🛸Drone Races Train Space AI

How drone racing is helping train AI for autonomous space missions

Key Points 

• Neural networks in AI function like human brains, which is promising for autonomous space missions involving complex spacecraft maneuvers.

• Drone racing has been selected as the ideal test environment on Earth to validate AI on robotic platforms for future space applications.

🛰️News - The European Space Agency's Advanced Concepts Team (ACT) has teamed up with Delft University of Technology's Micro Air Vehicle Laboratory in the Netherlands to train neural network AI systems to race drones. The goal is to test AI's capability to manage complex maneuvers, which may be used in future space missions.

🤔Why use AI? Neural networks in AI function like human brains by transmitting information between "neurons." This process is promising for autonomous space missions involving complex spacecraft maneuvers. 

Traditionally, humans plan maneuvers on Earth and upload directions to spacecraft. However, with neural-network AI, spacecraft could autonomously adjust their courses using real-time data, making rapid corrections in response to the unpredictable nature of space.

🥸Insider's perspective - Dario Izzo, scientific coordinator of ESA's ACT, explained that in space, every onboard resource must be used as efficiently as possible, including propellant, energy, computing power, and time. He noted that a neural network approach could optimize onboard operations, thereby boosting mission autonomy and robustness.

However, as with all space mission components, this technology must first be tested on Earth. Izzo shared that this is why they have chosen drone racing as the ideal environment to test neural architectures on real robotic platforms, aiming to build confidence in their future use in space.

Former OpenAI, Tesla exec Andrej Karpathy launches AI education platform

Key Points 

• The San Francisco-based company plans to leverage the latest advancements in generative AI to build teaching assistants that support student learning alongside human teachers.

• Eureka Labs' first product, LLM101n, is an undergraduate-level course designed to help students train their own AI.

☕News - Andrej Karpathy, a former exec at Tesla and OpenAI, has launched Eureka Labs, a new education platform that's fully centered around AI technology. 

🕵🏻‍♂️What's the vision? The San Francisco-based company plans to leverage the latest advancements in generative AI to build teaching assistants that help students navigate their coursework.

According to Karpathy, these AI assistants or personalities would collaborate with human teachers, creating a learning environment where, as Karpathy suggests, "anyone can learn anything." This means teachers would still design course material, but they'd be supported by these AI assistants.

Karpathy's post also suggests a future where these assistants could emulate real individuals, akin to Meta’s celebrity chatbots or Character AI’s character chatbots.

🤓What's more? Even though Eureka Labs' goal is to develop AI teaching assistants, Karpathy mentioned their first product will be LLM101n, an undergraduate-level course. This course will help students train their own AI, a smaller version of the teaching assistant Eureka Labs hopes to eventually scale up.