OpenAI files ‘digital voice assistant’ trademark application

Sam Altman may have Siri and Alexa in his sights after OpenAI filed a 'digital voice assistant' trademark application

OpenAI appears to be venturing into the realm of digital voice assistants, challenging established entities like Apple's Siri and Amazon's Alexa. It has filed a trademark application encompassing "digital voice assistants" and a "voice engine," suggesting a potential new product launch. Despite the speculative nature of this move, as companies often file for trademarks that never materialize into products, OpenAI has signaled forthcoming releases, including a significant upgrade to its ChatGPT model. Sam Altman, OpenAI's CEO, has hinted at various important announcements preceding any details about a GPT-5 model. Additionally, OpenAI has filed for trademarks for future models such as GPT-6 and GPT-7, covering functionalities like conversation simulation and predictive analytics, with these applications still under review. The company faced a setback when its attempt to trademark "GPT" was rejected for being too descriptive. Meanwhile, the "voice engineer" trademark application is pending and targets a wide array of capabilities, from voice recognition to multilingual translation and text-to-voice conversion, aligning with OpenAI's existing TTS API and Whisper speech recognition model. OpenAI has not yet offered any comments on these developments.

• Stability AI Announcement - Emad Mostaque has stepped down as CEO of Stability AI and from the company's Board to focus on decentralized AI. The Board has appointed COO Shan Shan Wong and CTO Christian Laforte as interim co-CEOs while they search for a permanent replacement. Chairman Jim O’Shaughnessy expressed confidence in their ability to guide the company, noting Mostaque's contributions to the firm's success, including hundreds of millions of downloads and leading AI models. This transition is seen as a pivotal moment for Stability AI to continue advancing in generative AI technology and maintain its industry leadership.

• Elvis Act Signed Into Tennessee Law to Protect Musicians From AI Deepfakes - Tennessee Governor Bill Lee signed the ELVIS Act into law, designed to safeguard musicians against unauthorized AI-generated deepfakes and voice cloning. The law extends existing personal rights protection to cover artists' voices, combating AI’s potential misuse. Amidst concerns over the rise of artificial intelligence infringing on copyright and intellectual property, this legislation criminalizes the use of AI to replicate a musician's voice without consent, punishable as a Class A misdemeanor. The act, celebrated by artists like Luke Bryan and supported by state legislative leaders, aims to uphold and defend Tennessee's vibrant musical heritage as the technology landscape evolves.

• Apple picks Baidu’s Ernie Bot for its iPhone 16 in China, report says - Apple is set to partner with Baidu to use the Chinese tech giant's AI technology for its iPhone 16 and other products in China, diverging from using its own AI model, which will be implemented elsewhere. This decision emerged during CEO Tim Cook's China visit, where he opened a new retail store, met suppliers, and discussed Apple's carbon-neutral goals. The move is seen as a compliance strategy and may influence Apple's competitiveness in the Chinese market, where its sales have declined by 24% in early 2024. Baidu's share prices rose following these developments. The partnership echoes Samsung's integration of Baidu's AI in its Galaxy S24 series for China, marking a significant trend in foreign tech companies adopting local AI solutions within the Chinese market.

• Apple’s iOS future could also include Anthropic - Apple is speculated to be exploring partnerships with major AI players like OpenAI, Google, and potentially Baidu in China to integrate chatbot functionality into iOS. Additionally, Mark Gurman of Bloomberg has hinted that Anthropic might also be a contender for collaboration. Furthermore, Gurman posits that Apple could enable developers to integrate generative AI deeply within the iPhone. In a subscriber-exclusive segment, Gurman reveals that the upcoming iOS 18 is expected to feature an overhauled and more customizable home screen.

• The tech industry can’t agree on what open-source AI means. That’s a problem - There's an ongoing debate about the definition of "open-source AI," prompting the need for clarity as the term impacts innovation and technology control. The Open Source Initiative (OSI), which sets the criteria for open-source software, is tackling this challenge by gathering a diverse group of 70 stakeholders, including big tech representatives, to establish a definition for open-source AI. The initiative highlights the complexity of aligning interests within the community, from activists to multinational corporations, to achieve a consensus that ensures fair play without stifling the concept's usage by leading tech firms such as Meta, which has expressed support for OSI's efforts.

• In One Key A.I. Metric, China Pulls Ahead of the U.S.: Talent - Research from MacroPolo indicates that China has surpassed the U.S. in producing top A.I. researchers, accounting for nearly half of the world’s elite in this field, while the U.S. produces approximately 18%. This marks a significant increase from three years prior when China's share was about one-third. The data was gathered by examining the backgrounds of researchers who contributed to the 2022 Conference on Neural Information Processing Systems (NeurIPS). The trend of Chinese researchers staying in their home country, rather than remaining in the U.S. post-Ph.D., is reversing a previous pattern. This shift in A.I. talent distribution is considered geopolitically significant as A.I. is key to economic productivity and innovation. Despite this talent shift, generative A.I. developments have been primarily driven by U.S. companies like Google and startups such as OpenAI, attracting substantial investment and potentially the interest of Chinese researchers amidst U.S.-China tensions.

• Johnson & Johnson MedTech Works With NVIDIA to Broaden AI’s Reach in Surgery - NVIDIA collaborates with Johnson & Johnson MedTech to integrate AI into surgical procedures, aiming to increase efficiency and enhance clinical decision-making in the operating room. Harnessing NVIDIA's advanced AI platforms – IGX for edge computing and Holoscan for medical device creation – J&J MedTech seeks to expedite the application of AI in surgeries. This partnership leverages J&J MedTech's considerable presence in global operating rooms and experience in healthcare professional education. The initiative, unveiled at NVIDIA's global AI conference (GTC), also proposes an open ecosystem to foster innovation, allowing third-party model deployment and facilitating real-time clinical insights to improve surgical outcomes. Their AI-driven tools could advance surgical analytics through continuous learning and collaboration, potentially leading to smarter surgical technologies that assist surgeons and optimize operational workflows.

Awesome Research Papers

Quite-STaR: Language Models Can Teach Themselves to Think Before Speaking - The "Quiet-STaR" paper introduces a novel approach to improve language models' reasoning abilities by having them generate internal rationales to explain future text. It extends the Self-Taught Reasoner (STaR) framework to enable language models to infer unstated rationales in arbitrary text, addressing challenges through techniques like a tokenwise parallel sampling algorithm. The paper demonstrates performance improvements on reasoning benchmarks and natural text perplexity without requiring task-specific fine-tuning, marking a significant step towards more general and scalable reasoning in language models. The paper represents a significant contribution to the field of natural language processing and artificial intelligence, as it suggests a method for LMs to self-improve their reasoning abilities without direct supervision or task-specific training.

Awesome New Models

Stable Code Instruct 3B - Stable Code Instruct 3B is a cutting-edge language model designed for enhancing code completion and facilitating natural language interactions in software development. It outperforms similar models in diverse coding tasks. Notably proficient in popular programming languages like Python, Javascript, and Java, the model also excels in less familiar ones such as Lua, thanks to its deep understanding of coding principles. Additionally, it adeptly handles advanced coding tasks, such as database queries and code translation. Its instruction tuning allows it to process complex technical instructions, showcasing strong capabilities in logical reasoning and technical narrative comprehension. This model's superior performance is documented in a technical report and is now accessible through Stability AI Membership, with resources available on Hugging Face.