Jessie Chen

Super proud of Mike and his team. Many people asked me over the last year why long context and reasoning are so important. My answer is simple: long context is, in other words, a much better memory that allows LLM to utilize its ability to comprehend incoming text fully which as we all know is a million miles better than RAG. A very clear use case is with processing huge documentation like corporate wiki. Now, the deal with reasoning over long context is also simple: if models can get an ability to process long context but do that poorly then what's the value is in doing? And that's why a proper benchmark like BABILong is so desperately needed. It helps you to open your eyes on how inefficent the majority of the modern methods for processing long context by LLMs are. It helps you to understand what the real goal is. Let's make LLMs understand long context better together!

18

2 則回應

Earlier this year, Jensen Huang - Founder / CEO of Nvidia - opined the following when asked where he thought the next amazing revolution was going to come from: - (Text transcribed verbatim) ***************** "This is going to be flat out one of the biggest ones ever, and I told you I was going to tell you, and I'm going to tell you. There's no question that digital biology is going to be it. For the very first time in our history, in human history, biology has the opportunity to be engineering, not science. When something becomes engineering, not science, it becomes less sporadic and exponentially improving. It can compound on the benefits of the previous years, and every company's contributions, every research's contributions, compound on each other. For the very first time, we know how to represent biology, understand the language of biology. We can represent the language of chemistry. It's never happened before. And now we're going to have incredible tools that bring the world of biology, which is very chaotic and constantly changing, and very diverse and complex, and bring it into the world of computer science. And that is going to be profound. And so if you happen to love this intersection, I think it's going to be rich with opportunities. It's going to be a giant industry." ***************** Now if you take this thesis and apply it to the world of #ClinicalDevelopment / #ClinDev in #LifeSciences, the opportunities to leverage #AI cuts across: * optimizing trial design * patient recruitment and retention * data analysis and prediction And the impact on the speed, cost, and efficacy of new therapeutics will be profound. And so I am really looking forward to joining some of the best minds in the intersection of AI and Life Sciences including Adam Petrich, MD | Arnaub Chatterjee | Prof Ava Amini from MSFT | Charbel Haber from Moderna | Fred Martin | Ittai Harel | Thaminda Ramanayake | Jasmine LaCoursiere | Gaurav Singal, MD former Chief Data Officer from Foundation Medicine | Orr Inbar | Sean Hilscher | Thorsten Wirkes from Bertlessman at the world's first AI Summit focused on clinical development in Boston on May 30th. Together we hope to demystify the role of AI in Clinical Development, discuss the risks and regulatory issues surrounding this field, and real-world applications and success stories, and the venture capital investment happening in this area. Apply to join us on May 30th - https://qhsummit24.com/

75

9 則回應

🤔How proficient are Claude 3.5 Sonnet and GPT-4o in Korean? Today, I'll discuss the issues and underlying causes in Claude 3.5 Sonnet's Korean output based on specific inputs. (I'll cover GPT-4o in the next post). For this test, I provided the following input to Claude 3.5 Sonnet: Translate the sentence 'Youngcheol is watching a movie with Jesse' into Korean for each context: 'Friends Talk, Polite Talk, Business Talk, Presentations, Formal Writing.' And these are the detailed information: -Speaker: female -Youngcheol: male, older, not a friend -Jesse: female, older, a friend ==Output Results Friends Talk: 영철 씨가 제시 언니랑 영화 보고 있어. Polite Talk: 영철 씨께서 제시 언니와 영화 보고 계세요. Business Talk: 영철 씨께서 제시 씨와 영화를 보고 계십니다. Presentation: 영철 씨는 제시 씨와 영화를 보고 있습니다. Formal writing: 영철 씨는 제시 씨와 영화를 관람 중이다. == There are several issues with the results, but let's focus on the use of '씨' for now. In these results, '씨' was used for '영철 Youngcheol.' For '제시 Jesse,' although '언니' was used appropriately in Friends and Polite Talk, it incorrectly changed to '씨' in the other three contexts. *Changing contexts doesn't alter the speaker's relationship with the characters, so '언니' should be maintained across all contexts. This is one of the most challenging parts for AI in Korean, which I will discuss in the next post. Regarding '영철 Youngcheol,' it seems Claude missed the condition that '영철 Youngcheol' is older than the speaker, which I pointed out, leading it to re-output the results. == Friends Talk: 영철 오빠가 제시 언니랑 영화 보고 있어. Polite Talk: 영철 오빠께서 제시 언니와 영화 보고 계세요. Business Talk: 영철 님께서 제시 씨와 영화를 보고 계십니다. Presentation: 영철 님은 제시 씨와 영화를 보고 있습니다. Formal writing: 영철 씨는 제시 씨와 영화를 관람 중이다. == So, is it correct now? Actually, '영철 오빠' is incorrect, and '영철 씨' and '제시 씨' are also incorrect. The most appropriate honorific in all contexts should be '영철 님.' You might wonder, 'Can't we use it?' But there's a difference between 'can be used' and 'is actually used.' Without sufficient context, we should prefer commonly used expressions. Why is '영철 씨' incorrect? How would you feel if someone younger called you by your name with '씨'? Can you easily call someone older by adding '씨' to their name? That's right, '씨' cannot be used for someone older. Interestingly, Korean native speakers use '씨' correctly without mistakes in daily life, but they do not necessarily understand all the restrictions of '씨.' Traditional Korean language education for foreigners often uses '씨' with every name, regardless of age or closeness. This results in humorous situations where classmates of the same or differing ages call each other '씨.' Perhaps influenced by this, AI knows the restrictions of '씨' well but continues to misuse it in examples. Through our proprietary Korean research data, we are building an AI model that outputs accurate Korean at a native level. You can check the Korean proficiency of our AI model through the link below. Thank you. https://lnkd.in/gCvuuxAZ

