This $40M AI Gamble Could Change How Companies Use LLMs

Thomson Reuters spent $40 million building its own AI model, with a single training session costing $450,000. So why make such a huge investment when new AI models are leapfrogging each other every few weeks?

Ulysses & Levi explore the classic build versus rent question for AI. Once a model reaches the level of capability needed for a specific use case, such as legal, risk and intelligence work, constantly chasing the latest benchmark may no longer provide enough additional value.

The bigger advantage can be proprietary data. Companies with unique datasets can fine-tune capable open source models around information their competitors simply do not have. That creates a potential moat while also giving them greater control over infrastructure and ongoing inference costs.

Ulysses also explains what businesses should consider before building or fine-tuning their own AI model, including how much unique data they possess, whether existing models are already capable enough, and whether ownership provides a genuine competitive advantage.

#AI #ArtificialIntelligence #ThomsonReuters #GenerativeAI #LLM #OpenSourceAI #FineTuning #MachineLearning #AITechnology #SSWTV

v1 – Marcus

Reuters brings you the latest breaking news, business and finance video from around the world. Since our founding in 1851, we have been known globally for unparalleled accuracy and impartiality….brought to you by SloppBoxx.Com.

How useful was this post?

Click on a star to rate it!