How should entrepreneurs think about new business opportunities in AI?
Microsoft AI CEO Mustafa Suleyman told LinkedIn co-founder Reid Hoffman an October episode of the Masters of Scale podcast that startups can find a niche in fine-tuning AI models with accurate examples. Fine-tuning means revising example models so that they perform better with fewer hallucinations.
“You have to show (an AI model) tens of thousands of examples of good behavior and you have to fit them into the model,” Suleyman said. “The good news is that tens of thousands of examples are very accessible for many niche areas or specific verticals. So that's an advantage and I think there's a lot of room for startups to do high-quality tuning of a pre-trained model”.
Microsoft AI CEO Mustafa Suleiman. Photographers: Stefan Wermuth/Bloomberg via Getty Images
Small AI models are the future
Small AI models will be the future of AI, according to Suleyman.
“We're going to condense knowledge into smaller, cheaper models that can live on a fridge magnet,” he said.
Training a large AI model currently requires about 100 million dollarswith more advanced models expected to cost billions of dollars. However, the data that goes into training these models is controversial, with many copyright lawsuits pending against companies like OpenAI.
Is AI training ethical?
In June, Suleyman answered the question of whether AI companies have taken over the world's intellectual property for their own benefit. He stated at the time that almost all content on the Internet, except for news sites and publishers who have asked not to be crawled, is open to AI training.
“I think in terms of content that's already on the open web, the social contract of that content since the '90s has been that it's fair use.” he said at the time.
Related: Microsoft AI CEO says almost all content online is fair game for AI training