Element AI set to release its first products
A little more than a year and a half after it raised over $100 million in venture capital, Montreal’s most prominent artificial intelligence company is preparing to release its first products.
“We’ve been on a long journey, building all the technology and figuring out all the right-use cases for us to start growing the business,” said Jean-François Gagné, CEO and co-founder of Element AI. “We’re now at the point where all of these products are becoming generally available, some in the first quarter of this year and some in the second quarter of this year.”
Element AI was started in the fall of 2017 by Gagné, Nicolas Chapados, Montreal venture capital fund Real Ventures and Université de Montréal AI scientist Yoshua Bengio, who pioneered the development of deep learning, an artificial intelligence technique that led to many of the recent breakthroughs in the field.
Element AI’s first products will be aimed at large companies — it hopes to win customers in the financial sector, like big banks, insurance companies and asset managers. It’s also aiming at companies in transportation and logistics.
One product is currently being tested with the Port of Montreal.
There, Element AI is working to make better predictions about how long trucks will have to wait before they pick up or drop off goods at the port.
It’s just one area where AI is increasingly being used.
“If you want to deploy any highly sophisticated piece of software, you will probably end up using some shape or form of artificial intelligence,” Gagné said.
But access to that technology can be difficult, he said, and many companies are using pre-trained AI models.
(Many contemporary AI applications, like machine learning, are “trained” by being fed a large amount of data. At the beginning, the trainer knows what the output should be and adjusts the system until it delivers the expected results — that sets the “rules” that the system will operate by.)
According to Gagné, pre-trained models, like traditional software, don’t give businesses enough flexibility in competitive, changing markets.
“Their rules are hard-coded or there’s a very long, long complex list of parameters that people need to switch on and switch off, that’s a very unnatural way of using systems,” he said.
Element AI’s products will be different, Gagné said, because regular users — people like insurance adjusters, truck dispatchers and cyber-security analysts — at the companies who buy the products will train the systems as they do their jobs.
“As they utilize the system over time, the system gathers more and more knowledge about their organization and how they want to operate and it improves in performance,” he said.
The company’s AI products will be able to do that because they’re made up of “core capabilities” — artificial intelligence “building blocks” that can be combined together in different ways.
For example, a “building block” that can read text (including handwriting) can be combined with a natural language processing block to “read” a document. For example, it can be used to identify the merchant, date and final total on a receipt but that same combination of “blocks” could also be used to read a mortgage application, insurance claim or a bill of lading, said Jerome Pasquero, a product manager at Element AI.
The company says it doesn’t want to replace people with machines.
“What’s really, really important for us is that there’s always a human in the loop, we’re not aiming at replacing, we’re aiming at augmenting,” said Christophe Coutelle, Element AI’s vice-president of marketing. “It’s like a very smart assistant that is able to help you and do all the laborious tasks or be able to take into account a bigger amount of data.” jserebrin@postmedia.com