Daily AI Newsletter: Millions forced to use brain as OpenAI's ChatGPT takes morning off

In today’s email:

🧑🏻‍⚕️ New AI System Detects Rare Epileptic Seizures
👀 Elon Musk reportedly diverted thousands of AI chips from Tesla to X
🌦️ New model that predicts global weather can run on a single desktop computer.
🧰 12 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.

AI apocalypse? ChatGPT, Claude, and Perplexity all went down at the same time

In the early hours of June 4, 2024, OpenAI’s ChatGPT, along with Anthropic’s Claude and Perplexity, experienced significant outages. The disruptions started around 7:33 AM PT, with ChatGPT going down, followed by issues with Claude and Perplexity, though the latter two were resolved more quickly. Users speculated that a broader infrastructure issue or excessive traffic caused these outages. Google’s Gemini also faced brief issues, according to some user reports.

During the outages, ChatGPT’s landing page indicated the service was at capacity, even updating the message to pirate speak, while Claude’s website displayed an error message about server components. Perplexity’s site showed an over-capacity message, stating, “We’ll be right back.” Both Claude and Perplexity started functioning again around 12:10 p.m. ET, though Perplexity experienced intermittent issues thereafter.

ChatGPT had just recovered from a lengthy overnight outage when the new issues began. OpenAI’s status page noted a major outage and the company’s efforts to investigate and resolve the problem. By 10:17 AM PT, ChatGPT’s service was restored, ending a multi-hour disruption.

Listen to a song about this 👆🏼

Raspberry Pi AI Kit available now at $70

The Raspberry Pi AI Kit, now available for $70, is a powerful addition for those interested in neural networks, AI, and machine learning on the Raspberry Pi 5. Developed in collaboration with Hailo, the kit features the Raspberry Pi M.2 HAT+ board preassembled with a Hailo-8L AI accelerator module. This combination allows for high-performance, power-efficient inferencing suitable for a variety of applications, including real-time AI vision. Key features include 13 tera-operations per second (TOPS) of inferencing performance, a single-lane PCIe 3.0 connection running at 8Gbps, and compatibility with various cameras. The M.2 HAT+ board and Hailo-8L module enhance the Raspberry Pi 5's capabilities, enabling users to build complex AI applications with low latency.

A major highlight of the AI Kit is its software ecosystem, which simplifies the integration of AI frameworks with the camera subsystem. The rpicam-apps suite of camera applications includes a post-processing template for real-time neural network inferencing. The kit also comes with Hailo Tappas post-processing libraries, allowing users to create advanced AI-based applications with minimal code. Additionally, Hailo offers a model zoo with a wide variety of pre-trained neural network models optimized for the AI Kit. Future integration with the Picamera2 framework will further streamline AI application development on the Raspberry Pi.

Setting up the AI Kit is straightforward, involving a few package installations and a reboot. Once set up, users can quickly start running AI demos, such as object recognition and pose estimation, through the rpicam-apps. The Hailo-8L co-processor can also be used in non-camera applications, thanks to the packaged API integrated with the GStreamer framework and support for native Python or C/C++ applications. This flexibility makes the Raspberry Pi AI Kit a versatile tool for AI enthusiasts and developers looking to enhance their projects with advanced AI capabilities.

Elon Musk reportedly diverted thousands of AI chips from Tesla to X

Elon Musk has reportedly redirected thousands of Nvidia-made AI chips from Tesla to his social media company, X, according to emails obtained by CNBC. This diversion could delay Tesla's acquisition of $500 million worth of processors, which are crucial for its transformation into a leader in AI and robotics. Despite Musk's statements about increasing Tesla's AI chip purchases to 85,000 by the end of the year, Nvidia emails suggest that many of these chips are being rerouted to X and its AI subsidiary, xAI. The reallocation includes 12,000 H100 GPUs originally meant for Tesla being redirected to X.

Musk's decision to prioritize X over Tesla could upset Tesla investors, who are counting on his promise to deliver fully autonomous vehicles. Tesla plans to reveal its first robotaxi in August, yet its Autopilot and Full Self-Driving features have faced scrutiny due to numerous crashes, some fatal. Musk defended the chip diversion by stating that Tesla's Austin factory isn't ready to handle the GPUs, estimating a $3-4 billion spend on AI chips from Nvidia in 2024. Meanwhile, xAI is competing with companies like OpenAI and Google to develop advanced generative AI applications, having recently raised $6 billion in funding.

Nvidia, driven by the high demand for its GPUs, has become the third most valuable company globally. CEO Jensen Huang highlighted the immense demand for GPUs in cloud computing and generative AI, noting that customers are consuming every available GPU. Nvidia reported a 200 percent revenue growth in the last quarter, reflecting the booming AI industry and the critical role Nvidia's chips play in powering these technological advancements.

New model that predicts global weather can run on a single desktop computer.

Artificial intelligence is revolutionizing weather forecasting with models like WeatherMesh and AIFS, which can run on a desktop computer. Leveraging the rich ERA5 dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF), these AI models have shown impressive accuracy, sometimes surpassing traditional physics-based models that require supercomputers. WindBorne Systems, co-founded by Stanford graduates, developed small, long-lasting weather balloons to gather atmospheric data globally, further enhancing AI forecasting capabilities.

AI weather models learn by ingesting vast amounts of historical weather data, recognizing patterns, and predicting future conditions. This method has proven effective, especially in tracking hurricane paths, although intensity prediction still lags behind traditional models. While physics-based models remain vital, AI's rapid advancements suggest a future where data assimilation and forecasting could be fully AI-driven, potentially revolutionizing the field within a decade.

As the 2024 Atlantic hurricane season begins, these new AI models are being closely evaluated. Although not yet used for official forecasts by the National Hurricane Center, they offer promising supplementary tools for meteorologists. The real-time performance of models like AIFS and WeatherMesh during this hurricane season will be crucial in determining their future role in operational weather forecasting.