Companies Are Developing Industry-Specific AI Models --- Bayer and others are working with Microsoft to tailor tech to their sectors

"Bayer is known for selling seeds. Now it sells seeds and artificial intelligence.

Microsoft last week announced it is working with the German pharmaceutical-and-agricultural group and other companies on specialized AI models fine-tuned on industry-specific data. The companies can now list and monetize those models on Microsoft's online model catalog.

For Bayer, that means an AI model fine-tuned with its data and designed to provide answers on agronomy and crop protection is available to be licensed by its distributors, new agricultural-technology startups and even potentially competitors. The model can answer questions about ingredients in an insecticide or whether a product can be applied to cotton, for example.

"A lot of folks have the same pain points that we have," said Sachi Desai, Bayer's VP of AI Go to Market and Partnerships. "There's a lot of ways to not only amortize our own cost by allowing others to collaborate off the same platforms or build on it, but also to uplevel the outcomes for our customers."

Bayer last week cut its full-year earnings target after a tough agricultural market hit its crop-science division, and said it heads into next year with a muted outlook and likely declining earnings.

Microsoft expects this new approach, built on its Phi family of small language models and preloaded with industry knowledge, will accelerate enterprise generative AI adoption, a yearslong effort built on the understanding that off-the-shelf AI models often don't cut it for business needs.

Companies now find it critical to augment today's general models with more industry-specific or business-specific data if they are going to be useful.

Model makers in turn have quickly staffed up teams to work with customers and have identified strategic partners to build out company and industry-specific artificial-intelligence tools.

For Microsoft, handing over the reins to enterprise partners is a question of knowing its limits. The tech giant said it has taken steps to develop more industry-focused models by leveraging public and even synthetic data sets -- but it said it could only go so far without rich data sets owned by companies like Bayer.

"In a lot of cases, the AI scenarios and the accuracy of it is only as good as the data you have," said Satish Thomas, corporate vice president of business and industry solutions at Microsoft.

Industrial automation provider Rockwell Automation, compliance tech provider Saifr -- which is part of Fidelity -- manufacturing analytics provider Sight Machine, automotive software company Cerence and Siemens Digital Industries Software, a unit of Siemens, also launched their industry models last week. Most will be available on Microsoft's Azure Catalog, although in Siemens's case, its model will be available in the Azure Marketplace or directly through Siemens.

Saifr, a startup born of Fidelity Labs, said it is releasing four models targeting financial-services compliance, including one that can be used to help suggest compliant language for content such as marketing materials and emails. Chief Executive Vall Herard said Saifr leveraged Fidelity's data and expertise to build the models.

"Fidelity has a history of compliance rigor and out of that process, you have human-reviewed data that simply doesn't exist in the amount that you need anywhere else. You cannot go out on the internet and find this data. You can't buy it," he said.

On the surface, the strategy here almost seems counterintuitive: Since the first day of ChatGPT's public release, corporate executives have been determined to ensure no proprietary business information makes its way into the training data of a model that could then be used by competitors or the general public.

But Saifr said none of the data was sensitive client information and instead was in the form of subject-matter expertise.

Herard said he hopes models made available on Azure Catalog will be used by financial-services companies and the software companies that serve them -- and ultimately make it easier for the industry to start using AI on a broader scale.

When it comes to enterprise adoption of AI, "Industry-specific and domain-specific models are going to be the game-changer," said Ritu Jyoti, general manager and group vice president of AI and data as well as global AI lead at research firm International Data Corp.

With these more-focused models, Jyoti said, enterprises can spend less time doing their own fine-tuning and get more value out of AI sooner. "This is going to become a lot more prevalent," she added." [1]

1. Companies Are Developing Industry-Specific AI Models --- Bayer and others are working with Microsoft to tailor tech to their sectors. Bousquette, Isabelle.  Wall Street Journal, Eastern edition; New York, N.Y.. 18 Nov 2024: B.5.