Good morning. It’s Friday, October 11th.

Did you know: Tesla stock is down 8% this morning despite their well-received announcements from the “We, Robot” event last night.

In today’s email:

• OpenAI’s Groundbreaking Benchmark

• Tesla’s ‘We, Robot’ Announcements

• Anduril’s Killer Drones

• 5 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

Can AI really compete with human data scientists? OpenAI’s new benchmark puts it to the test

OpenAI’s MLE-bench puts AI to the test with 75 Kaggle-style data science challenges, assessing its ability to plan, troubleshoot, and innovate. In trials, the o1-preview model, paired with AIDE, achieved medal-level performance in 16.9% of cases. However, the benchmark reveals AI's ongoing struggle with adaptability and creative problem-solving, areas where humans excel.

By making MLE-bench open-source, OpenAI hopes to set a new standard for evaluating AI’s real-world capabilities. While AI shows potential to complement human experts, it continues to face challenges in matching the nuanced decision-making required for complex data science tasks. Read more.

Tesla’s ‘We, Robot’ event: the biggest news and announcements

Tesla Cybercab 

Elon Musk has introduced the Tesla Cybercab, a prototype robotaxi designed for full autonomy without traditional controls like steering wheels and pedals. This sleek vehicle, priced at approximately $30,000, features upward-opening doors and inductive charging.

Musk claims the Cybercab could provide rides for around $0.20 per mile, significantly cheaper than city bus fares. Set for production by 2026, pending regulatory approval, it is touted to enhance safety by reducing accidents 10 to 20 times compared to human drivers. Read more.

Optimus bot 

Several humanoid Optimus robots made their debut, showcasing their ability to perform everyday tasks like delivering packages and watering plants. CEO Elon Musk noted that the robots could handle various functions, such as walking dogs and serving drinks, with a projected price range of $20,000 to $30,000.

Live demonstrations mainly featured the robots waving and interacting with attendees and not really executing complex tasks yet. Musk stated that a more advanced version of Optimus could be available for external sale by next year. Read more.

Robovan

Tesla CEO Elon Musk introduced the Robovan, an unconventional passenger vehicle, at the recent "We, Robot" event. Featuring a sleek, Art Deco-inspired design and no visible wheels, the Robovan can seat up to 20 passengers or transport goods.

Musk said it has the potential to tackle high-density transport needs, projecting travel costs as low as 5-10 cents per mile for services like team transportation. Designed for integration with Tesla’s autonomous ride-hailing service, Tesla Network, the Robovan represents a notable entry into the electric van market, competing with recent offerings from Volkswagen and Mercedes. Read more.

Anduril unveils autonomous AI drones designed for "simple, flexible and lethal precision firepower"

Anduril Industries has introduced two new autonomous aerial vehicles (AAVs) equipped with AI capabilities: the Bolt and the Bolt-M. The Bolt serves as a reconnaissance drone with a range exceeding 20 kilometers, an endurance of over 45 minutes, and a weight of approximately 5.4 kilograms, making it suitable for reconnaissance and search and rescue operations.

In contrast, the Bolt-M, which weighs between 5.9 and 6.8 kilograms, is designed to deliver precise firepower to ground troops and can engage targets from any angle, ensuring high precision in complex environments. Both drones utilize advanced computer vision and machine learning software, allowing operators to track targets from user-defined standoff positions, even when obscured.

The company is also engaged in the U.S. Air Force's Collaborative Combat Aircraft program to develop highly autonomous drones. Read more.

• AI wins another Nobel, this time in Chemistry: Google DeepMinders Hassabis and Jumper awarded for AlphaFold

• Hugging Face makes AI development easier than ever with Gradio 5 release

• OpenAI's GPT-4 matches facial recognition algorithms without explicit training in biometrics

• Tx-LLM: Supporting therapeutic development with large language models

• Meta AI will launch in six more countries today, including the UK 

• NVIDIA Health & Life Sciences Launches Two Downloadable Inference Microservices 

• US plan to break up Google's search dominance threatens profit engine, AI growth 

• AI helps discover more than 160,000 new virus species

• After winning Nobel for foundational AI work, Geoffrey Hinton says he's proud Ilya Sutskever 'fired Sam Altman'

• Zoom will let AI avatars talk to your team for you

• This AI powered ‘tongue’ can tell Coke and Pepsi apart

• Google makes its Imagen 3 image AI available to all Gemini users

• Japanese multimodal AI model Aria is open source and beats many competitors

• DeepMind’s Michelangelo benchmark reveals limitations of long-context LLMs

• Writer’s Palmyra X 004 takes the lead in AI function calling, surpassing tech giants

• Suno Unveils 'Replace Section' Feature for Pro Users to Edit Songs and Add Instrumental Breaks 

• Physicists Reveal a Quantum Geometry That Exists Outside of Space and Time 

• AMD launches AI chip to rival Nvidia's Blackwell

• Meta Unveils Free Course on Multimodal Llama 3.2 AI Model via DeepLearning.AI

📄 Only-IF:Revealing the Decisive Effect of Instruction Diversity on Generalization

📄 Personalized Visual Instruction Tuning  ​​

📄 Aria: An Open Multimodal Native Mixture-of-Experts Model

📄 GLEE: A Unified Framework and Benchmark for Language-based Economic Environments

📄 Pixtral 12B 

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X!