It’s a busy time in language technology as companies race to carve out a place in a rapidly evolving world. And Phrase is no exception. Company Chief Executive Officer Georg Ell, Chief Product Officer Simone Bohnenberger, and Vice President of AI Research Alon Lavie recently sat down with Nimdzi Insights’ Josef Kubovsky to discuss the latest in the company’s efforts. That includes the technology behind Phrase Translate, which aggregates 30 different machine translation providers to help customers select the best service for their needs.
“We’ve been doing (machine language) solutions for five-plus years, and in the last two years, we’ve been building the workflow tool of the future: the custom configurator, the orchestrator that allows you to pull every API everywhere,” Ell told Kubovsky.
According to Ell, the company has been focusing on three pillars as it creates a future-proof suite of services: artificial intelligence (AI), workflow, and scale.
“What I think people will see in the next release is a very strong focus on AI and some new capabilities for us there, as well as a continuation of historic capabilities, and then the next generation of the workflow tools,” Ell said.
It’s a dynamic world to work in, as much for Phrase’s customers as for the company executives themselves. According to Bohnenberger, many are still determining how developing technologies will precisely fit into their workflows.
“We have the honor of working with industry-leading institutions … who are not just large institutions, but sometimes still growing at 40-60% annually themselves. So (these are) really big cutting-edge and innovative institutions,” Bohnenberger said. “What we hear throughout the board is that, in general, executives keep on asking their teams to innovate and play with large language models (LLMs). We’ve seen very little going into production because there are still so many question marks, and there’s such a requirement for supporting infrastructure around it.”
Much experimentation and development needs to occur before a fully formed vision of future work processes emerges. But Lavie believes there’s much reason to be excited.
“Just exploring and understanding and developing those proof of concepts is amazing,” he said. “We can do that today. As this stabilizes, we will find out basically what this will distill into and which capabilities are really most important for what we’re trying to do. And we’ll develop access to stable LLMs that will be able to do this reliably and at scale. And that’s what we’re kind of building on here at Phrase as well, and that’s the way we’re thinking about it.”
