Investing in Generative AI with Hexo AI
Generative AI is opening up a world of opportunities and Hexo is building the underlying infrastructure to enable that.
With Generative AI, we are at the cusp of another defining decade in the human-machine partnership. Backing founders at the earliest stage in their journey enables us to support truly orthogonal and experimental companies that are bound to define the trajectory of how technology companies are built in the future. One such company is Hexo. Kunal Bhatia and Vignesh Baskaran, the founders of Hexo, met at our first-ever Antler India Residency and embarked on a journey to build in one of the most exciting spaces of recent times.
Separating signal from noise: What’s happening in Generative AI
When you take a step back from all the hype, what we are witnessing right now in the Generative AI space has classic parallels to the playing out of the democratization of any major technology. A realization of the immense potential of a technology that will create a foundational shift in the world. Just as mobile, cloud and web3 did, but at a much larger scale. With one key difference, however. With Generative AI, progress in fundamental breakthroughs is happening in tandem with the emergence of interesting products, at an extraordinary pace. As organizations like OpenAI and StabilityAI publish new research work every few weeks, we are also seeing the likes of DreamStudio, Runway, Lex, GitHub Copilot, Copy.ai and Jasper take off. This is bringing about a Cambrian explosion that will result in the creation of generational companies. And it is something we just cannot afford to ignore.
It’s still very early and moving very fast. GPT-3 was released less than three years ago with ~200 billion parameters; the upcoming GPT-4 has ~1,000,000,000,000 (a trillion) parameters. In just about 5 days since its launch, ChatGPT has reached a million users. And we are just getting started.
This new era of AI is going to transform the way work is done and productising AI will unlock many opportunities. To read more about how we anticipate the Generative AI space evolving read our take on AI as humanity’s miracle machine.
Introducing Hexo, the infrastructure backbone for enterprises to integrate Generative AI
As more and more people consume content digitally, the need to generate very high volumes of image and video-based content has increased significantly, not just for consumers but for enterprises too. The new-age digital workflow relies on visual data more than ever. However, the speed and cost of generating useful image content at scale have not seen any significant improvements in efficiency and there is still an extremely heavy reliance on designers and templates to deliver the final output. This has meant that the demand for image content exceeds the speed of supply, and the gap only continues to increase.
As we looked for solutions to the above problem statement, we stumbled upon Stable Diffusion and the ability to generate images out of raw lines of text. Though Stable Diffusion has enabled image generation from text, currently the output does not have high utility, restricting its enterprise usage.
This is because the current products give one-shot input-output which is not suitable for the iterative, complex and custom nature of enterprise image creation work. For enterprises and developers, integrating AI-generated image creation into their existing workflows still remains challenging. While companies can attempt to build their proprietary image generation engines, pipelines and workflows - customizing large language models, hiring specialized ML developers and setting up a GPU cluster is a complex and expensive task. This is why we are backing Hexo.
As its first offering, Hexo is building an open-source image generation API that offers a wide range of controls to make image generation much more predictable and accurate. This “fine-tuner” product for enterprises will enable them to easily build custom image generation engines based on the company’s design language, products, characters, IP etc and embed text to image generation in their workflow. Think of an application for a game developer. Using Hexo, they can create and integrate the most complex game assets for a fraction of the cost and time. And it doesn’t stop there, from custom film storyboarding to animation and VFX, this API solution can be used to build several nifty tools on top, unlocking significant value add for enterprises, brands, and creative professionals.
Besides building a simple plug-and-play infrastructure, Hexo creates a solid product moat through their own search and image generation tool along with a massive repository of pre-existing images. This, we believe, is the first step towards democratizing image generation.
Only a couple of months into their journey building Hexo (after first meeting at the Antler India Residency), Kunal and Vignesh have made significant progress showing immense promise in keeping up with this extremely fast-evolving space. Vignesh is a deep learning expert (Top 1% Kaggle) and led multiple machine learning research teams globally. Kunal is a three-time founder, having earlier built SuperLearn (EdTech) and Switch (IoT). Given Vignesh's engineering strength and Kunal's business skills, we are sure this is a team that is rightly suited to build AI products for scale.
We are thrilled to lead Hexo’s pre-seed funding round and partner with them in this journey of building in the Generative AI space. We are still in the very early stages of AI but when the cost of creativity tends to zero, the possibilities truly become endless and we are excited to partner with Hexo going into this new age of AI.
Read more about Hexo’s pre-seed round funding as featured on Entrepreneur India and Entrackr.
If you are a budding Indian entrepreneur or looking to jump into the founder journey, get in touch with us and let's get started!