1

How do we make AI agents think and act? 🤖 Today's episode with Shunyu Yao (and special co-host Harrison Chase from LangChain!) is probably our best agents episode so far: - Origins of ReAct and how it inspired LangChain - Agent-Computer Interfaces and tool design - How to benchmark AI agents - The future of AI products Now live on Substack and YouTube 📫

32

6 則回應

PTN raw deals data Sept 10, 2024: Chai Discovery / $30m / N/A / AI for drug discovery / OpenAI, Thrive Capital PIN AI / $10m / N/A / Open-source Personal Intelligence Network (PIN) / A16z CSX, Hack VC Mangomint / $35m / USA / Salon and spa software / Altos Ventures, SaaStr Fund, OpenView Venture Partners Mangomint (Venture Debt) / $12m / USA / Salon and spa software / Avidbank Superluminal Medicines / $120m / USA / Generative biology, chemistry, and ML for medicine / RA Capital Management, Insight Partners, NVentures, Gaingels CardiaTec / $6m / UK / AI for cardiovascular disease drug discovery / Montage Ventures, Continuum Health Ventures, Laidlaw Ventures, Apex Ventures, angel investors double jump.tokyo / $10m / Japan / Web3 solutions, blockchain games / SBI Investment, Gate Ventures, Sony Group Corporation, Taisu Ventures, Kitadenshi Inc. Odigos / $13m / USA / eBPF-based observability platform / Venture Guides, Salesforce Ventures, Mango Capital, Firestreak Ventures, angel investors CSA Medical / $53m / USA / Medical device for chronic bronchitis / TVM Capital Life Science, Yonjin Venture, SV Health Investors, Ascension Ventures, First Analysis, Blue Heron, Intersouth Partners Cercli / $4m / UAE / HR and Payroll platform / Afore Capital, COTU Ventures, Y Combinator, Rebel Fund, individual investors Calyxia / $35m / France / Cleantech, biodegradable microcapsules / Lombard Odier Investment Managers, Bpifrance, Astanor OrsoBio, Inc. / $67m / USA / Treatments for obesity and related disorders / Ascenta Capital, Woodline Partners, Samsara Biocapital, Longitude Capital, Enavate Sciences, NuevaBio, Eli Lilly and Company NetNow / $1m / Canada / Trade credit automation platform / Ripple Ventures, Centre Street Partners, Antler, Motivate Venture Capital, Day One Ventures Candid Therapeutics / $370m / USA / Autoimmune disease drugs / Venrock Healthcare Capital Partners, Fairmount, TCGX, venBio Partners Bambusa Therapeutics / $15m / USA / Bispecific antibodies for I&I disorders / BVF Partners L.P., Dawn Biopharma, Salvia GmbH, INCE Capital Korint / €5m / France / Insurtech / Ventech, 360 Capital Citroniq Chemicals / $12m / USA / Carbon-negative materials / Lummus Technology Ventures, other investors Espislogen / £12m / UK / Immunoglobulin E (IgE) antibodies for cancer / British Patient Capital, Novartis Venture Fund, Epidarex Capital, 3B Future Health Fund, ALSA Ventures MirrorWeb / $63m / USA / Digital communications archiving software / Mainsail Partners

4

Ethernet always wins. It has already adopted most of the features of Infiniband. Now, Meta reports Ethernet is more reliable.

