Transforming the Middle Mile: How AI is Revolutionizing Logistics with Gatik

Table of Contents

1. Introduction
2. The Story Behind Gatic AI
3. The Growth of Gatic AI
4. The Hype Surrounding AI
5. The Use of ML Techniques in Autonomous Vehicles
6. Challenges with ML Techniques in Autonomy
7. Observability and Transparency in AI
8. Conclusion

Introduction

Welcome to the future of supply chain, presented by Dynamo. In each episode, we sit down with leaders in the industry as we build the future of the supply chain together. In this episode, our host Santosh Sankar interviews Apeksha Kamava, the co-founder and chief engineer at Gatic AI. They delve into the world of AI intersecting supply chain and discuss the story behind Gatic AI, the growth of the company, challenges with ML techniques in autonomy, and the importance of observability and transparency in AI.

The Story Behind Gatic AI

Apeksha Kamava starts by sharing her background in robotics and AI, specifically in the navigation of robots. She explains the challenges of navigating different environments and the intersection of machine learning in identifying behaviors of agents around robots. Apeksha, along with her co-founders Surgeon and Gotham, all come from a robotics background, bringing a mix of experience in AI and conventional robotics. They founded Gatic AI in 2017 with a vision to develop autonomous vehicles for short-haul logistics, focusing on bringing the technology to a near-term, realistic application that solves customer pain points.

The Growth of Gatic AI

Since its founding in 2017, Gatic AI has experienced significant growth. The company now boasts a team of over 130 people spread across different states in North America. They have successfully deployed their autonomous vehicles in several markets, including Texas, Toronto, and Arkansas. Moreover, Gatic AI has collaborated with major customers such as Walmart, Lowe's, and Kroger, gaining valuable experience and insights from these deployments. With each deployment and advancement, Gatic AI continues to mature its technology and moves closer to productization.

The Hype Surrounding AI

Apeksha acknowledges the hype surrounding AI, particularly with recent advancements such as GPT and Transformers. She highlights that while there is value in these technologies, the hype often exceeds the actual capabilities of the models. The Perception of AI as a tool that can solve everything or achieve sentience is misleading. Apeksha emphasizes the need for a more realistic understanding of AI and its near-term benefits without overpromising on what it can currently achieve.

The Use of ML Techniques in Autonomous Vehicles

Apeksha explains how ML techniques are utilized in delivering an autonomous experience in vehicles. She breaks down the three key components: sense, plan, and act. Sensors and cameras perceive the vehicle's surroundings, while ML algorithms classify and identify objects such as vehicles, pedestrians, and traffic lights. The vehicle then plans the appropriate action based on this information and executes it. Apeksha further distinguishes between level 2/3 autonomy, where the driver is responsible for taking over when needed, and level 4 autonomy, where the vehicle can handle all situations without driver intervention.

Challenges with ML Techniques in Autonomy

While ML techniques have improved perception and decision-making in autonomous vehicles, Apeksha highlights the challenges. Deep learning models require vast amounts of data for training and lack explainability in their decision-making process. She describes the limitations of data-driven approaches, stating that although they often produce correct outcomes, understanding why they make certain decisions is difficult. To address this, Gatic AI employs a hybrid architecture that combines deep learning models with deterministic conventional methods. This allows for better observability and explainability, reducing the reliance on massive amounts of data.

Observability and Transparency in AI

The topic of observability and transparency in AI arises, discussing the need for creators to prioritize these aspects. Apeksha acknowledges that current models, like Transformers, lack sufficient observability and explainability. While there have been attempts to Visualize attention networks, it remains a challenge to maintain explainability as models Scale up and require more data. Observability and transparency should be driven by the specific application and its tolerance for inaccuracies. For highly critical applications like autonomous vehicles, ensuring a high level of certainty is paramount.

Conclusion

In conclusion, Apeksha Kamava provides valuable insights into the story behind Gatic AI, the growth of the company, the hype surrounding AI, and the challenges faced in using ML techniques in autonomous vehicles. She emphasizes the importance of observability and transparency in AI, encouraging creators to prioritize these aspects. As Gatic AI continues to advance its technology and deploy autonomous vehicles in various markets, it remains committed to solving customer pain points and delivering safe and efficient autonomous solutions.

Highlights:

• Apeksha Kamava shares the background and vision behind Gatic AI, a company focused on developing autonomous vehicles for short-haul logistics.
• Gatic AI has experienced significant growth since its founding and has deployed its autonomous vehicles with major customers such as Walmart, Lowe's, and Kroger.
• While there is value in recent advancements in AI, the hype surrounding technologies like GPT and Transformers often exceeds their actual capabilities.
• ML techniques play a crucial role in enabling autonomous vehicles, particularly in perception and decision-making systems.
• The challenges with ML techniques include the requirement for massive amounts of data and the lack of explainability in decision-making processes.
• Gatic AI employs a hybrid architecture that combines deep learning models with deterministic conventional methods to improve observability and explainability.
• Observability and transparency are crucial considerations for AI creators, depending on the specific application and its tolerance for inaccuracies.