Malted AI — an Edinburgh-headquartered start-up — has raised £6m to scale up its development of “distilled” GenAI models for enterprises. The round was led by Hoxton Ventures with participation from Creator Fund and angel investors.
The company's distillation process uses a “teacher” system that takes a company's raw data and feeds it into a series of small, efficient language models that the company uses to perform specialized tasks. You can complete it.
“It’s the exact opposite of what a lot of people are focusing on from a larger, more general perspective. [AI model] is better,” says Ian Mackie, co-founder and CEO of Malted AI.
For small models
Malted AI's focus on small-scale language models reflects broader trends happening in generative AI. For several years, generative AI has followed the logic that larger models (trained with more data) produce better results. Companies like Paris-based Mistral and Silicon Valley's Meta are pouring significant resources into training models that rely on a fraction of the amount of data that OpenAI's GPT-4 does.
General-purpose chatbots like ChatGPT that are expected to respond to all kinds of queries benefit from large training datasets, while more specific business applications rely on smaller, cheaper-to-run models. You can get equivalent results from .
“At scale, it could easily be 10 to 100 times cheaper.” [to use smaller models]” says McKee. “In the long run, there are definitely cost benefits to really small, focused models.”
Malted AI is working on “several” pilots with large companies to develop language models for retrieving complex information from large internal datasets and finding regulatory issues within text documents. We support the construction of applications that utilize it. By using distillation techniques to generate specialized training data, the company says it can develop “smaller, more focused, and better-performing AI models at a fraction of the cost.” says.
scottish story
From small startups to foundational model companies attracting hundreds of millions of dollars in investment, there are many companies looking to help businesses leverage generative AI for enterprise use cases.
“Our focus is on being known as best-in-class when it comes to distillation and creating really small, focused AI models,” Mackie says.
Malted AI also sets itself apart by being based in Edinburgh, one of Europe's leading AI hubs, outside of London and Paris. Mackie argues that Scotland has a lot of potential, even if it is not yet developed into a major AI company, because of its strong research universities.
“Glasgow is one of the world's largest research areas for information search and retrieval systems. Edinburgh has some of the most talented NLPs [natural language processing] People,” he says. “There are some really great researchers out there, but there aren't a lot of interesting venture capital-backed startups at the moment. So we want to give people really exciting jobs and help them work at the cutting edge without having to go to London or Silicon Valley. We are striving to
Mackie co-founder Carlos Gemmell previously worked in Meta's language modeling department. The Malted AI team (currently 10 people) also incorporates experience from DeepMind, Amazon, and Bloomberg.
The startup plans to use the new funding to continue funding technology development and build a commercial market entry strategy with early partners.