8

At Fireworks, we’re driven by a bold vision: to deliver simple API access to the full spectrum of knowledge. In my recent discussion with Sonya Huang and Pat Grady on Sequoia Capital's Training Data podcast (https://lnkd.in/eV3YnWTu ), we explored the transformative potential of Compound AI Systems—where tens, hundreds, or even thousands of models and components work together to achieve extraordinary results. As AI rapidly shifts from training to inference, startups are facing significant challenges with latency and cost. Fireworks is eliminating these obstacles by optimizing AI inference for speed and scalability, empowering businesses to deploy real-time, high-quality AI with ease. If you're an AI startup founder ready to accelerate your product development, there’s no better time to get started with Fireworks and bring your AI innovations to market faster than ever before.

284

5 則回應

Jensen Huang, the Founder and CEO of NVIDIA, delivered an exciting keynote discussing his latest innovations and upcoming AI trends. We're thrilled that he mentioned GliaCloud Co., Ltd. and InfinitiesSoft Solutions Inc., two startups that SparkLabs Taiwan invested in early on. We believe this is the best time to develop AI, and Taiwan is playing a crucial role. SparkLabs Taiwan is all-in on AI. Let's work together and make a difference. #DreamBigGoGlobal #DBGG _ David Chen Peng Agnes Dewey Hsu Wenyu Chen Mallory (Meng-Yao) Chien Kenny Chen Joe (游詩喬) Yu

271

1 則回應

🚀 Agora Integrates with OpenAI to Launch Real-Time Conversational AI SDK 🚀 Agora Agora, Inc. has launched its Conversational AI SDK, integrated with OpenAI’s Realtime API, which is set to revolutionize real-time, voice-driven AI experiences across multiple industries. This integration is a game-changer for developers, allowing them to effortlessly bring natural, voice-driven AI into a wide array of applications. Whether it’s 24/7 customer support, concierge services, health and wellness, education, language learning, or even gaming, the new SDK provides endless possibilities for more interactive and engaging AI-powered user experiences. Key features of the Agora Conversational AI SDK include: Seamless real-time voice interaction with OpenAI’s GPT models Ultra-low latency response times, thanks to Agora’s global real-time network powering 60 billion minutes of interaction per month Built-in AI echo cancellation and background noise suppression to ensure clear, accurate voice processing in any environment This is a huge leap forward in bringing lifelike AI experiences into everyday life. Companies like Grepp and Wyze are already leveraging this integration to enhance personalized learning and IoT devices. As Agora's CEO, Tony Zhao, put it: "The integration with OpenAI will unleash a wave of creativity and utility in the AI space unlike anything we have seen to date." 🌐 Developers can now tap into the full potential of real-time conversational AI, opening doors to new innovations that were once thought impossible. Learn more about how you can integrate the Agora Conversational AI SDK into your applications and enhance user engagement through natural, voice-driven interactions. 🔗 Check out more here: https://lnkd.in/grxmQgJc #ConversationalAI #RealTimeAI #VoiceDrivenAI #AIInnovation #AgoraSDK #OpenAI #ArtificialIntelligence #AIInHealthcare #AIInEducation #AIInGaming #CustomerSupportAI #IoTDevices #TechInnovation #VoiceAI

2

Yesterday, I attended the "Future R&D Innovation Strategy Joint Forum" held at the National Assembly of Korea. This forum was a crucial event to convey the opinions of various sectors regarding the future innovation of science and technology R&D. A significant discussion point was the recent reduction in the government-supported budget for science and technology R&D in Korea. Many experts voiced concerns about the negative impact this reduction could have on technological progress. R&D budgets are essential for fostering innovation and technological advancement, and such budget cuts could severely affect the nation's long-term competitiveness. Additionally, there was a strong call from the patent attorney community for the establishment of a specialized court for technology disputes. With the increasing complexity of technology disputes, a specialized court is seen as necessary to effectively handle these issues. This would greatly contribute to promoting technological innovation and providing companies with a secure environment to invest in R&D confidently. This forum highlighted the critical importance of continuous support and policy improvements to ensure South Korea's advancement as a leading nation in science and technology. We hope for ongoing support and systemic enhancements to drive Korea forward as a technological powerhouse.

