During KubeCon (almost) everyone agreed: the future of AI is cloud-native and open source. Open models and infrastructure should ensure that we understand and use AI better.

KubeCon threatens to get off to a bad start on Wednesday morning: a brief internet outage leads to a standstill at the registration desk. After a few minutes, the deafening sound of ID cards rolling out of the printers turns the initial grumbling into sighs of relief.

There is a rush, but the Eiffel Tower and its employees are full of enthusiasm Tricolores waved and led us to the right room. French chauvinism also finds its way into an international event like KubeCon.

Ten years of Kubernetes

When the morning debates begin, the keynote room is packed. KubeCon is celebrating a special event this year: Kubernetes is celebrating its tenth anniversary in June this year. Organizer CNCF doesn’t want to let this go unnoticed. Prijanka Sharma, Managing Director of Cloud Native Computing Foundation, is proud to announce that this year’s conference has a record number of visitors: 12,000 people came to Paris. This announcement alone triggers thunderous applause in the hall.

“Today, Kubernetes supports workloads everywhere, from the cloud to the edge. This extensibility has always been the great strength of Kubernetes. In these ten years we have seen it learn to crawl, walk and run. Kubernetes will fly in the next decade,” announces the enthusiastic hostess. The biggest players in the tech industry have also received an invitation to the birthday party: on the first day alone, Google, Microsoft, Intel, Nvidia, VMware and the home player Mistral are invited to the stage.

AI is the next big change: it’s no longer knocking on the door, it’s already part of your interior.

Prijanka Sharma, Managing Director CNCF

Era of exuberance

The jet stream that will push Kubernetes to new heights is AI. The current technology hype is also a recurring theme during KubeCon. “AI is the word of the week, you see or hear it everywhere. The AI ​​era has arrived,” said Jorge Palma from Microsoft. Palma is not exaggerating: one AI Hub is set up on the exhibition floor where visitors can discover the wonderful world of artificial intelligence.

The AI ​​era will be an era of “exuberance,” predicts Sharma. “We have seen this with every major change for humanity, from the invention of the railway to the invention of the smartphone. AI is the next big change: it’s no longer knocking on the door, it’s already part of your interior.”

However, there is also room for a critical comment on the AI ​​revolution, which would be considered heresy at many other conferences. Sharma asks the audience who has tried a prototype project using AI. The hands go up quickly. But when it comes to the follow-up question of who has already managed to get these pilot projects into production, there aren’t as many hands left.

“Anyone can develop a prototype, but the challenge lies in implementing AI in daily operations,” says Sharma, explaining the conclusion of her survey. “Many companies fear that the introduction of AI will lead to higher costs. There is a need for standardization to increase the interoperability of AI systems.” “Kubernetes is the best place to run AI models, but I think there is still a lot we can do to make it easier for everyone,” true Google engineer Clayton Coleman agrees.

Always cloud native?

Sharma has a clear answer to the question of implementation: “Developments in AI are progressing very quickly. Cloud native means rapid development. It is the only form of infrastructure that can still keep up with the pace of AI. For me, AI and cloud native are the be-all and end-all: they can exist perfectly separately, but when you put them together you get a perfect combination.”

For Kehan ​​​​Salihbegovic, VP Developer Relations With Docker, the truth is more nuanced. “AI is ubiquitous today and companies want to be a part of it, but are getting lost in the ecosystem. Many companies have come to this conclusion only locally is not the best practice, however Cloud only is not always ideal either. It’s about finding the right mix of on-premises infrastructure and just enough cloud, but not too much. We believe it’s important to give developers the choice of which tools and platforms they want to use.”

“The big advantage is that with Kubernetes we can build on a platform that has been around for ten years. This means that we do not have to start from scratch as before. But the average data scientist shouldn’t know anything about Kubernetes. There is a need for a new role within organizations, that of the “AI engineer”: someone who can balance platform engineers and data scientists,” says Lachlan Evenson, Principal Program Manager from Microsoft and board member of CNCF.

Local only isn’t the best approach, but cloud only isn’t always ideal either. It’s about finding the right mix of on-premises infrastructure and just enough cloud, but not too much.

Kehan ​​​​Salihbegovic, VP Developer Relations Docker

AI belongs to everyone

Everyone at KubeCon has their own opinion about the direction in which AI should develop. The call to keep AI models open was heard loudly several times during the conference. “The tools we use for AI are now almost all open source. “How can we build what we want if the models aren’t?” asks Ollama’s Jeffrey Morgan during a panel discussion.

Marc Collier and Jonathan Bryce from the OpenInfra Foundation also share this opinion. “An AI model is much more than just software, it also includes data, how it is trained, how it produces results… Many of the concerns surrounding AI arise from the mystery that surrounds the technology. We can only understand models if they are made openly available. Transparency is the strength of open source,” explains Collier.

Bryce continues: “The privatization of AI limits who has access to the models. Companies don’t necessarily do this with bad intentions, but it can have dire side effects. History has shown many times that concentrating technology for a few players does not work.”

Wild developments

Collier expects “wild developments” in the field of (generative) AI, especially within multimodal models. “The most interesting models will not be based on text or images, but on a combination. Video will be the next frontier that generative AI will bridge. Models will become increasingly better at understanding our physical reality. It is still difficult to predict what impact this will have on our population and society.”

“You’re not bad at making predictions,” Bryce jokes to his colleague. “The beauty of open source lies somewhere in the unpredictability. Our community is about much more than just developing code: open source is a way to bring people together around collective problems and find solutions that benefit everyone. Users often come up with the best use cases themselves. That’s why it will also be important for AI to put the technology into as many hands as possible.”

History has shown many times that concentrating technology for a few players does not work. Put the technology in the hands of as many users as possible and they will find the best use cases.

Jonathan Bryce, Executive Director of the OpenInfra Foundation

After hearing many opinions and views on the triangle between AI, cloud native and open source, we are ready for some fresh air. Under the wonderful Parisian spring sun, we let the words simmer in our minds. Kubecon will not be the last conference where AI is the main topic. The tech industry can’t stop talking about it, but no one knows where the AI ​​craze will end. In a year, the open source world will put its cards on the table again.