According to the Bloomberg article, both Microsoft and Apple are dropping their plans to join the OpenAI board due to increased scrutiny from global watchdogs over big tech's influence in AI.

Regulators in the US and Europe have expressed concerns about Microsoft's sway over OpenAI, as Microsoft already owns 49% of the company and has integrated OpenAI's services into its own platforms. The article states that the board resignation is unlikely to resolve the US Federal Trade Commission's concerns about Microsoft's partnership with OpenAI.

Microsoft had recently acquired Anthropic, another prominent AI company, in a move that seemed to fly under the radar by not officially acquiring the company but rather hiring its CEO Mustafa Suleyman. This suggests Microsoft is trying to navigate the increasing regulatory scrutiny around big tech's consolidation of AI capabilities.

The article notes that the UK is also looking into Amazon's $4 billion collaboration with Anthropic, the company behind the impressive AI model Claude. Regulators appear to be concerned about the concentration of power in the hands of a few major tech companies when it comes to AI development and deployment.

Overall, the departure of Microsoft and Apple from the OpenAI board seems to be a strategic move to distance themselves from the perception of an "incestuous" relationship between big tech companies in the AI space, in an effort to appease regulators and promote more competition in the field.

Anthropic, the company behind the impressive AI model Claude, has recently released several new features for their flagship product. Let's dive into the details:

1. Fine-tuning for Haiku: Anthropic has now enabled fine-tuning for their Haiku model, which is their fastest and most cost-effective model in the Amazon Bedrock platform. This allows users to fine-tune the model to better suit their specific needs, such as acting as a moderator on an example data set. In testing, fine-tuning improved classification accuracy from 81% to 99% and reduced prompt tokens per query by 89%.

2. Artifact Publishing: Anthropic has expanded the capabilities of Claude's "Artifact" feature. Artifacts are a unique UI element that allows Claude to output lengthy responses, such as code snippets or visualizations, in a separate window while keeping the main conversation clean. Users can now publish and share these artifacts online, making it easier to collaborate and showcase their work.

3. Concurrency Enabled by Default: Anthropic has released version 0.2 of their open-source library, Llama, which now includes concurrency enabled by default. This allows Llama to serve multiple requests simultaneously, using only a small amount of additional memory per request. This enables use cases such as handling multiple chat sessions, hosting code completion LLMs, processing different parts of a document in parallel, and running multiple agents with different models at the same time.

These new features from Anthropic demonstrate their commitment to continuously improving and expanding the capabilities of the Claude AI model. The fine-tuning for Haiku, artifact publishing, and concurrency enhancements all contribute to making Claude a more versatile and powerful tool for users across various applications.

Elon Musk's AI company, xAI, has been making significant moves to build its own server farms and secure the necessary computing power to compete with industry leaders. According to the report, xAI has ended talks with Oracle on expanding their server rental agreement due to disagreements over timelines and power supply concerns. Instead, xAI is now building its own AI data center in Memphis, Tennessee.

The company is purchasing Nvidia chips from Dell and Supermicro for this project. xAI already rents about 16,000 Nvidia chips from Oracle and plans to build a supercomputer with 100,000 Nvidia GPUs for training its Grok 3.0 model. Elon Musk has commented on this, stating that the reason for building this massive system internally is to ensure that xAI's fundamental competitiveness depends on being faster than any other AI company.

This strategy of vertical integration and building their own computing infrastructure is consistent with Musk's approach at Tesla and SpaceX, where the companies have become largely self-sufficient. The goal is to have full control over the "steering wheel" rather than being a "backseat driver" when it comes to the development of their AI models.

The competition for AI dominance is heating up, and Elon Musk's move to build this powerful GPU cluster is a clear indication that xAI is determined to catch up and potentially surpass industry leaders like OpenAI. This race for the fastest and most capable AI models is likely to continue, with companies vying for the edge in computing power and model development.

Venture capital firm a16z has been actively purchasing large quantities of Nvidia's powerful H100 GPUs, with the goal of providing these resources to their portfolio companies. This strategy aims to give a16z an edge in AI-focused investment deals, as access to cutting-edge hardware can be a significant advantage in the rapidly evolving field of artificial intelligence.

The firm, which manages $42 billion in assets, has been renting out these GPUs to many of its portfolio companies, ensuring that the startups it invests in have the necessary computing power to develop and deploy advanced AI models. Additionally, a16z plans to expand this "oxygen" program, as they describe it, to more than 200,000 GPUs in the future.

