Microsoft, Apple Drop OpenAI Board Plans as Scrutiny Grows

Microsoft Corp. and Apple Inc. have decided to forgo board roles at OpenAI amid increasing regulatory scrutiny over Big Tech's influence on artificial intelligence. Microsoft, having invested $13 billion in OpenAI, will step down from its observer role, a move prompted by regulatory pressures from the US and Europe. These regulators have expressed concerns about Microsoft's dominance in AI, particularly through its integration of OpenAI's services into its platforms. Despite the withdrawal, this change is unlikely to alleviate concerns from the US Federal Trade Commission and other regulators regarding Microsoft's close ties with OpenAI.

The decision reflects broader scrutiny of Big Tech's AI partnerships. Both US and European regulators are examining Microsoft's relationships with AI startups, including a probe into its deal with Inflection AI. Similar investigations are also targeting Amazon's collaboration with AI company Anthropic and Nvidia's dominance over AI chips. This scrutiny stems from fears that Big Tech's investments in AI could lead to concentrated control over the most advanced AI technologies, limiting competition and innovation in the field.

Microsoft's departure from OpenAI's board follows last year's turmoil when OpenAI's directors fired and subsequently reinstated CEO Sam Altman. Moving forward, OpenAI plans to host regular stakeholder meetings with partners and investors to enhance collaboration on safety and security issues. Despite the board changes, OpenAI remains committed to its mission, continuing to foster partnerships while navigating the regulatory landscape.

OpenAI Partners With Los Alamos to Test AI’s Value for Lab Work

OpenAI and Los Alamos National Laboratory (LANL) have announced a groundbreaking partnership to explore the safe use of multimodal AI models in bioscientific research. This collaboration aims to evaluate how AI can assist scientists in laboratory settings, focusing on biological capabilities and safety. The initiative aligns with the recent White House Executive Order, which directs national labs to assess frontier AI models. By leveraging AI technologies like GPT-4o, OpenAI and LANL aim to enhance scientific research and mitigate associated risks, following a tradition of public-private sector cooperation in advancing health care and bioscience.

The partnership will involve an evaluation study to test GPT-4o's capabilities in performing and troubleshooting standard laboratory tasks. This includes using GPT-4o's vision and voice systems to support tasks such as genetic transformation, cell culture, and cell separation. The goal is to quantify how these models can improve task accuracy and completion rates for both novices and experienced professionals. This study will extend previous research by incorporating practical lab techniques and multimodal inputs, allowing for real-time, interactive problem-solving.

This collaboration underscores the potential of AI to revolutionize scientific research while emphasizing the importance of safety and risk assessment. OpenAI's Chief Technology Officer, Mira Murati, and LANL's deputy group leader, Nick Generous, highlight the dual benefits and risks of AI in science. The joint effort aims to set new standards for AI safety and efficacy, demonstrating the critical role of public and private sector collaboration in harnessing innovation for the greater good. As AI capabilities advance, this partnership seeks to pave the way for future innovations that benefit humanity while ensuring robust safety measures.

Exploring Multimodality with OpenAI: Romain Huet

Romain Huet, leading the developer experience at OpenAI, discussed the transformative journey of AI and OpenAI's mission to build AGI (Artificial General Intelligence) to benefit humanity. He highlighted the importance of iterative deployment and the role of developers in achieving this mission. Huet reminisced about the early days of OpenAI's developer platform, which started with GPT-3 in 2020, and how it evolved with the launch of ChatGPT, capturing the world's attention. The introduction of GPT-4 brought significant advancements, including enhanced reasoning, creativity, and the ability to analyze and interpret data across multiple modalities such as text, images, and audio, leading to the creation of GPT-4 Turbo and GPT-4 O, a flagship model that integrates multimodal capabilities in real-time.

Huet demonstrated the capabilities of these advanced models through live demos, showcasing how GPT-4 can engage in natural conversations, understand and generate emotional tones, and interpret visual data. He illustrated practical applications, such as using GPT-4 to improve a travel app's responsiveness and employing the model to read and summarize text from a book. Furthermore, he presented the future potential of AI through Sora, a diffusion model that generates videos from simple prompts, and the voice engine model, which creates custom voices based on short clips, emphasizing the seamless integration of various modalities to create more natural and efficient human-computer interactions.

Looking ahead, Huet outlined OpenAI's focus on four key areas: enhancing textual intelligence, making models faster and more affordable, enabling model customization for specific needs, and advancing the development of AGI agents that can interact with the world through multiple modalities. He emphasized the company's commitment to supporting developers and startups, encouraging them to build AI-native products that leverage these cutting-edge technologies. Huet concluded by inspiring the audience to embrace the future of AI, highlighting the exciting opportunities for bold builders to shape the next generation of software and contribute to the ongoing AI revolution.

AI speech generator 'reaches human parity' — but it's too dangerous to release, scientists say

Microsoft's new AI speech generator, VALL-E 2, can replicate human voices using just a few seconds of audio, achieving a level of quality comparable to human speech. Developed by Microsoft researchers and detailed in a paper on arXiv, VALL-E 2 uses advanced techniques like "Repetition Aware Sampling" and "Grouped Code Modeling" to generate fluid and natural speech. It has reached human parity, meaning its output matches or exceeds the quality of human speech in benchmark tests.

Despite its impressive capabilities, VALL-E 2 will not be released to the public due to concerns about potential misuse, such as spoofing voice identification or impersonating specific speakers. Microsoft has stated that VALL-E 2 is purely a research project with no plans for public release or product integration. This decision aligns with increasing worries about the ethical implications of advanced voice cloning and deepfake technologies.

The researchers see potential future applications for AI speech tech in various fields, including education, entertainment, journalism, and accessibility features. However, they emphasize the need for protocols to ensure ethical use, such as obtaining speaker approval and developing synthesized speech detection models to prevent misuse.

Other stuff

• I Banned Most Overused Words in ChatGPT Responses 🔥🔥

• Defeated by A.I., a Legend in the Board Game Go Warns: Get Ready for What’s Next 🔥

• Has Artificial Intelligence Co-opted the Sparkle Emoji?

•  Odyssey - Hollywood-grade visual AI

• AWS App Studio promises to generate enterprise apps from a written prompt 🔥

• Destroying Russian Tanks Is Just The Start For U.S. AI Drone Autopilot

• Internet of Agents: Weaving a Web of Heterogeneous Agents for Collaborative Intelligence 🔥

• Summary of NATO's revised Artificial Intelligence (AI) strategy

Hope you enjoyed today's newsletter

⚡️ Join over 200,000 people using the Superpower ChatGPT extension on Chrome and Firefox.

OR