- Industry-specific data will be crucial to the next stage of AI evolution
- Expect data alliances to form, both between model builders and partners as well as industry-specific groups
- Integration between data sets will also be key, OpenText told us
You’ve heard about artificial intelligence (AI) model libraries. But what about data hoards? As large language models begin to branch into smaller, niche iterations, access to industry-specific data will be key to making them relevant, OpenText CMO Sandy Ono told Fierce.
“That’s what’s valuable for the customer,” Ono said of the ability to plug AI models into specialized data sets. And she would know, given OpenText is in the information management business and is thus well attuned to what enterprise customers are looking for.
Take a healthcare customer, for example. Their AI model “doesn’t need to tap into the whole internet but it taps into all the laws and regulations for healthcare for my country, so therefore when I’m making decisions or I’m generating the next step, it’s taking that into account.”
So where will these data sets come from? In a word, alliances.
Partnerships on the data front are already starting to emerge. Indeed, we reported earlier this week on Extreme Network’s move to feed Intel’s device data into its network AI.
And late last year, dominant force OpenAI launched a Data Partnerships program, through which it is working to build open source and private datasets for training AI models. On the private front, these partnerships give it access to non-public information like archives and metadata.
“We’re interested in large-scale datasets that reflect human society and that are not already easily accessible online to the public today,” OpenAI wrote at the time.
Google appears to be working on a similar data partnership initiative to drive its AI forward.
But Ono said vertical players could also form their own alliances.
“When you’re an industry, you inherently know where to go,” she said. “In insurance, you have GuideWire and things like that. So, folks who are those data providers will start to lean in and form these alliances where it matters.”
There’s one more piece of the puzzle: integration.
“You integrated with applications five years ago because you wanted the workflow to be smooth. Now, we need the data to be smooth so the AI can work on top of it,” Ono said. “The customers will drive it, because they’ll be the ones saying ‘well, why doesn’t this connect with that? Why do I have to go to another library?’”
“This is one of those things I think we’re going to evolve into over the next three to six months,” she concluded.