In this digest, we’ll get a quick look at the most exciting updates in AI and tech: OpenAI’s groundbreaking self-checking AI model, key executive departures, and an exciting collaboration on a new AI device. Plus, you’ll explore Meta’s leap into augmented reality with its advanced AR glasses and Uber’s moves to expand its autonomous vehicle services.
AI
Meta Connect 2024: new products and AI innovations unveiled
At the Meta Connect 2024 event, which occurred on September 25-26, Meta revealed a range of new products and updates in VR, AR, and AI. The biggest announcement was the Meta Quest 3S, a more affordable mixed-reality headset. Designed for entry-level users, it offers high-quality VR experiences with a new Qualcomm processor and upgraded Horizon OS.
Meta has also unveiled Orion, their first true augmented reality glasses, which blend the look of regular glasses with immersive AR capabilities. They are already being hailed as the most advanced AR glasses ever created. Orion represents a significant advancement, offering large holographic displays and personalized AI assistance, all within a lightweight and stylish design. The glasses allow users to perform tasks like video calling, managing calendars, and accessing digital content, all while staying engaged with the physical world around them. Currently available to Meta employees and select testers, Orion marks a key step toward Meta’s future consumer AR products.
Additionally, Ray-Ban Meta Smart Glasses received software updates, including live language translation, QR code scanning, and Spotify integration. Overall, Meta Connect 2024 highlighted Meta’s commitment to making immersive technology more accessible and integrated with daily activities.
OpenAI’s September shifts: progress and challenges
OpenAI launches o1: a self-fact-checking AI model
OpenAI recently introduced its new generative AI model family, known as “o1,” which includes two versions: o1-preview and o1-mini. One of the standout capabilities of o1 is its ability to fact-check itself. It can take additional time to analyze a question before generating a response, which helps improve the accuracy and relevance of its answers. The model is designed to perform better in reasoning tasks, allowing it to tackle complex problems by synthesizing multiple pieces of information. This makes it suitable for tasks like legal analysis and coding.
Although Altman suggested that o1 learns underlying concepts from data, experts caution that current AI models primarily identify statistical patterns rather than genuinely understanding concepts. o1 has been trained using reinforcement learning, which involves rewarding the model for correct answers and penalizing it for incorrect ones. This approach encourages better performance over time.
OpenAI claims that o1 outperforms previous models, such as GPT-4o, especially in areas like data analysis, science, and multilingual capabilities. For instance, o1 reportedly achieved high scores on math competition problems and performed well in coding challenges. Despite its advancements, o1 is not without drawbacks. It can be slow in generating responses and is still prone to errors and “hallucinations,” where it might generate incorrect information confidently.
OpenAI intends to enhance the o1 model’s capabilities further, possibly experimenting with versions that can reason over extended periods. The company is also navigating regulatory challenges, pushing for government involvement in AI safety testing while facing criticism regarding its stance on specific legislation.
OpenAI transforms the safety committee into an independent oversight board
Next, OpenAI has announced that its safety and security committee will transition into an independent board oversight group chaired by Carnegie Mellon professor Zico Kolter. This new structure will focus on critical safety measures related to model development and deployment, including overseeing safety assessments for OpenAI’s latest AI model, o1. The committee will have the authority to delay model releases until safety concerns are adequately addressed and will enhance cybersecurity through internal measures and industry collaboration for sharing threat intelligence.
Additionally, OpenAI aims to increase transparency about its safety practices by publishing detailed system cards that outline model capabilities and risks. The company plans to pursue independent safety assessments and collaborate with external organizations and government agencies to establish industry-wide safety standards.
Shocking departures: Top Executives at OpenAI step down
In a shocking turn of events, OpenAI’s Chief Technology Officer, Mira Murati, announced her departure from OpenAI after over six years. In a heartfelt post on X, Murati expressed her decision to leave as a difficult but necessary step for her personal exploration. “There’s never an ideal time to step away from a place one cherishes, yet this moment feels right,” she shared, reflecting on her extraordinary privilege working with the OpenAI team.
Murati’s exit comes just a week before OpenAI’s annual developer conference, raising eyebrows given her significant role in the company’s projects, including ChatGPT and DALL-E. CEO Sam Altman responded to her announcement with gratitude for her contributions, highlighting her support during challenging times. Her departure adds to a series of recent high-level exits at OpenAI, including co-founder John Schulman, who left to join rival Anthropic in August 2024, president Greg Brockman taking extended leave, and former product manager Peter Deng’s earlier exit. Interestingly, Durk Kingma, one of the co-founders of OpenAI, has announced he’s joining Anthropic, where he’ll work remotely from the Netherlands.
