KubeCon China - don’t say OpenAI, say ‘open AI’

While Western governments and proprietary tech companies have tended to regard China with suspicion in recent years, it is now the world’s second largest contributor to the open-source and cloud-native computing community worldwide. 

That was the context for the combined KubeCon and CloudNativeCon event moving to China this week. Specifically, to former British territory Hong Kong, which – though part of the People’s Republic of China since 1997 – still retains its own currency (the privately issued Hong Kong dollar), traditions, and a majority Cantonese-speaking population. Even the cars still drive on the left.

Even though KubeCon China has also joined forces with the Open Source Summit and AI_dev conference too, this feels like a focused local event, especially when compared with the massive Paris summit in the Spring (see diginomica, passim). But it is also a welcoming and collaborative space, in line with the open-source community’s values. The atmosphere is friendly, open, excited, and enthusiastic: not the China we hear about in the West, perhaps, but one that is very much in evidence at KubeCon.

The Hong Kong jamboree is perhaps one tenth of the size of KubeCon Paris, yet still buzzy and engaged, especially in the always-busy exhibition hall – fittingly for a nation that now contributes ten percent of updates and innovations to the open-source movement.

Opening the event, Executive Director and CEO of the Linux Foundation Jim Zemlin said:

We have almost 800,000 developers who work on Linux Foundation projects every single day, which technically makes us the world's largest software company. But we don't really employ most of those developers. They work all over the world, in digital companies and organizations. What we do is all the stuff that you need to facilitate this great development, and it continues to grow year over year. In fact, our organization is growing about 30% annually, and that’s a testimony to how important open-source software is.

But for Zemlin the dominant theme is clear:

This week, as I've been traveling throughout China, the only thing that people want to talk about is AI. But one of the things that is important to understand – before we talk about specific open-source projects – is the AI stack. Most people who are interested in artificial intelligence are focused on the top of it. For example, just yesterday in Beijing, I saw a talk from Alibaba who are creating an AI application for early detection of pancreatic cancer – amazing work: an application that is already saving lives in China by helping to detect pancreatic cancer as early as possible. So that's where most of the development is taking place.

But underneath those open-source applications are a bunch of other things. […] And in open source, there are a few specific areas where there are opportunities to take advantage of collective development, and these aren't all just about the source code. First of all, in fine tuning specialized models: I think this is an area where there's a real opportunity for people to collaborate on a set of standards that make it easier for enterprises to deploy rag models and to really deploy enterprise AI.

He continued:

We're already seeing incredible breakthroughs in open source with Large Language Models. Models like Mistral and Llama 3 are incredible, cutting-edge frontier Foundation Models that are rapidly catching up with, and in some cases surpassing, the leading frontier Foundation Model, which is, of course, OpenAI’s GPT. And in model hubs, Hugging Face is the clear leader here. And there's a whole ecosystem of open models that people can download and use to make AI applications.

Safety

So, what about that key issue for policymakers worldwide: AI safety? Zemlin said:

This is a particularly good opportunity for open source to address concerns around AI safety, and to co-develop tools and standards to help with content authenticity and privacy, to help fight deep fakes, fraud, and to help root out model bias. Open Source is particularly helpful in AI safety because of the transparent nature of the development process.

He added:

This week, the Cloud Native Computing Foundation is here, and while the CNCF is called that, it might as well be called the ML Ops Foundation, because all large language models and machine learning tools are deployed on cloud-native infrastructure.

Lacking

In other areas of AI, the lack of open policies is beginning to cause problems, Zemlin suggested:

On the data side is probably one of the biggest opportunities and challenges. As data becomes the building block for LLMs in the United States, we're already seeing companies cut off access to the ability to crawl data. We can see a whole new world emerging where data is sold to train LLMs, but it's only available to the people who have the most money to buy it. So, there is an opportunity for us to present an open alternative to closed data. But that is something that's going to demand resources and a very focused approach.

Then with tongue firmly in cheek, he added:

Some people, when they say, ‘open AI’, they mean the company OpenAI. But when I say it, I mean the open space of AI. And there are so many good folks [in our community] who are working on developing a common understanding of what ‘open AI’ really means.

But on that point, is China itself really as open as KubeCon would suggest? Chris Aniszczyk is CTO of the CNCF, the host of these events in different parts of the world. He said:

We have a lot of great contributions from China across our whole community. If you look at the straight-up raw data of contributions, China is the number two contributing country across all of our projects. Indeed, some projects are actually born in China.

Even so, in the world of realpolitik and complex international relations, how do organizations like the CNCF, the Linux Foundation, and others in this community, balance their own open cultures with the political ramifications of having China as the number-two contributor to a US-based, if global, organization and movement?

He said:

At the end of the day, open source has always been a global endeavour. People have contributed from all of the world, and open source has always been considered something that is meant to achieve that. At the Linux Foundation, for example, there is great work behind the scenes to ensure that things are, legally, kosher.

You were at KubeCon Paris. Now we're here at KubeCon China and we are doing the first KubeCon India later this year, in December. So, we are going to where our community is and trying to work together. The best ideas will be from everywhere. It's truly the best way to develop software. As long as the code is as good and works, and it improves something!

My take

In this world, the source code and not the source of the code takes precedence. More exclusive content from the conference this week.