Taking Generative AI into the physical world with Raquel Urtasun from Waabi

12 Nov 2024 (10 days ago)
Taking Generative AI into the physical world with Raquel  Urtasun from Waabi

Waabi and the Autonomous Trucking Industry

  • Waabi is a company developing a fleet of autonomous Long Haul trucks using generative AI, with the goal of bringing level four trucks, requiring no humans, to robustify the supply chain and make roads safer for everyone (1m31s).
  • The company's technology is very general and can be applied to any four-wheeled vehicle, but they chose to focus on trucking first due to the growing driver shortage, pervasive safety issues, and the potential for automation to solve these problems (2m5s).
  • The trucking industry is a trillion-dollar market with a clear need for this technology, making it an attractive area to build a product that can scale quickly (2m46s).
  • Raquel Urtasun, the CEO and co-founder of Waabi, has a background in AI and has worked in the transportation space, including with Uber, and has been working in AI for 25 years and in self-driving for almost two decades (3m8s).

Waabi's Technology and Approach

  • Urtasun started as an academic before moving to industry, where she focused on moving from building cage technology to really pushing the boundaries of what is possible (3m17s).
  • Waabi uses simulation to train their self-driving trucks before taking them out on the road (53s).
  • Replacing Long Haul trucks with autonomous vehicles will not require a large infrastructure overhaul (48s).
  • Raquel Urtasun got her start in academia (46s).
  • The goal is to deploy generative AI in the real world, and an opportunity was identified three and a half years ago to disrupt the traditional approach to autonomous driving, which was manual and capital-intensive (3m22s).
  • A different technology approach was taken with the idea of a single AI system that can solve the necessary tasks for driving, now referred to as Gen AI or foundational model (3m46s).
  • This technology is transformational, and a product launch is planned for next year, which will be much faster, more capital-efficient, and more scalable than others in the market (4m15s).
  • The conviction to start building Waabi came from two decades of research in AI, which provided the necessary compute, data, and next-generation technology to build a skill-level solution (4m43s).
  • Waabi's technology is a different generation of generative AI that is interpretable, provably safe, and suitable for safety-critical applications like self-driving cars (5m22s).

Impact on the Workforce

  • The development of self-driving cars may replace human labor, but it's believed that the workforce will be reformed and deployed in better jobs as the technology scales (6m12s).
  • The motivation behind building this technology is to empower humans to have more humanized jobs, such as those that allow for a better work-life balance, unlike the challenging conditions faced by long-haul truck drivers (6m35s).
  • The development of generative AI technology is expected to create better jobs, such as remote assistance solutions, and will not negatively disrupt the labor market, but rather create more jobs as the technology scales over time (6m53s).

Waabi's Progress and Funding

  • Despite being founded just a couple of years ago, Waabi has made significant progress in scaling its technology, with plans to launch physically next year, thanks to its innovative approach and simulator that can mimic reality (7m31s).
  • Waabi's simulator has enabled the company to train the brain of its autonomous driving vehicle to be much safer and handle safety-critical situations that are impossible to capture in the real world, making it 10 to 20 times faster and more capital-efficient than the rest of the industry (8m3s).
  • Waabi has been well-capitalized with two successful fundraising rounds and is currently testing its self-driving trucks on public roads in Texas, doing commercial loads with some of the largest shippers in North America (8m41s).
  • The company plans to launch its trucks without a human driver next year, following a much safer path thanks to its breakthrough in simulation technology (8m59s).

Testing and Expansion Plans

  • Waabi chose Texas as the location for testing its trucks due to the state's positive regulatory environment, its strategic location for US freight, and favorable weather conditions (9m14s).
  • The company is headquartered in Canada but has a significant presence in the US, with its US headquarters in San Francisco, and plans to expand to other markets, including Canada, in the future (9m26s).
  • The United States is a suitable place to start and expand a self-driving car business due to its economic perspective and the ability to cover the entire country from east to west, north to south (10m52s).

Challenges of Self-Driving and Waabi's Solution

  • Texas, in particular, is a great place to start, despite having traffic and varying weather conditions such as tornadoes, heavy winds, hail, and snow (10m34s).
  • Training self-driving cars to drive through different terrains and weather conditions is a complex challenge that cannot be solved by simply coming up with engineer solutions for every possible scenario (11m35s).
  • The traditional approach to training self-driving cars has been non-scalable and cumbersome, as it is impossible to enumerate all possible scenarios and handle each situation individually (12m17s).
  • Waabi's approach to training self-driving cars involves using a simulator, which is a digital clone of the world, to learn from data and experience without risk (12m53s).
  • The simulator allows the system to learn by experiencing and understanding the consequences of its actions, such as what to do in case of a deer crossing the street or a child on the road (13m2s).
  • Waabi's approach is differentiated from the industry and is more similar to how humans learn (13m18s).

