Patsnap built its own LLM, but will profits come?

IN FOCUS

In today’s newsletter, we look at:

• Patsnap invested ‘7 digits’ to develop an LLM. Will it pay off?
• Hurdles beyond the hype of GPT-4o’s impact on businesses
• How a new AI-focused collaboration between the US and Singapore risks being a talent bridge to nowhere

---

Welcome to the first edition of The Prompt! Delivered every fortnight, this free newsletter breaks down the biggest stories and trends in AI. Get all our content, including The Prompt, by registering for a Tech in Asia account.

Hi there,

Could AI have rights at some point in the future? This question was raised during a panel discussion I attended at last week’s SuperAI conference in Singapore.

My first reaction was incredulity: Is this more exaggeration from the cheerleaders of what is an already overhyped sector?

On further reflection, though, that possibility might not seem so far-fetched. After all, companies have rights and obligations even though they’re not people. Also, societal attitudes change. The modern animal rights movement only got off the ground after the publication of Animal Liberation: A New Ethics for Our Treatment of Animals in 1975.

Unlike animals, of course, AI isn’t sentient – not yet, anyway. Could it ever be?

The fact that we’re even having these discussions speaks to the profound ways that AI will affect us going forward. It will have a significant impact not just in tech and business, but also across society, from defense and the law to labor and education.

That’s why here at Tech in Asia, we believe it’s important to deepen our coverage and understanding of AI and everything related to it.

Much of the attention – and capital – in this space is focused on generative AI and its foundation layer: the large language models (LLM) that enable applications such as ChatGPT to respond to questions and Midjourney to generate fantastical images.

This week, our Big Story story looks at Patsnap, which runs a global patent database. The Singapore-based unicorn has spent millions of dollars building a proprietary LLM on which its AI assistant Hiro is based. Patsnap’s LLM incorporates datasets from both its own data and Meta’s open-source LLaMA.

Why did Patsnap decide to do this? After all, existing LLMs developed by the likes of OpenAI and Anthropic are considered best in class.

Patsnap co-founder Guan Dian admits that it’s “very hard to measure” the return on investment for this initiative. However, she maintains that the company has “good reasons” to develop its own LLM.

Do read the piece and figure out for yourselves whether this decision makes sense.

---

THE BIG STORY

---

THE HOT TAKE

---

DEEP READS

---

FYI

---