonstrates how rapidly the opportunity for generative AI is expanding. The months following ChatGPT’s launch last year focused on companies building large-scale language models. OpenAI-led companies included names like Anthropic, Cohere, and Inflection AI (which just announced his latest $1.3 billion funding round).
Today, the focus is shifting to all technologies and the skills required to use them. These acquisitions have really skyrocketed the value of companies that can create purpose-built models on demand or apply generative models at scale to specific industries.
But while the shift to acquisitions highlights the centrality with which many tech companies believe the technology will soon be incorporated into their offerings, the main focus for most is partnerships, and in some cases are supported by equity investments.
Earlier this month, Salesforce launched an ambitious plan to build generative AI into many of its products, doubling the amount allocated for investing in AI startups to $500 million. The news made headlines on the heels of his participation in his $450 million investment in Anthropic.
Similarly, Oracle earlier this month participated in a $270 million investment round in large model company Cohere. Oracle then pledged to work closely with the startup to incorporate the technology into other services for its own customers.
Using a partnership sealed by an equity stake seems like a logical first step. Gain instant access to critical technology. Big industry competition has always gotten the most attention, but the technology industry has long relied on such close partnerships to deploy important new technologies.
But if large language models become the future of technology, as many in the industry believe, will the biggest tech companies be happy to outsource development?
This question is of particular importance to Microsoft. A $10 billion-plus investment in a majority of OpenAI’s minority shares has been overshadowed by a wave of acquisitions and other investments over the past few weeks.
Earlier this year, Microsoft CEO Satya Nadella said in an interview with the Financial Times that the relationship is a form of “codependency” and that OpenAI is dependent on Microsoft and vice versa. said there is. He said AI companies need supercomputers built and tuned by Microsoft to train models.
Whether that’s the formula for long-term stability in key technology partnerships, as Nadella argues, is another question. This makes Microsoft dependent on others for its core technology. There is no guarantee that OpenAI will feel the need to work with Microsoft’s Azure cloud beyond the initial contract term. It’s also not yet clear exactly how OpenAI’s own ambitions will develop, and whether it will one day become a direct competitor to Microsoft.
But with the technology industry burgeoning to target attractive new markets, such considerations tend to be postponed to another day.
richard.waters@ft.com