4

𝐂𝐡𝐢𝐧𝐚 𝐨𝐧 𝐭𝐡𝐞 𝐑𝐨𝐚𝐝 𝐭𝐨 𝐂𝐨𝐦𝐦𝐞𝐫𝐜𝐢𝐚𝐥 𝐒𝐮𝐜𝐜𝐞𝐬𝐬 𝐢𝐧 𝐑𝐨𝐛𝐨𝐭𝐚𝐱𝐢𝐬 🚖 It looks like China will be the first to achieve commercial success in robotaxis, and it's not DiDi leading the way, but Baidu, Inc.. Baidu representatives have announced that their Apollo Go division will become profitable next year. In the last quarter of 2023 alone, Apollo Go completed over 839,000 trips, with approximately 45% of orders in Wuhan being fully autonomous. The company has revealed that the 6th generation Apollo robotaxi will cost around 200,000 yuan (28,169 USD) – less than half the cost of the previous generation. 𝐊𝐞𝐲 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: - 𝘐𝘯𝘤𝘳𝘦𝘢𝘴𝘦𝘥 𝘛𝘳𝘪𝘱𝘴: Over 839,000 trips in Q4 2023. - 𝘏𝘪𝘨𝘩 𝘈𝘶𝘵𝘰𝘯𝘰𝘮𝘺: 45% of trips in Wuhan were fully autonomous. - 𝘊𝘰𝘴𝘵 𝘌𝘧𝘧𝘪𝘤𝘪𝘦𝘯𝘤𝘺: New generation robotaxi costs less than half of its predecessor. - 𝘗𝘳𝘰𝘧𝘪𝘵𝘢𝘣𝘪𝘭𝘪𝘵𝘺 𝘔𝘪𝘭𝘦𝘴𝘵𝘰𝘯𝘦: Expected to reach breakeven by Q4 2024 and become profitable by 2025. Baidu's success is not just due to technological innovations but also legislative support. Autonomous vehicles need real-world driving experiences to learn effectively, and China’s regulatory environment has been conducive to this development. https://lnkd.in/dNSGrem7 #Innovation #AutonomousVehicles #Robotaxi #Baidu #ApolloGo #Technology #Transport #FutureOfMobility 🚗✨

4

🚀 Webinar Announcement: Registration Now Open! 🚀 Join us on 24th July, Wednesday, 9:00 AM - 10:00 AM Singapore Time, for an exclusive webinar with thought leaders in robotics and global expansion. Speakers: 🔹 Pham Quang Cuong - Co-founder & CEO at Eureka Robotics, Dr. Phạm Quang Cuong is a pioneer in robotics and automation, renowned for his groundbreaking work at Eureka Robotics. With a strong academic background and innovative approach, he is shaping the future of intelligent robotics. 🔹 Joyce Zhang Gray - CEO & Founder at Alariss Global, Joyce is an alumna of Stanford Graduate School of Business, Princeton, and Harvard. She was the first employee at fintech Human Interest and has worked with Microsoft Asia-Pacific, Groupon China, World Bank's IFC, and the Federal Reserve Bank of New York. She is also a term member of the Council on Foreign Relations and a former Elected Director of the Harvard Alumni Association Board of Directors. 🎟️ Free Registration - Limited Seats! https://lnkd.in/epdXaYaJ Secure your spot today and be part of an insightful discussion on the journey from academia to industry and global expansion strategies. We look forward to your participation and valuable insights. Don't miss out on this opportunity! 💬 Drop your thoughts and questions in the comments below. Let's get the conversation started! #Webinar #Robotics #GlobalExpansion #Innovation #TechLeaders #IndustryInsights #AlarissGlobal #EurekaRobotics #FreeWebinar #LimitedSeats

5

🎓 Highlights from Jensen Huang's Caltech Commencement Speech: 💪✨ Importance of Perseverance & Dedication: Huang emphasized perseverance in one's career, sharing lessons from leading NVIDIA. 🚀💡 Revolutionary Impact of Accelerated Computing: Discussed NVIDIA's contributions to accelerated computing, enhancing performance, reducing costs, and saving energy. 🤖📈 Breakthrough in Deep Learning: AlexNet's 2012 success marked the AI revolution, transforming NVIDIA from a graphics company to an AI powerhouse. 🌐🔍 Potential of Generative AI: Generative AI's vast potential, with NVIDIA's innovations creating new computing models and industries. 🏭🔗 AI Factory Concept: Introduced the concept of AI factories, envisioning future industrial revolutions centered on producing intelligent tokens. 🌱🔄 Resilience & Adaptability: Shared stories of personal and corporate resilience, essential for overcoming setbacks. 🗓️✅ Prioritizing Important Work: Stressed tackling the most important tasks first to start each day successfully. 🌸🛠️ Craft & Dedication: Illustrated the importance of craft and dedication with a story about a Japanese gardener, encouraging graduates to find and commit to their craft. 💻🔧 Transformation of the Computing Industry: Explained how computing is fundamentally transforming, impacting all industries. 🌟🚀 Future Hope: Urged graduates to find a lifelong passion, prioritize important matters, seize opportunities, and develop their "superpowers." #Caltech2024 #JensenHuang #AIRevolution #CareerAdvice https://lnkd.in/gbHNvZh3

