AI promised to profoundly change the world, so why hasn’t it?
From healthcare to manufacturing, finance, climate, communication, and travel, to how we live and work. AI can help solve any problem that can be represented by data, assuming the right algorithms and enough computational resources.
Today we have all the ingredients to realize the benefits of AI. More data is being generated through billions of phones and IoT devices than ever before in human history, and we have near unlimited storage and compute access in the cloud. Meanwhile AI hardware has made enormous improvements in the cost and performance of raw compute, and the rate of AI research papers published yearly is now vastly superlinear.
Despite the promise and initial successes of AI, it has yet to impactfully penetrate our lives – we don’t have widespread autonomous cars and buses, preventive medicine still relies on doctor’s visits and stethoscopes, drug discovery remains a speculative investment game, global food security is unsolved, farming still relies largely on hard human labor, and real robotic automation at home is still for the Jetsons.
Instead, we are getting more targeted advertising, better social media recommendations, cute AR glasses, predictive credit scoring, and “smart” speakers. These successes may be ubiquitous, but they are also far from realizing the full potential we see coming out of ML research on a daily basis.
AI is a triumvirate of data, algorithms, and hardware, and we are rich in all three - so why isn’t it getting fully utilized? The reason may not be apparent to many, because the truth is deeper than the stream of AI research updates coming from the world’s largest tech companies (MAGMA aka Big Tech) and the press.
If AI is so important, why is its software so bad?
AI software is fundamentally broken because it was designed to be used by the full stack researchers, engineers and architects that built the technology in the first place – it was never built as a product. This software was built by Big Tech companies to solve their own problems, and now the rest of the world runs on “trickle down infrastructure”. The consequence of this is that only the largest and most business-impactful uses of AI get built and deployed in practice, and even then, only if your needs align well with the in-house needs of Big Tech companies.
Why is this? AI software today is monolithic and research quality, fractured into technology silos that follow the organization chart of the Big Tech companies who built it. It was created by researchers to do research, and the quick evolution of AI has left no time to pause and build it right. Instead we've just piled on more and more layers of complexity over time, resulting in an industry that is struggling to maintain and scale fragmented custom tool chains that differ between research and production, training and deployment, server and edge. AI systems have become a vast sea of incompatible technologies that only the largest and most sophisticated companies can brute force to achieve their goals.
Isn’t Big Tech going to solve the world’s AI problem for us?
The successes at deploying AI come from teams who amassed talented armies of AI researchers and developers. Big Tech uses their vast compute and financial resources to further the priorities of their products and core business, including their own clouds, phones, social networks, and AI hardware. Despite their incredible contributions to the field, it doesn’t make business sense for Big Tech to lift AI for the world (spanning all hardware, clouds, and ML frameworks), and the rest of the world cannot expect them to do so. This unfortunate truth imposes an industry-limiting ceiling on the rest of the world’s ability to use this technology to fight problems outside the Big Tech focus areas, including some of the most significant socioeconomic and environmental problems the world faces. That's not a future we choose to accept.
These companies make incredible contributions to the progression of AI, but for AI to reach its full potential, we need an independent company that doesn’t prioritize its own hardware, its own cloud infrastructure, its own mobile phones, or its own research. We need a neutral company that does what is in the best interest of users and companies around the world. We need to incorporate the learnings from the rapid growth in AI software into a next-generation of technology that is designed for usability and generality, for all organizations in all problem domains.
Today, the most pressing problem facing companies who aren’t Big Tech is the ability to productionize AI within performance, cost, time, and talent bounds. Their opportunity cost means innovations not making it to market, inferior product experiences, and ultimately a negative impact on their bottom line. But for society as a whole, this means a longer timeline to solving some of the world’s most critical problems with AI. We don’t have the time to wait for trickle-down AI software from the world’s largest companies. AI can change the world, but only once fragmentation has been repaired and the global AI developer community can take high quality infrastructure for granted, rather than fighting to get the infrastructure to work.
If not Big Tech, who will solve this for the world?
Modular is building a next-generation AI developer platform focused on usability, velocity, and flexibility. Our platform unifies popular AI framework front-ends via common interfaces, and enhances access and portability to a wide range of hardware backends and cloud environments. We are rebuilding core developer workflow tooling to be more expressive, usable, debuggable, reliable, performant, and scalable. Our tools can be easily adopted into existing workflows without rearchitecting or rewriting your code, meeting you where you are today and enabling you to realize improved productivity and performance improvements at lower cost. We will accelerate AI value capture and lower time to market for our customers.
But we’re not content with the status quo. We believe that the full potential of AI will be realized when it can permeate applications in a fine-grained way - AI shouldn’t be something you need to shape your application around. Our platform is built of layered, composable infrastructure components that can be remixed and extended to enable a wide-variety of use cases, while also allowing domain experts to innovate without needing to know how the whole system works. We have seen first-hand how a layered and composable approach can unlock new use cases that haven’t been imagined.
To truly fix AI infrastructure, we are both solving “hard tech” problems (like compilers for exotic heterogeneous compute technologies) as well as building seamless end-to-end developer workflows. To achieve this, our team is made up of the software architects, engineers, and product leaders who have already built world leading developer platforms and the current generation of AI frameworks. We have listened to thousands of users and developers across many years, felt the pain ourselves, and are deeply focused on building products users love.
What does this new world look like?
Our success means global developers will be empowered by access to AI software that is truly usable, portable, and scalable. A world where developers without unlimited budgets or access to top talent can be just as efficient as the world’s largest technology giants, where the efficiency and total cost of ownership of AI hardware are optimized, where organizations can easily plug in custom ASICs for their use cases, where deploying to the edge is as easy as deploying to a server, where organizations can use any AI framework that best suits their needs, and where AI programs seamlessly scale across hardware so that deploying the latest AI research into production “just works”.
We see an AI industry that isn’t dependent on the timeline and capabilities that the top Big Tech companies decide they need. We see an industry that moves even faster and is less fragmented. We see an industry where innovation thrives at all levels of the stack, allowing developers to focus on bringing new innovations to market in their domains of expertise, and to build a better future for all of us. We see an industry with high velocity, lifting us from the “Era of AI Research” to the “Era of Production AI”. That’s a world worth fighting for and a world we can look back on and be proud of.
That’s the world Modular is creating, and the time to do it is now. A better future starts today, and we are assembling the best talent on the planet to do it. We are proud to have raised $30M from GV, Greylock, Factory, SV Angel and notable angels who are backing us - investors who understand AI infrastructure deeper than anyone and recognize its importance to the world.
Join us in building a Modular AI platform for the World.
