technology innovation

Former Sequoia Partner Michelle Fradin, Involved in FTX Investment, Joins OpenAI

Michelle Fradin, a former partner at Sequoia Capital known for her key role in the investment in FTX, has transitioned to a new position at OpenAI, where she will spearhead data strategy, acquisitions, and operations. This move marks a significant step in her professional journey, intertwining her expertise in venture capital with the rapidly evolving landscape of artificial intelligence.

At Sequoia Capital, Fradin was integral in shaping investment strategies, particularly in the cryptocurrency sphere, most notably with FTX. Her tenure at Sequoia witnessed a dynamic period, especially in the wake of the FTX collapse, leading to significant shifts within the firm. Beyond her investment acumen, Fradin played a pivotal role in Sequoia’s exploration of AI and its integration into various industries. This experience provided a foundational understanding of the interplay between technology, business, and investment, fueling her transition to a more AI-focused role.

Fradin’s interest in technology and its commercial applications was evident early in her career. Starting at McKinsey, she gained insights into leadership and organizational structures before moving to Google, where she led the Creative Lab team, delving into e-commerce, payments, and AI/ML products. This phase was instrumental in honing her storytelling skills and scouting early-stage investments for Google. Her pursuit of understanding what constitutes a great business led her to Hellman & Friedman, a private equity firm, further solidifying her investment prowess. It was her move to Sequoia that synergized her passion for investing, serving others, and continual learning.

At Sequoia, Fradin was involved in groundbreaking discussions on the role of large language models (LLMs) like ChatGPT in innovation, observing their growing integration into products across various companies. She contributed to Sequoia’s engagement with 33 companies, spanning seed stage startups to large enterprises, to understand their AI strategies and the evolving landscape of AI applications. Her work highlighted the adoption of language model APIs, the importance of retrieval mechanisms for enhancing the quality of AI outputs, and the increasing interest in customizing language models for specific contexts.

Michelle Fradin’s move to OpenAI is a testament to her deep understanding of both the venture capital world and the transformative potential of AI. Her journey from Sequoia Capital to OpenAI reflects a broader trend in the technology sector, where AI is increasingly becoming central to business strategies and operations. As she embarks on this new chapter, her experience and insights are poised to make a significant impact in shaping OpenAI’s data strategies and future innovations.

Rabbit Inc Introduces AI-Powered r1 Device for Mobile Interaction

Rabbit Inc, a Los Angeles-based AI startup, recently unveiled its cutting-edge mobile device, the Rabbit r1, at CES 2024, marking a significant stride in the realm of mobile technology. The r1, priced at $199 and available for pre-order, is poised to redefine our interactions with digital devices through its innovative features and user-centric design.

At the heart of the Rabbit r1 is the Large Action Model (LAM), a groundbreaking operating system that transcends the traditional app-based interfaces. LAM, distinct from the generative AI models like ChatGPT, is adept at understanding and acting upon human intentions, facilitating tasks like booking tickets or ordering groceries without requiring specific app integrations. This marks a departure from the conventional approach of downloading and navigating through multiple apps, offering a more streamlined and intuitive user experience.

The r1 device itself is a testament to Rabbit Inc’s commitment to innovation and user convenience. Unlike typical smart devices that rely on smartphones for operation, the r1 functions as a fully standalone device with both Wi-Fi and cellular connectivity. It features a 2.3 GHz MediaTek Helio P35 processor, 4 GB of memory, 128 GB of storage, and a USB-C port. Moreover, it is equipped with an empty, factory-unlocked SIM card slot, ensuring versatility in connectivity. The battery life of the r1, designed for all-day usage, aligns with the device’s promise of convenience and efficiency.

Rabbit’s dedication to privacy and security is evident in its technology. The company ensures that no user credentials for third-party services are stored, with all authentication processes occurring within the respective service’s login systems. This approach grants users full control over their data and interactions with rabbit OS. Additionally, rabbit’s data infrastructure adheres to major industry standards, ensuring robust security and encryption.

