Recently I was named by Medium as one of the top writers on the platform on the subject of Artificial Intelligence which was great, but that made me think of all the incredible people in the space whose take on AI would help enrich this blog and expose our readers to the broader world of technology.
One of the first folks that came to mind was Element AI co-founder Jean-François Gagné, a friend of Jimoh’s and mine who has been doing incredible stuff over the past few months including raising $102 million, launching a $57 million dollar AI fund, working with awesome companies around the world, and recently opening up a London office.
I made a point of catching up with Jean-Francois to get his thoughts since the raise and his take on what’s happening on AI and deep-learning, which you’ll find below. Enjoy!
The Element AI team takes time out of their day to strike a pose in their office in Montreal, Canada
We offer an array of products and services that help large enterprises to transform their business to run on AI. Our platform focused on Enterprise AI, meaning that it uniquely deals with complex problems, relatively small amounts of data to train from, and very high-performance thresholds required by global organizations. We help our customers go from ideation, road mapping, implementation and continuous improvement/maintenance. We are the only company that possesses deep expertise and capabilities in the full AI spectrum, deep learning (Vision, NLP, etc.), Deep Reinforcement Learning, Operations Research, and many more…
I would describe it as any piece of software that takes care of a task by itself where the software has “evolved” the way to solve/execute by learning from experiences (data), and therefore can continue over time (exposure) to improve itself by executing the task.
It is an approach where the concepts (representation) and what they represent from a math model perspective have been evolved from the problem instead of having been defined by a person. Each representation layers one on top of another that builds on the knowledge of the previous. ex.: a pixel, a shape, an object, a ball, a tennis ball, etc…
Being the AI trusted partner to the Global 2000 in the transformation toward running on AI. Providing the platform and the professional expertise to go through the necessary change management.
For me the density of quality of the talent is clearly a reason, but I think what really sets it apart is the open collaboration happening at the research level. We had a paper at NIPS last years where there were six groups represented, Facebook, Microsoft, Google, DeepMind, Elesment AI and University of Montreal… you will never see that elsewhere.
That anyone using it extensively needs to have, on top of stellar data governance practices, an AI governance team that is very well staffed and established. Biases can build up very easily, transparency on why the model is performing in certain ways is never trivial, as powerful as the tech, there is a lot of new skillsets required to get it to perform at its maximum. Very similar to cars, they need maintenance and tuning, etc. its the same with AI models.
We are starting to see significant progress on the decision-making front (AlphaZero is an example) and combined with the progress we have made on perceptive task, (Vision, Signal processing, Voice, NLP) the opportunities to redefine product and services are immense. We are clearly going through a full replacement cycle on all software systems in the enterprise across the economy in the next 10 years, and I can’t think of a better time for new entrants like Element AI to make their way in and change the current power dynamics.
There are so many people to look to as resources for AI, you can’t possibly follow them all, so here’s a smattering of some favorites.
For the great big pie-in-the-sky aspirations of how AI will change the world, check out the book Life 3.0. A little more focused on the present is Machine Platform Crowd.
The most accessible video is probably Kevin Kelly in this talk here. Here’s a year in review from a research perspective here which is quite good. Probably our favorite video at the company for easily explaining what a neural network is is this one. This special report by The Economist in June 2016 is one of the most sober pieces I’ve read, and has aged relatively well.
Interestingly, recently a site called Canada.ai launched and they are doing a great job curating articles about AI, and is a great place to go get an idea of what’s going on in Canada.
My blog, jfgagne.ai, is a little more focused on those who are already a bit informed, but definitely a place to catch early ideas from someone with a rare vantage point on how the industry is developing around the world.
Bonus: two great newsletters for staying up to speed! Jack Clarke’s at Import AI for latest, significant developments in AI research which short tidbits that sum up what’s important here. And, Machine Learnings, which is good for following the more mainstream (though still informed) dialogue on AI, see it here.
Recognize that you need high-quality data to train your models, no matter what you are going after. This is hard, complex work and must be done with companies that have that data or you need to build it! Building advanced solutions that really use AI, means that the Startup playbook must change from the typical SAAS/cloud/mobile Startup to something new. That means that you need more cash, more time, and more pain until you are going to get it right, but once you do the barrier to entry keeps going up (as your models improve with more data) which puts you in a great position.
CEO & Co-Founder of ROSS Intelligence. International speaker on the subjects of AI, legal technology, & entrepreneurship and has been featured in publications such as The New York Times, BBC, Wired, Bloomberg, Fortune, Inc., Forbes, TechCrunch, the Washington Post, and the Financial Times.