But surprises just began. Following Mira Murati’s departure, Chief Research Officer Bob McGrew and VP of Research Barret Zoph announced their departures just hours after CTO Mira Murati’s exit. CEO Sam Altman confirmed the resignations of Bob McGew and Barret Zoph in his post on X, emphasizing that the three execs made independent decisions but coordinated their timing to ensure a smooth transition. This wave of departures follows a string of high-level exits, leaving only three of the original 13 founders at OpenAI. As Altman puts it, leading at OpenAI is “all-consuming” amid rapid growth and relentless challenges.
Does Sam Altman plan to shift OpenAI to a for-profit structure?
So, what is OpenAI’s closest future? In the spotlight of multiple departures of top people at OpenAI, including CTO Mira Murati, news has surfaced that CEO Sam Altman is poised to receive equity in the company for the first time. OpenAI reportedly plans to transition from its nonprofit structure to a for-profit benefit corporation, similar to competitors like Anthropic and Elon Musk’s xAI.
According to Reuters, this restructuring will allow the nonprofit board to retain a minority stake while relinquishing control, making OpenAI more appealing to investors concerned about the company’s return cap. However, this move raises questions from the AI safety community about the company’s ability to maintain accountability as it advances toward superintelligent AI. The exact details and timeline for the transition are still being worked out.
Sam Altman collaborates with Jony Ive to create an innovative AI device
According to Jony Ive’s New York Times profile, he is collaborating with Sam Altman, CEO of OpenAI, to create a new AI-powered device. The project reportedly grew out of casual dinners between the two last year, and while many specifics about the product remain undisclosed, the aim is to develop a device that offers a more socially harmonious computing experience than current smartphones, like the iPhone.
Ive’s design firm LoveFrom, is spearheading the creative aspect of the product, and industrial designer Marc Newson is also involved. Though the startup’s name and launch timeline are still unknown, they are in the process of fundraising, with contributions from Laurene Powell Jobs’ Emerson Collective and Ive himself. The company aims to raise up to $1 billion by the end of the year.
Robotics
Rabbit’s r1 to get major LAM update
Rabbit, a tech company known for its r1 device, which features a web-based “Large Action Model” (LAM) agent, is set to release a significant update on October 1. Rabbit aims to improve the r1’s capabilities, allowing it to perform a wider range of tasks on any website. CEO and founder of Rabbit Inc. Jesse Lyu admitted that the company initially set expectations too high since the r1’s debut in early 2024. The upcoming update will introduce a generic version of the LAM agent that can break down tasks — such as buying concert tickets or registering a website — and execute them by interacting with web elements.
Despite earlier limitations, the update promises a more versatile experience. However, there are ongoing concerns about the agent’s performance, user data privacy, and the lack of a clear “killer app.” Lyu defended Rabbit’s choice to focus on a web-based platform instead of a traditional app, emphasizing that this approach allows for broader functionality while avoiding conflicts with major tech companies like Apple and Google. While the LAM agent shows potential, Lyu cautioned users that it remains a work in progress.
Uber expands its autonomous fleet with WeRide in Abu Dhabi and Waymo in U.S. cities
With two significant partnerships, Uber is making major moves in the autonomous vehicle space. The ride-hailing giant recently announced a deal with WeRide to bring its robotaxis to Abu Dhabi later this year. This is part of Uber’s broader strategy to integrate autonomous vehicles into its platform. WeRide, a Chinese AV company, received a license to test its vehicles in the UAE and is now pushing into the Middle Eastern market.
In addition, Uber is expanding its collaboration with Waymo, allowing users in Austin and Atlanta to hail Waymo robotaxis starting in early 2025. This partnership follows their successful launch in Phoenix in 2023. As Uber continues to secure partnerships with autonomous vehicle companies like WeRide and Waymo, it’s positioning itself as a leader in the autonomous ride-hailing sector.
Nuro changes strategy to sell self-driving technology instead of operating delivery vehicles
Nuro, a self-driving vehicle startup created by Jiajun Zhu and Dave Ferguson, is pivoting from owning and operating a fleet of autonomous delivery bots to licensing its autonomous driving technology to automakers and mobility companies. The shift comes after financial strains from maintaining its fleet. Nuro will now focus on offering Level 4 autonomous driving products and partnering with manufacturers for consumer vehicles with automated driving systems. This strategy aims to make Nuro’s business more sustainable and extend its operational runway. Although it hasn’t signed new partnerships yet, Nuro has existing relationships with Uber and Toyota.