The r1’s design and development are backed by a team of experts, including Kaggle Grandmasters and former Google engineers. Jesse Lyu, the Founder and CEO of Rabbit Inc and a two-time Y Combinator alumnus, brings his experience from founding Raven Tech, a startup that pioneered conversational AI operating systems and was later acquired by Baidu. This team’s expertise underpins the r1’s advanced capabilities and potential for revolutionizing human-computer interactions.

Rabbit Inc’s vision for the r1 extends beyond just a novel gadget; it aims to fundamentally alter how we engage with technology. By providing an intuitive and efficient operating system that anticipates and acts on user intentions, rabbit plans to enhance our daily digital interactions significantly. With the additional $10 million raised in Series A funding, the company is poised to further refine its hardware and operating system, potentially setting a new standard in AI-powered mobile technology.

ChatQA: A Leap in Conversational QA Performance

The recently published paper, “ChatQA: Building GPT-4 Level Conversational QA Models,” presents a comprehensive exploration into the development of a new family of conversational question-answering (QA) models known as ChatQA. Authored by Zihan Liu, Wei Ping, Rajarshi Roy, Peng Xu, Mohammad Shoeybi, and Bryan Catanzaro from NVIDIA, the paper delves into the intricacies of building a model that matches the performance of GPT-4 in conversational QA tasks, a significant challenge in the research community.

Key Innovations and Findings

Two-Stage Instruction Tuning Method: The cornerstone of ChatQA’s success lies in its unique two-stage instruction tuning approach. This method substantially enhances the zero-shot conversational QA capabilities of large language models (LLMs), outperforming regular instruction tuning and RLHF-based recipes. The process involves integrating user-provided or retrieved context into the model’s responses, showcasing a notable advancement in conversational understanding and contextual integration.

Enhanced Retrieval for RAG in Conversational QA: ChatQA addresses the retrieval challenges in conversational QA by fine-tuning state-of-the-art single-turn query retrievers on human-annotated multi-turn QA datasets. This method yields results comparable to the state-of-the-art LLM-based query rewriting models, like GPT-3.5-turbo, but with significantly reduced deployment costs. This finding is crucial for practical applications, as it suggests a more cost-effective approach to developing conversational QA systems without compromising on performance.

Broad Spectrum of Models: The ChatQA family consists of various models, including Llama2-7B, Llama2-13B, Llama2-70B, and an in-house 8B pretrained GPT model. These models have been tested across ten conversational QA datasets, demonstrating that ChatQA-70B not only outperforms GPT-3.5-turbo but also equals the performance of GPT-4. This diversity in model sizes and capabilities underscores the scalability and adaptability of the ChatQA models across different conversational scenarios.

Handling ‘Unanswerable’ Scenarios: A notable achievement of ChatQA is its proficiency in handling ‘unanswerable’ questions, where the desired answer is not present in the provided or retrieved context. By incorporating a small number of ‘unanswerable’ samples during the instruction tuning process, ChatQA significantly reduces the occurrence of hallucinations and errors, ensuring more reliable and accurate responses in complex conversational scenarios.

Implications and Future Prospects:

The development of ChatQA marks a significant milestone in conversational AI. Its ability to perform at par with GPT-4, coupled with a more efficient and cost-effective approach to model training and deployment, positions it as a formidable tool in the domain of conversational QA. The success of ChatQA paves the way for future research and development in conversational AI, potentially leading to more nuanced and contextually aware conversational agents. Furthermore, the application of these models in real-world scenarios, such as customer service, academic research, and interactive platforms, can significantly enhance the efficiency and effectiveness of information retrieval and user interaction.

In conclusion, the research presented in the ChatQA paper reflects a substantial advancement in the field of conversational QA, offering a blueprint for future innovations in the realm of AI-driven conversational systems.