Radical Reads

Solving self-driving

By Jordan Jacobs, Co-Founder & Managing Partner


Photo Source: Waabi, Photo by Natalia Dolan

This week AI pioneer Raquel Urtasun announced the launch of Waabi and her goal of building the next generation of autonomous driving technology.

When we started Radical Ventures, we set out to invest in extraordinary people working to solve the world’s biggest challenges. We are very proud to continue our long-time relationship with Raquel and to invest in Waabi’s work to change the world for the better.

Raquel is a renowned expert in deep learning, computer vision, and autonomous driving. She most recently was Chief Scientist of Uber Advanced Technologies Group (ATG) and led the team in Toronto. Raquel is also a Professor at the University of Toronto and a co-founder of the Vector Institute for AI alongside Geoffrey Hinton, Richard Zemel, and other amazing professors (plus Tomi, Ed Clark and me).

Waabi’s breakthrough, AI-first, approach was developed by a team of world leading technologists who followed Raquel from Uber ATG to found Waabi. Using deep learning, probabilistic inference, and complex optimization, the team has created software that is end-to-end trainable, interpretable, and capable of highly complex reasoning. Raquel has already assembled a team of nearly fifty engineers and researchers that will initially focus on deploying Waabi’s software in logistics, specifically long-haul trucking, an industry where self-driving technology stands to make the biggest and swiftest impact due to a chronic driver shortage and pervasive safety issues.

Waabi means “she has vision” in the Ojibwe language, and wabi refers to the concept of ‘beauty in simplicity’ in Japanese culture, which are apt descriptions respectively for Raquel and Waabi’s work. Raquel’s decision to launch Waabi in Toronto, Canada reflects the significant momentum in the AI ecosystem, and reinforces Canada’s position as home to world-leading talent. The existence of Waabi in Toronto is also an affirmation of the vision we laid out in 2016 centred around building the Vector Institute for AI and the Pan-Canadian AI Strategy we (and especially Tomi) helped write, and which the federal government adopted in 2017 as the world’s first national AI strategy. Our belief was that governments should focus on supporting basic research and education through the establishment of ambitious research institutes/graduate schools that can anchor AI ecosystems and become magnets for talent, businesses, and capital, which together can create an ecosystem flywheel.

Raquel was born in Spain, studied at MIT and UC Berkeley, and chose Toronto as her home based on the talent and diversity of the community. When Raquel joined Uber ATG and it became a sponsor of Vector, there was criticism in some corners about her decision to join a foreign company and its sponsorship of Vector. We believed then that this view misunderstood reality, which was that Raquel is determined to solve autonomy and had many offers to leave Toronto to do so. We argued that keeping Raquel, her pipeline of students and building a world-class team around them would preserve the opportunity to build one of the most important industries of the future in Canada, and inevitably result in new startups spinning out of that group. Four years later there are multiple startups launched by Uber ATG Toronto alumni, with Waabi being the most prominent.

As we discussed in the Globe and Mail this week, Raquel continues to play an integral role in the Canadian and global AI ecosystem, drawing new AI talent to join her mission. We believe Waabi is one of the few teams capable of solving the  challenge of autonomous driving. In a tech ecosystem where the flywheel is really flying (marked lately by the seemingly daily announcements of a new Canadian tech unicorn) Raquel’s decision to build Waabi — and to do so in Canada — is a signal to the world and worthy of celebration.

AI News This Week

  • US Launches Task Force to Open Government Data for AI Research   (The Wall Street Journal)

    The Biden administration launched an initiative to make more government data, including health care and driving records, as well as demography, available to artificial intelligence researchers. The National Artificial Intelligence Research Resource Task Force is made up of a dozen members from government, industry, and academia including Fei-Fei Li – a Stanford professor and founder of Dawnlight, a Radical Ventures portfolio company. There is a growing effort across the US government to ensure the country remains at the vanguard of technological advancements.

  • DeepMind’s path to AGI  (VentureBeat)

    Despite Hollywood’s depictions of superintelligent machines, creating human-level, generalizable, AI remains an elusive challenge. Some researchers have been using specific processes to replicate intelligent abilities such as vision, language, reasoning, and motor skills. In contrast, scientists at UK-based AI lab DeepMind submitted a new paper to the peer-reviewed Artificial Intelligence journal, arguing that the simple but powerful principle of reward maximization with trial-and-error exploration will be enough to allow artificial general intelligence to emerge.

  • The current state of affairs and a roadmap for effective carbon-accounting tooling in AI | Sustainable Software   (Microsoft DevBlogs)

    AI’s impact on climate change is a growing dilemma. On the one hand, significant energy is required for training AI systems. On the other hand, AI can create efficiencies in industrial processes to help reduce emissions. Last week, Abhishek Gupta of the MAIEI wrote a research summary of an influential research paper that explores this conundrum. The researchers propose four metrics that are crucial to understanding the impact of specific computations: the training server location, the server’s energy grid, the training duration, and the hardware’s make and model. The researchers also produced the next iteration of their web-based ML CO2 calculator tool called CodeCarbon which is integrated into the computation workflow. This allows operators to address low hanging fruit that can immediately decrease related emissions. These metrics and tools were covered in a previous edition of Radical Reads.

  • When AI becomes child’s play   (MIT Tech Review)

    Kids say the darndest things! Researchers designing voice systems for children’s entertainment are seeing that algorithms trained in pattern recognition struggle to understand kids. Children are notoriously inconsistent, which is challenging for recognition algorithms looking for patterns and consistency. More parents are turning to audio to cut down on screen time, especially for young children, but tech companies have yet to develop a system with sufficient contextual awareness to successfully interact with children.

  • Six of the smartest applications of AI in business  (Forbes)

    “Nearly 90% of CEOs confirmed that AI is considered mainstream technology in their offices in 2021.” While these applications are useful, companies considering implementing AI should aim to reimagine an entire core business processes, journey, or function end to end.

Radical Reads is edited by Leah Morris (Senior Director, Velocity Program, Radical Ventures).