This move by a16z highlights the growing importance of AI capabilities in the startup ecosystem. By securing access to high-performance hardware, the venture capital firm can offer its portfolio companies a valuable resource that may be difficult for others to obtain, potentially giving them a competitive edge in the race to develop and deploy innovative AI-powered solutions.

The scarcity of these powerful GPUs, combined with the increasing demand from AI-focused startups, has made it challenging for companies to acquire the necessary computing resources on their own. By leveraging its resources, a16z aims to position itself as a preferred partner for AI-driven startups, further strengthening its position in the rapidly evolving technology landscape.

OpenAI and Los Alamos National Laboratory, one of the leading U.S. national laboratories, are collaborating to study how artificial intelligence can be safely utilized by scientists in laboratory settings to advance bioscientific research.

The partnership involves developing evaluations to understand the capabilities and limitations of multimodal AI models, such as GPT-4, in assisting humans with tasks in a physical laboratory environment. This includes conducting biological safety assessments for GPT-4 and its unreleased real-time voice systems to determine how they could be employed to support bioscience research.

The goal of this collaboration is to provide a framework for the responsible and safe deployment of frontier AI models in laboratory settings, ensuring that the technology can be leveraged to enhance scientific discovery and innovation while addressing potential risks and ethical considerations.

Nintendo, the iconic video game company, has taken a firm stance against the use of generative AI in its video games. According to reports, Nintendo has become the largest company in the gaming industry, and perhaps the world, to explicitly say "no thank you" to the integration of generative AI.

The company's decision is in contrast to the approach taken by many other game developers, who are actively exploring ways to incorporate AI technology into their games. Nintendo, however, values the preservation of its well-established intellectual property and the unique gaming experiences it has crafted over the years.

In an investor call, the CEO of Nintendo expressed the company's stance, stating that while game development and AI technology have always been closely related, the use of generative AI can raise concerns about intellectual property rights. The CEO emphasized that Nintendo has the expertise to create optimal gaming experiences for its customers and that the company aims to continue delivering value that is unique and cannot be easily replicated by technology alone.

This decision by Nintendo reflects the company's commitment to maintaining its creative control and the integrity of its iconic characters and franchises. It also highlights the ongoing debate within the gaming industry about the role of AI in content creation and the potential impact on the industry's traditional development processes.

As the gaming landscape continues to evolve, it will be interesting to see how other major players in the industry respond to Nintendo's stance and whether the use of generative AI in video games becomes more widely adopted or faces further resistance from companies prioritizing their unique creative identities.

Google DeepMind has released a video showcasing how Gemini 1.5's long context window can help robots navigate the real world. The key points are:

• A limited context length makes it challenging for many AI models to recall environments as they navigate.
• With Gemini 1.5's million-token context window, the robots were able to recall key locations from a previous tour, such as "Lewis's desk" and "temporary desk area".
• The robots were then provided with additional multimodal inputs, such as map sketches, audio requests, and visual cues. With these inputs, the robots could carry out various actions for different people.
• The researchers have published a paper with more details on these experiments, which can be found in the description below.

Overall, this demonstrates how Gemini 1.5's long-term memory capabilities can enable robots to better navigate and interact with real-world environments, by recalling and utilizing contextual information from previous experiences.

The latest update to Tesla's Full Self-Driving (FSD) software, version 12.4.2, showcases an impressive capability - the ability to anticipate a pedestrian crossing the street. In the video shared, the Tesla vehicle does not wait for the pedestrian to start crossing, but rather slows down as it senses the pedestrian's movements, indicating a likely intention to cross the street. This proactive behavior, where the vehicle anticipates the pedestrian's actions rather than just reacting to them, demonstrates a significant advancement in Tesla's autonomous driving technology.

The video highlights how the Tesla's sensors and algorithms are able to detect subtle cues in the pedestrian's movements, even before they make a definitive move to cross the street. This level of situational awareness and predictive capability is crucial for ensuring the safety of both the vehicle occupants and pedestrians in real-world driving scenarios.

The smooth and seamless manner in which the Tesla slows down in response to the pedestrian's movements is a testament to the refinement of the FSD software. This update underscores Tesla's commitment to continuously improving its autonomous driving capabilities, making the roads safer for all users.