If Weaviate was a typical human child, this is about the stage of development when we would begin using the future tense in speech. And, at about four, kids begin using words that are 100% understandable to strangers, not just family members.
Of course, we’ve been thinking about the future from the very beginning; our fourth birthday is a rare occasion when we’ll dwell on the past for a moment. But coincidentally, one thing that has happened recently is analogous to that hypothetical four-year-old child using language that strangers can understand: In the last few months, we have found that when we talk about vector databases, we’re more widely understood too!
For better or worse, AI-native companies will see ChatGPT as a watershed moment
The media’s recent obsession with generative AI and ChatGPT in particular was a bit of an early birthday present for Weaviate. Like so many watershed moments in tech—or indeed, in life—this happened at first very slowly, and then all of a sudden.
Around 2014 people tinkered with machine learning and models, thinking of new ways of storing data and representing it. In 2016, Google—which of course had become a global behemoth on the strength of keywords—announced that it would, henceforth, be “AI-first”. A year later, a group of Google scientists delivered a seminal paper, “Attention Is All You Need”, that conceptualized transformers. Seeds were sown.
“I was at a conference in Greece—it was all about linked data—and there was a guy there from the EU who complained that he’d been tasked with compiling a database of all the rivers, lakes, and seas in Europe. ‘The problem,’ he said, ‘is that in German, the word for lake is see.’ My suggestion was, why not use machine learning and just let the model figure it out? Now, that is the go-to answer but back then, it was blasphemy.” - Bob
It was another couple of years before we founded SeMI Technologies, the company that became Weaviate. At the beginning we were influenced by the popular notion of the Internet of Things. We realized that pushing towards the edge meant that companies were dealing with many disparate databases. Our first thought was that we could create a product that unified them.
“That was also a semantic search use case. But as we tried to build the technology there was a point where we got this crazy idea: What if we don't do it on the schema? What if we do it on data? Once we saw the value in doing it on the data, the next obvious question was, shouldn't we also store the data? Because if we want to have efficient retrieval, it doesn't make sense to store it somewhere else.” - Etienne
At the time, we were enchanted with the magic of machine learning and the idea of navigating in a vector space to find things that were logically related. We saw the value in giving people that ability. But sometimes it seemed that only we saw it.
It took a long time to convince people to let machine learning train the model and use the information from the model embedding to organize, store, and retrieve data. It was literally a paradigm shift, and it didn’t come naturally to most potential clients. For a while, we’d give demos in which we used a publishing business as an example. Clients would politely listen and then say, “That’s nice, but we don’t have a magazine.” We had a pitch deck; we raised zero dollars.
“Early on—before we had customers or users reaching out to us and saying, ‘I have a problem that can be solved with vector search’—there was doubt. But the belief in Weaviate was so much bigger than the doubt that it was obvious we should keep going.” - Etienne
As we forged ahead with Weaviate, we gradually built up our company and community of users. Investment, when it came, seemed to happen organically.
And then suddenly…
In the history of AI technology, ChatGPT is just another step. But in terms of enterprise interest in AI, it marks the beginning of a new era. We’re no longer patiently answering questions about which verticals and what use cases. C-level execs are demanding to know how their companies will use generative AI. Customers, including major global brands, reach out to us having already identified specific problems they want our help in solving.
“I'm surprised at what's happening now—not because people get it, it was obvious people would eventually get it—but it was there for a long time; people were talking about it, people were showcasing it. The thing that takes me aback is how suddenly it’s all going now. We’ve gone from zero to hearings in the U.S. Senate in a few months.” — Bob
That change sparked growth. That growth’s forced change
At that hypothetical four-year-old child’s birthday party, the focus would be on cake and treats. But for us as the founders of a company that’s reached four, the tendency is to look back, and take stock of where we are now.
We’ve accomplished many of the things that we set out to accomplish back in 2019. And as a result, our day-to-day jobs are now a lot different than they were. There are new satisfactions that come from watching Weaviate grow; analogous, we suppose, to the feelings a parent has watching a child develop. But we both have to admit that change brings new challenges and sometimes it means doing less of the things we loved when our company was just a handful of people.
“The difficult thing as a founder is that you want to do everything. You wear too many hats, and if someone else doesn't do something, you tend to say, Okay, I'm wearing 13 hats. I can just put on a 14th hat. Now, I’m reversing that and giving those hats to other people, trusting that they’ll do the right thing. What is left over after that is where I can create the most value.” — Etienne
Etienne was basically our Founding Engineer (although he claims not to have even been familiar with that term at the time). He was the one person who, hands-on, built the first product. From there his role evolved into something like an engineering team lead, then into something like a VP, Engineering. Now it’s evolving further into a classic CTO role. Bob’s role and challenges have changed as much.
“A friend who leads a much larger business than Weaviate warned me that as we grew I’d face two big challenges. One was context switching, and he was right about that. I have days with back-to-back meetings on completely different topics, and that is exhausting. The second thing is, so often when people send these things, I read them, I see them, but the thing he warned me about was, Nothing that comes to me is going to be easy. Even at our current size, every easy problem–including many things that I would have dealt with in the old days–is now managed away. So, the only questions that reach me are hard ones! Even ones that call for a binary yes-or-no answer involve a lot of money or make an impact on people.” — Bob
We recently published a blog post on our plans to grow Weaviate in a “cellular” fashion. We believe that strategy will allow us, as founders, to contribute maximum value while empowering other managers and team members to grow the company, and their careers. We’re living in a unique moment in terms of popular interest—and enterprise acceptance—of AI.
Happy Birthday to us, and thanks to all the people (you know who you are!) that helped us reach four.
Weaviate is open source, and you can follow the project on GitHub. Don’t forget to give us a ⭐️ while you are there!