1

South Korean #AI startups are on the brink of joining the global unicorn club. Despite challenges, the funding momentum for companies like Rebellions Inc. and Upstage Co. is promising. The rapid growth in venture capital reflects a strong belief in their potential. The government and institutional investors must continue supporting these deep-tech innovations to secure Korea's position in the AI industry. What do you think? Do you see this momentum continuing? #KoreanStartups #AIInnovation #VentureCapital

66

1 則回應

✨ It was an honor to participate in Huawei's Amplify Intelligence, Advance Industry forum in #Shanghai, where #thoughtleaders, #academics, and #regulators came together to explore the transformative potential of #AI across various sectors. As the Co-Founder of AllStarsWomen DAO and an Adjunct Professor at The Hong Kong University of Science and Technology, I had the opportunity to contribute ideas on the global regulatory landscape surrounding AI and how it impacts different industries, from business opportunities, operations automations to ethics, security, and the responsible deployment of these powerful technologies. Luna Xie, Director of Huawei’s Digital Media Affairs Department, welcomed us warmly to Huawei Connect 2024. She encouraged a deep exchange of insights on the evolving role of AI in shaping industries & society. Next came some amazing keynote speakers and a heated panel with #AI Academia and tech leaders from #Singapore, #Bahrain and #Thailand. 🇸🇬 Hong-Eng Koh (高宏荣), Global Chief of Public Services Industry Scientist at Huawei, showcased Huawei’s One Cloud with AI strategy, will drive inclusivity & personalization in education & healthcare, helping underserved populations access high-quality AI and Cloud services. James Sun, Vice President of Huawei’s Global Public Sector, discussed Huawei’s Smart Campus solutions, demonstrating how AI cloud computing is improving education systems globally for effective learning. Rose Pan, Head of the Info Sector Greater China at BSI, addressed the critical intersection of cybersecurity & AI innovation / adoptions. As for the panel, it was moderated by the tech savvy Javier Ferrández Soto 韩小平, from Huawei Social Media Team, with insights from: Dr. Jassim Haji, President of the Artificial Intelligence Group in Bahrain, highlighted AI’s potential in the oil & gas industry by improving predictive maintenance & spill detection. Ben Leong, Associate Professor at the National University of Singapore, provided a balanced perspective on AI’s impact on the job market, explaining that while routine, well-defined tasks will likely be automated, people-centric and creative roles will continue to thrive. Narinrit Prem-apiwathanokul, Vice President of the Thailand Information Security Association, emphasized the critical need for collaboration between industries & governments as both sectors navigate the emerging challenges & opportunities presented by AI. As for me, I challenged those who think Generative AI will not change the way we work or do business. Most operational and administrative jobs will be replaced. New jobs will be created. Regulations on Generative AI calls for global collaborations among all regulators so it will be affective for all matters related to #IP, #security, #moral, #ethics, and human safety. 💫 Indeed it was a heated debate but it was exciting to see how the ideas shared here will drive forward the next wave of innovation and growth! 💥 #HuaweiConnect2024

90

18 則回應

Key Points from Jensen Huang's Quote on AI: 1. Current State: $1 trillion in data centers use outdated general-purpose computing. 2. Future Shift: Transition to accelerated computing with GPUs, replacing up to $2 trillion in infrastructure. 3. AI Industrial Revolution: AI produces valuable floating-point numbers (tokens) at high volumes, similar to a new industrial revolution. 4. AI Factories: New data centers will produce these AI tokens, transforming electricity into AI capabilities. 5. Economic Impact: AI tokens will drive $100 trillion in new products, services, and industry productivity. 6. Need for Financial Platforms: Platforms like Stripe will be crucial to monetize and transact in this new AI-driven economy.

13