Introduction
Imagine you're a chef who loves to cook delicious meals. Instead of always buying ingredients from the store, you decide to grow your own vegetables and raise your own chickens. This way, you have complete control over what goes into your dishes. That’s kind of what big companies like Thomson Reuters are doing with artificial intelligence (AI) — they’re choosing to build their own AI systems instead of renting them from others.
What is it?
Artificial intelligence (AI) is like teaching computers to think and learn, just like humans do. One type of AI is called a language model, which is good at understanding and creating human-like text. Think of it as a super-smart writing assistant that can help with emails, reports, or even creative stories.
When companies use AI, they have two main choices:
- Rent AI: They use someone else's AI system, like those from OpenAI or Anthropic. It's like renting a car — you get access to something powerful, but you don’t own it.
- Build AI: They create their own AI system from scratch. It’s like building your own car — it takes more time and money, but you have full control.
How does it work?
Thomson Reuters decided to build their own AI, which they named Thomson. They used a system called Qwen, which was made by Alibaba. Qwen is a large language model, meaning it’s trained on a huge amount of text from the internet, books, and other sources.
But here’s the twist: this AI works best when it has access to Thomson Reuters' own content — like legal documents from their Westlaw database. It's like a teacher who knows the curriculum of their school best. If the AI is trained on a company's own data, it becomes much better at helping with that company’s specific tasks.
Building an AI system is not easy. It costs a lot of money and requires many experts. Thomson Reuters spent around $40 million over two years to create their AI. That’s a lot of money, but they believe it’s worth it for control and performance.
Why does it matter?
There are several important reasons why companies are choosing to build their own AI:
- Control: When you own your AI, you can decide exactly how it’s used and what data it sees. This is especially important for sensitive industries like law or finance.
- Security: By keeping everything in-house, companies can better protect their data from being seen by outsiders.
- Customization: An AI built for a specific company can be trained to understand that company’s unique needs and style.
For example, a legal firm might want an AI that understands court procedures and legal jargon better than a general-purpose AI. By building their own, they can make sure it’s tailored to their work.
Key takeaways
- AI language models are tools that can understand and create text, like a smart writing assistant.
- Companies can either rent AI or build their own.
- Building your own AI gives more control, better security, and custom performance.
- Thomson Reuters spent $40 million to create their own AI because it helps them do their work better and keep their data safe.
Just like how a chef might choose to grow their own ingredients to make the perfect dish, companies are choosing to build their own AI to meet their exact needs.
