➀ A senior staffer from OpenAI's AGI readiness team has left the company; ➁ The former employee, Miles Brundage, warns about the lack of readiness for AGI at OpenAI and globally; ➂ Brundage's departure led to the dissolution of the AGI readiness team.
Recent #OpenAI news in the semiconductor industry
➀ OpenAI receives one of the first engineering builds of NVIDIA's new Blackwell DGX B200 AI system; ➁ The system features 8 x B200 AI GPUs with up to 1.4TB of HBM3 memory and 72 petaFLOPS of training performance; ➂ OpenAI expresses excitement about the new system and its potential to accelerate AI compute.
➀ NVIDIA is considering joining a funding round for OpenAI, valuing the startup at over $100 billion. ➁ Apple and Microsoft are also reportedly in talks to participate in the funding round. ➂ Thrive Capital is leading the funding round, investing around $1 billion.
➀ Intel passed on investing in OpenAI in 2017, missing the opportunity to be part of generative AI's early stages. ➁ The potential investment was reportedly worth $1 billion for a 15% stake. ➂ Intel's CEO at the time, Bob Swan, did not see generative AI as a significant market opportunity.
1. OpenAI announces new AI search feature called SearchGPT; 2. SearchGPT allows users to search for information using natural language and provides quick answers with relevant sources; 3. The feature is currently in prototype phase and requires users to sign up for a waiting list to try it out.
➀ Lee Kefu has been engaged in a demanding schedule, including public events, speaking engagements, and meetings over the past month;
➁ His busy lifestyle is linked to his role as the founder of Zero One Universe, which is facing challenges and requiring his full attention;
➂ He expresses concerns about OpenAI's potential dominance in the AI industry and the implications of pre-training models on AI development;
➃ Lee emphasizes the importance of realistic strategies for AI companies and the need to adapt to the market environment.
➀ OpenAI CEO Sam Altman confirmed that ChatGPT-5 won't be released this year; ➁ The company is delaying GPT-5 to early next year, focusing on updates to existing models; ➂ GPT-5's release date is uncertain due to the complexity of outgoing models and computing power limitations.
➀ OpenAI is developing its first inference chip in collaboration with Broadcom and TSMC to support its AI systems;
➁ In addition to NVIDIA chips, OpenAI plans to use AMD chips to meet its growing infrastructure needs;
➂ OpenAI has researched various plans to diversify chip supply and reduce costs, including internal design and manufacturing, but has temporarily abandoned this ambitious plan due to high costs and time requirements;
➃ OpenAI has formed a team of about 20 people led by Richard Ho and Thomas Norrie to develop the chip, aiming to launch the first AI chip in 2026;
➄ OpenAI is exploring other parallel solutions and may introduce more partners in the future;
➅ The development of self-developed chips is part of a broader trend among major tech and AI companies to develop dedicated hardware to meet the unique needs of AI workloads.
➀ OpenAI, Broadcom, and TSMC are collaborating on an AI inference chip; ➁ This move deviates from OpenAI's previous strategy of raising funds to build a network of fabs for AI IC production; ➂ Intel has also shifted its focus to inference ICs, possibly due to Nvidia's strong position in the AI training IC market.
➀ Orion is expected to be released before the end of 2024; ➁ It may be 100 times more capable than GPT-4; ➂ The use of synthetic data in training raises concerns about model output.
➀ OpenAI has finally released a ChatGPT app for Windows, following its earlier release on macOS; ➁ The app is available for ChatGPT Plus, Enterprise, Team, and Edu users, but not for free users; ➂ The app supports Windows 11 and Windows 10 (version 17763.0 or newer), and offers various models including GPT-o1 and GPT-4o with Canvas; ➃ The app allows quick access via Alt+Space and supports file interaction and DALL-E image generation; ➄ However, the app lacks Standard or Advanced Voice mode, which may impact convenience and efficiency.
➀ In the 14th century, Dick Whittington set out for London believing its streets were paved with gold, much like Palo Alto today; ➁ Sam Altman, worth $2 billion, made his first real money in 2012; ➂ OpenAI is seeking $65 billion for expansion, with current spending at $7 billion and expected revenue of $3.7 billion
➀ OpenAI CEO Sam Altman proposed a bold plan to raise $7 trillion to build 36 semiconductor factories and data centers globally; ➁ TSMC executives considered the plan absurd, calling Altman 'podcasting bro'; ➂ Japanese officials laughed at the plan during a meeting with Altman.
➀ OpenAI CEO Sam Altman proposes a $7 trillion plan to build 36 semiconductor factories and data centers; ➁ TSMC executives call the plan 'absurd' and nickname Altman 'podcasting bro'; ➂ OpenAI is exploring a $1 trillion scale investment; ➃ Altman seeks to establish data centers in the US to please the government; ➄ OpenAI is in talks to negotiate $65 billion in financing to support its operations.
➀ OpenAI is facing a challenging period with high-level leadership changes and a potential setback in its massive financing round; ➁ Apple's decision not to participate in the latest $6.5 billion round could affect the company's structure and strategy; ➂ Concerns are raised about OpenAI's focus on product development over research, and the impact on its culture and mission.
➀ TSMC executives reportedly dismissed OpenAI CEO Sam Altman as a 'podcasting bro'(cn: 嘴炮达人 or 满嘴跑火车); ➁ Altman's proposed investment plans in AI were considered 'absurd' by TSMC; ➂ The scale of the investment would be in trillions of dollars, raising concerns among TSMC execs.
➀ The cost of training AI models has surged in the past year, as indicated by data from Epoch AI; ➁ The training cost for Gemini, a large language model from Google, ranges from $30 to $191 million; ➂ OpenAI's ChatGPT-4, released in 2023, had a technical creation cost between $41 million and $78 million.
➀ OpenAI is developing its first in-house chip with TSMC using the A16 Angstrom process, aimed at enhancing Sora's video-generation capabilities. ➁ The chip development was initially planned with TSMC but was later canceled due to profitability concerns. ➂ The new chip could potentially boost Apple device sales as it integrates with Apple's generative AI features.
➀ TSMC's A16 process, which has not yet entered mass production, has already attracted significant attention. ➁ Apple has reserved the initial capacity of TSMC's A16 process, while OpenAI, the developer of ChatGPT, has also joined in reserving the A16 capacity due to its long-term demand for self-developed AI chips. ➂ OpenAI's strategic partnership with Broadcom and Marvell for ASIC chip development is expected to be produced at TSMC's 3nm family and subsequent A16 processes.
➀ Google unveiled its new lineup of Pixel 9 phones, including the $1,799 Pixel 9 Pro, at its Made by Google event. ➁ The company also announced Gemini Live, a direct competitor to OpenAI's advanced AI models. ➂ Google is expanding its AI-powered search summaries to six more countries: India, Brazil, Japan, the U.K., Indonesia, and Mexico.