The AI Boom and Waabi's Positioning

  • The AI boom from the last two years has made it easier for Waabi to pitch their idea, as people now understand the concept of generative AI and its potential to revolutionize the physical world (14m3s).
  • The company's AI-first approach, which was not well understood when it was first started, is now more relatable due to the success of models like CH GPT (14m15s).
  • However, the increased awareness of AI has also led to more questions and concerns about the technology, such as hallucinations and safety, which Waabi addresses by explaining the why behind their different type of generative AI technology (14m47s).
  • Waabi's technology is efficient, generalizes to the unknown, and is provably safe, making it an attractive solution for investors and partners (15m0s).
  • The company's founder comes from an AI background, which sets them apart from other founders in the self-driving domain who come from a robotics background (15m20s).
  • The distinction between AI-first companies and others has become more interesting, and Waabi is expected to be one of the companies leading this trend (15m37s).
  • Despite the challenges in the market, Waabi was able to close a massive Series B fundraising round of $200 million last June, which shows that investors are still interested in innovative AI companies (16m5s).
  • The fundraising process was successful because Waabi's technology was seen as a disruptor for the entire self-driving industry and beyond, with tremendous opportunity to build a large company for the physical world (16m40s).
  • Waabi is a disruptor company that has had a successful series of fundraising with an incredible roster of investors, which can be attributed to its unique approach to building self-driving trucks (16m51s).

Risk Mitigation and Safety

  • The concept of risk is a significant consideration when building a company in the self-driving space, as the consequences of failure can be severe, such as harm to people, unlike building a B2B marketing software (17m10s).
  • The risk of not doing anything to improve road safety is also significant, with 30,000 people dying in the US alone each year, and this number is not improving despite technological advancements (18m0s).
  • To mitigate risk, Waabi has elevated safety to the next level, which is referred to as "safety 2.0," by building everything to minimize risk in both development and deployment (18m27s).
  • Waabi's approach is the opposite of the typical approach in the driving industry, where teams quickly put a prototype on the road, instead, Waabi builds a simulator to test and train the vehicle before building it (18m53s).
  • This approach allows Waabi to enhance the safety of its development and make a scientific case to prove that its vehicles will be safer than a human driving a regular truck (19m12s).
  • Waabi's approach is also focused on deploying a small fleet to start commercial operations, rather than driving millions of miles to discover rare cases (19m36s).
  • Investors appreciate Waabi's tough approach to safety and its ambitious mission to build a simulator that is the same as the real world (19m51s).
  • Waabi aims to build a high-performing track from the beginning and achieve this goal in less than two years, which many people initially thought was impossible (20m6s).
  • Despite the skepticism, Waabi has been executing its milestones on time or even earlier than expected, with the same capital as predicted, which is unusual in the industry (20m17s).
  • The company plans to launch driverless trucks next year without compromising safety, and investors appreciate Waabi's ability to hit milestones (20m38s).

Waabi's Deployment Strategy and Advantages

  • To accommodate self-driving trucks, the majority of the industry is thinking of launching a terminal-to-terminal model, but Waabi's technology allows it to go directly to the end customer from day one, removing the need for terminals (21m6s).
  • This approach reduces friction and cost, as the lanes used today are very short and adding a terminal and local shuttle adds significant cost that customers may not be willing to take (21m41s).
  • Waabi's technology is designed to interact with humans safely, reduce risk, and learn from experience, without requiring dedicated lanes (21m59s).
  • The company believes that very little infrastructure is needed to deploy this technology and save lives, and it cannot wait for infrastructure changes to happen (22m20s).

