Dillon Erb is CEO of Co-founder of Paperspace, a leading provider of AI computing in the cloud. Dillon along with his co-founder Daniel Kobran built Paperspace to accelerate the next generation of cloud AI by making high-performance computing and ML development frameworks accessible to every developer.
Paperspace is the leading provider of high performance cloud computing. We offer a range of products powering everything from virtual desktops for businesses, 3D graphics and rendering for the VFX industry, Machine Learning pipelines for developers and technology companies, and gaming rigs for individuals.
Tell us about your journey into Artificial Intelligence? How exciting is it for you to be a part of an AI-heavy ecosystem?
When we first got started with Paperspace, we knew that GPUs were going to be important. While this bet was the right one to make, we would be lying if we told you we knew how fundamental this transformation would ultimately become. In the last few years, we’ve witnessed one of the biggest shifts in the history of computing: GPUs have moved from graphics pipelines to general compute devices, unlocking entirely new possibilities and workflows. This evolution has brought about an explosion of software built specifically for parallel compute, and with that a whole new set of applications like deep learning. But these modern applications do not fit nicely into the web services model which is what the public cloud was built to support. So we built Paperspace as an elegant front-end to GPU compute.
Over time, we uncovered a bigger problem. Development bottlenecks were were hindering innovation in AI. Researchers and developers were spending a bulk of their time managing infrastructure instead of developing deep learning models. And the tools available to them — AWS, Azure, Google Cloud, etc. — add complexity to an already complex process. So we evolved Paperspace to abstract the complexities of developing machine learning applications. Our latest product, Gradient°, puts Facebook-grade AI tooling in the hands of every developer. And this is just the beginning for us at Paperspace.
What are the core tenets of your AI and machine learning roadmap?
The first major one is accessibility. The tools available for building AI today are costly — both from a time and resource perspective — and keep developers from realizing their true potential. Community is also a core value at Paperspace. We are dedicated a building a strong body of developers through knowledge sharing and our ongoing contributions with institutions like Fast.ai and universities such as NYU, Columbia, Georgia Tech and others.
Second on the list for us is transparency & power. AI should not be a black box and reserved for Google-level developers and large corporations. The only way to deliver on the promise of AI is to open it up to everyone.
What are your predictions for AI in IT, Marketing and Sales?
AI will take over every business process. Any manual heuristic will be replaced by an online, real-time prediction engine. We’re already starting to see its impact on core business functions like sales and marketing.
Also, very soon, AI won’t just be a “department” or “research arm” — it’ll merge with development and engineering teams across the enterprise.
How do you see GPU cloud platforms evolving with the maturity of AI and Machine Learning?
AI has prompted a need for more powerful processing. This has given rise to special-purpose hardware to meet these high-performance computing needs. Left and right, we’re seeing new chip startups like Graphcore, Nervana, Mythic, DeePhi, etc. enter the scene. In fact, according to CB Insights, chip start-ups are poised to raise $1.6 billion this year alone.
Also, we’re going to see the ecosystem get richer for developers. That
means more hardware and software that simplifies development processes like Paperspace. We’re also going to see an explosion of tools for pipelining that perform core functions like monitoring, testing, profiling and more.
How do you intend to expand your horizon to meet business revenue objectives and automation standards?
On the tech side, we’ll continue building our platform with more tools and supporting more language integrations. We’ll build in more functionality to our existing tools, making them deeper and richer. The ecosystem will also be a big focus for us. We’ll continue to integrate with the big clouds — AWS, Azure, Google Cloud — as well as emerging, best-in-class startups like QuiltData.
Our end goal is to make it easier and faster to do real production ML. This opens up new audiences and market opportunities. Today our core users are researchers and advanced practitioners, but we’re committed to building a future where all developers are AI developers.
How can AI/machine learning help to build better infrastructure?
AI and machine learning require more powerful computing. This has pushed the traditional computing paradigm to its limits and reinforced the fact that Moore’s law is nearing its needs. The shift to GPUs and now dedicated AI chips is case in point.
We’re also seeing new chip companies emerge to cater to specific use cases such as audio, visualization and voice processing, among others. These new chips are purpose-built and are redefining computing from a hardware perspective.
Define your ‘Ideal Customer’ profile? How do you leverage predictive analytics and AI to zero onto your customers?
Whether you’re an enterprise or a small business, Paperspace is a fit for developers of all levels. We focus on abstracting the complexities of building AI applications so it’s accessible to everyone.
How do you consume all the information on AI and other emerging technologies for advertising and branding?
First and foremost, we listen to our customers and to our community. We’re also involved in the Hacker News and Reddit communities.
We also learn a ton and keep a pulse on things at top conferences like NIPS, ICML, Spark + AI Summit and others.
There is also a strong cadre of journalists promoting the movement like the good folks at the New Stack, Programmable Web, SD Magazine, TechTarget, and others.
In 2018-2020, what are the biggest challenges in the adoption of AI/ML?
The future of AI lies in giving developers access to tools to simplify the development process. Just to get started today, you need the manpower, resources and time that only large corporations have — and that’s threatening innovation. It’s also creating a massive skills gap, which is only exacerbating the issue.
How do you make AI deliver economic benefits as well as social goodwill?
Giving anyone access to AI means a distribution of power. This means no single company can monopolize the technology and use it for their own purposes. AI has the power to shape societies, redefine health and fuel global markets; it also has the power to do evil if it falls into the wrong hands. This is precisely why we must all work together to democratize AI.
Tell us about your collaboration with educational institutions?
We currently work with top universities like NYU, Columbia, Georgia Tech and others. We give students access to Paperspace because we believe the future of AI lies on the shoulders of this generation of students.
We also invest heavily in thriving online communities like Fast.ai, Udacity, and others. These communities are what’s driving current adoption. We share a common philosophy of developing AI in the open and giving access to free resources, so everyone has a chance to learn.
Our primary goal is to become the tool these new developers learn on.
What is your vision in making AI and Machine Learning readily available to enterprises?
Every company out there will need to change the way they do business. They will have to reconfigure processes to harness the power of AI. All functions — internal and customer-facing — will be impacted. We want to accelerate this process with elegant tooling that makes this transition painless.
The Crystal Gaze
What AI start-ups and labs are you keenly following?
What technologies within AI and computing are you interested in?
A big thing today is a technique called reinforcement learning, which is inspired by behavioral psychology. Also, media production / generative images through the use of GANs (generative adversarial networks) is also becoming an area of interest.
As an AI leader, what industries you think would be fastest to adopting AI/ML with smooth efficiency? What are the new emerging markets for AI technology markets?
Right now, neural networks / deep learning perform computer vision tasks with great efficiency — detecting cancer, driverless car mapping, etc. — but soon it will make its way into every industry.
What’s your smartest work related shortcut or productivity hack?
Arxiv Sanity Preserver — easy-to-use repository for research papers
The Tab Wrangler browser app that quietly kills all those extra tabs to avoid clutter.
Tag the one person in the industry whose answers to these questions you would love to read:
Thank you, Dillon! That was fun and hope to see you back on AIthority soon.