Let Data Flow in the Sunlight

See you in Haidian this Sunday.

@Ruirui Wu

After writing Xiao Wu’s Journey Through the Data World last week, I got to know more people at model labs and data vendors. Many of them seem to see us as an information hub: their first reason for reaching out is often to ask me to introduce them to one another.

People at model labs ask me: Do you know any data vendors selling data in verticals like XX or XX? People at data vendors ask: Do you know what kind of data that model lab or major vendor needs right now?

The data industry seems to have always had a great deal that goes unsaid, creating striking information gaps. There are legitimate reasons for this. No model lab wants its data needs laid bare for everyone to see, and data vendors want their products to remain distinctive.

But it has also produced some strange phenomena. There is little point in continuing to discuss all the shady rumors flying around; they are just the usual excesses that come with rapid growth. What matters more are the issues tied to raising the ceiling of intelligence: Who should establish fair standards for data acceptance? What kinds of next-generation models will emerge from each lab’s preferences in data? What will the long-term relationship between data vendors and model labs look like? Questions like these remain hidden behind the rumors.

This also reflects certain tendencies across the venture capital industry. A few years ago, VCs were generally willing to talk about the projects they had invested in, and startups were happy to share their progress. Now things seem to have split into two extremes: either everything is shrouded in mystery, or the publicity is a highly calculated PR campaign. The former invites baseless speculation, while the latter sets companies up for a brutal fall if their products fail to live up to the hype.

These issues need to be discussed openly. If the industry could reach a basic consensus on acceptance standards, establish the necessary communication between buyers and sellers, and develop a reasonable shared understanding of the value of different types of data, then the data industry — after a full year of explosive growth — might no longer have to remain at this primitive stage, where deals depend on relationships and information asymmetry, and people can still get rich by exploiting loopholes.

Data is one of the key factors determining the ceiling of intelligence (compute is a topic for another day), so developing industry standards is enormously important. Compared with that, even the most sensational rumors are nothing more than idle entertainment.

So this Sunday, November 22, Elsewhere will host an event about data. It will take place in Haidian, Beijing, and we will share the exact location with those who register.

Confirmed attendees so far include heads of data, vertical tech leads, and algorithm specialists from model labs; founders of leading data vendors; and Xiao Wu, the protagonist of the piece above.

To make sure everyone can speak freely and discuss the issues that really matter, we are capping attendance at 30 people. If you have experiences or ideas to share, we welcome you to sign up.

The theme of the event is “Let Data Flow in the Sunlight.” The phrasing comes from a slogan used many years ago: “Let Power Operate in the Sunlight.” As people said at the time: Shadows breed abuse and corruption; sunlight brings fairness, communication, and efficiency.

What we want to say is that the same should be true of the data industry.

Our thanks to the partners who helped make this event possible: Zhihu and AGI House.

Scan the QR code to register

Cover image: Job Adriaenszoon Berckheyde, The Old Stock Exchange of Amsterdam, 1670, Museum Boijmans Van Beuningen