Raquel Urtasun's Journey and Waabi's Founding

  • Raquel Urtasun has a background in academia, where she spent a long time, and she still works in academia while running Waabi (22m32s).
  • She never thought she would launch a company, but her focus on building a new algorithmic concept and an artificial brain that can do tasks that humans can do led her to pursue this path (22m57s).
  • The initial focus was on building and advancing the technology to achieve tremendous positive disruption, which required overcoming significant scientific challenges (23m19s).
  • The goal later shifted to deploying the technology and realizing its positive impact, leading to a transition from academia to industry (23m44s).
  • Joining Uber as an executive provided valuable experience in building a business, learning from the CEO, and understanding what it means to be a great CEO (24m7s).
  • The experience at Uber was instrumental in learning about hyperscaling and building a successful business, which has been fundamental to the success of Waabi (24m27s).
  • The decision to leave Uber and start Waabi was driven by a desire to deploy the technology in a way that was not possible within the existing industry structure (24m35s).
  • Building Waabi allowed for the creation of a company that combines the best people and technology to deploy the technology effectively (24m41s).
  • The journey of building Waabi has been an incredible experience, and the company is poised to transform the world (25m12s).
  • The transition from an academic background to entrepreneurship was facilitated by the experience at Uber, which provided valuable insights and skills (25m26s).
  • The decision to start Waabi was driven by the realization that it was not possible to transform an existing company like Uber into something new, and that a new approach was needed (26m20s).

Waabi's Unique Approach and Future Plans

  • Many companies are struggling to fully adopt a generative AI approach due to a lack of mindset, people, and structures to support this direction, which is why Waabi was built with a different DNA, rooted in bringing research and engineering together into a single team (26m47s).
  • Waabi's approach is different from other companies in the industry, making it impossible for publicly traded companies to replicate without significant changes, which can be challenging for investors (27m9s).
  • Waabi is planning to launch its generative AI technology in the freight and trucking industry next year, with the goal of driving trucks at scale in North America and potentially beyond (27m38s).
  • In the next five years, Waabi plans to expand its technology to other form factors, such as robot taxis, warehouse robotics, or humanoid robots, which can help with daily tasks (27m46s).
  • The technology developed for trucking can be reused on these new form factors, allowing for quick deployment and enabling opportunities to enhance people's lives, especially with the aging population (28m5s).

Infrastructure Considerations and Human Element

  • The idea that replacing long-haul trucks with autonomous vehicles requires a large infrastructure overhaul is not necessarily true, as autonomous trucks can still operate within existing infrastructure, such as pulling into bays (28m44s).
  • However, adapting to new technologies like robot trucks may still require some effort and time, especially in different markets and states with varying infrastructure (29m37s).
  • One potential aspect to consider is that while autonomous trucks may replace human drivers, they may still require more people to support their operation (30m1s).
  • The trucking industry is not ideal for human drivers due to its negative aspects, such as being away from family, low pay, and health risks, making it a suitable industry for technological replacement (30m19s).
  • Replacing human labor with technology in this industry could be a thoughtful way to move people into more meaningful and better-paying jobs, such as working on the infrastructure side or in terminals (30m51s).
  • Upskilling is not necessarily a bad thing, as it could lead to people getting into better-paying jobs, and the trucking industry already struggles with a lack of interested workers (31m1s).

Waabi's Generative AI Focus and Potential

  • The company's approach to building trucks, focusing on being a generative AI company rather than a self-driving car company, sets them up for potential expansion into other categories in the future (32m26s).
  • This approach is distinct from other companies in the self-driving car category, which are essentially AI companies, and is likely influenced by the company's AI background rather than a robotics background (32m45s).
  • The trucking industry is not an industry with an oversupply of people interested in doing the work, so the introduction of AI technology is not likely to displace a large number of existing workers (31m47s).
  • The company's focus on generative AI could allow them to expand into other areas beyond self-driving trucks, making their approach more versatile and adaptable (32m32s).

Waabi's Background and Testing Approach

  • The company Waabi is working on developing autonomous trucks, with a background in AI rather than mobility or transportation, which may bring a fresh perspective to the industry (32m58s).
  • Waabi's approach to training their trucks involves using online simulators to create various scenarios and virtually driving the trucks multiple times before testing them on the road (33m21s).
  • This approach is encouraging, as it shows that the company is taking a thorough testing process, unlike some other AI companies that may seem to be rushing to put their products on the road without proper testing (33m44s).
  • Waabi is currently conducting simulations in Texas and plans to start testing on the road next year, which will be an interesting development to follow (34m18s).
  • The company's focus on testing and safety measures is reassuring, especially considering the challenges of autonomous vehicles, such as dealing with unexpected situations like a tire bursting (35m6s).
  • In the event of a tire bursting, the autonomous truck is expected to automatically sense the issue and pull over, but it raises questions about how the situation would be handled without a human driver present (35m10s).
  • The development of autonomous trucks is a complex issue, and it will be important to see how companies like Waabi address the challenges that arise during testing and deployment (34m22s).

Overwhelmed by Endless Content?