Q&A with Jerry Ye

Listening to the artificial intelligence of Chinese business, he draws a completely different picture of how technology can be used than usual.

After finding as a data scientist at Meta, Caltech graduate Jerry E. returned to China in 2017 by investor’s will to create AI global company.

Startup, Keith, sells AI software and related retail equipment. He now claims that customers such as Starbucks, Xiaomi and Unilever. Last week, Whale announced that he had raised $ 60 million in the CD Financing Series, which included Bosch Ventures, Singtel and MDI Ventures. The startup is considered Hangzhou as China’s Chinese and Singapore headquarters as a global base.

This week I caught up with you in Beijing when he was walking around the city where he had a small team. This interview was reduced and edited for clarity.

What is the influence of the US-Kita Trade War?

We still believe that there is a global opportunity even if there is tension in the US. Business is a business. Our customers are global: Procter & Gamble, Watsons, Starbucks.

Who wins the AI ​​race?

The largest two markets are still the United States and China. Application layer will take off at a different way: marketing and selling in China; Very, very small vertical sectors in the US, such as in transport or legal. South -East Asia is likely to accept scripts from the US or China markets.

What limits AI development? Extended semiconductors?

For China, we do not see the lack of AI computing power. All clouds (companies) have a lot of training force. Many AI companies really no longer train (model).

Ultimately, the chips on the device (often related to edge calculations, or processors built into devices such as smartphones, home appliances, etc.) become powerful enough to run AI models and can provide data privacy. Ultimately, most of the competition will take place at the border rather than at the data center.

Where AI first make money for business?

Marketing and sales because they are very close to money, and the profitability of investments is easier to calculate. If you take AI in marketing and sales for seven days, on the eighth day, you will see the results.

What does this mean for the future content?

The basic concept of marketing is communication, narrowing the gap between the product understanding and the brand that sells the product.

We talk a lot about personalization. But traditionally we did not have enough content, sufficient power to give each person personalized content. With a deeper understanding of AI and algorithms to overcome the content and customer decision -making faster and easier.

Generative AI certainly gives us many opportunities to create a huge number of videos.

AI also allows us to send unique sections of photos and videos in social media pictures. This has become more and more advertising.

But if you use a lot of generative II, they transmit a lot of fake information. Most platforms are basically antgenerative videos with II. My personal opinion is that real -life videos are still needed. For generative II, it is very difficult (conveying) real emotions, a real human touch.

How do companies use AI today?

Our company serves Starbucks in China and Southeast Asia.

Whenever you buy Starbucks coffee, a huge amount of II is used to improve customer experience. Consumers can now get a coffee order within 5 minutes, 98% of the time. If one place is too busy to handle online orders on the top of orders in the store, AI analysis can send these online orders to another place nearby.

How much money does a whale earn? What are your costs?

Over the last three years, we have grown by about 100% (per year), and over the next two years we will support 100% annual growth. Last year, we brought a profit of $ 230 million ($ 31.39 million) and expecting a 400 million yuan revenue this year. We have about 500 customers of enterprises.

We spend about 1 million yuan per month in China for operating costs – outside China, it is more expensive, so the total amount is about 3 million yuan a month (36 million yuan a year).

Our biggest problem is to meet local data needs and security in each region.

Which big linguistic model AI is better?

Different models have different opportunities. For Claude, they are really good in creating web -stytes, pdfs. Twins, videos. This is how to have a PhD in different fields. For most of our business customers, they do not want to be fixed in one model.

Many businesses want to save their data private. That is why they use open source models, deployed in their data processing centers.

Chinese and Hong Kong shares are traded on Wednesday, when investors continued to evaluate the optimism of the US-European Union.