Blockchain Bond Platform Folds, Tech Giants Unite for Metaverse OpenUSD, UK ￡40M in IoT and 5G Innovation, Meta's Persona-Driven Chatbots and more!
Hi Everyone,
Welcome to QX Snapshots - a weekly recap of the key news on emerging technologies. In this newsletter, you will get a "digest" of latest info on AI, Quantum Technology, Industrial Metaverse and Enterprise Blockchain.
Hope it brings you value :)
[Quantum Technology] BMW Group, Airbus and Quantinuum collaborate to fast-track sustainable mobility research using cutting-edge Quantum Computers. Airbus, BMW Group, and quantum technology firm Quantinuum have partnered to accelerate research into the use of quantum computers to simulate quantum systems, particularly focusing on the chemical reactions in fuel cells. The collaboration successfully modelled the oxygen reduction reaction (ORR) on a platinum-based catalyst using Quantinuum's H-Series Quantum Computer. ORR, which converts hydrogen and oxygen into water and electricity, currently limits the efficiency of fuel cells, requiring a large amount of platinum catalyst. Therefore, better understanding of this reaction could help improve efficiency. This project is seen as a significant step toward a sustainable energy transition, benefiting products like metal-air batteries with enhanced efficiency. BMW Group sees quantum computing's potential in researching new materials for more efficient processes, while Airbus sees its benefits in the quest for sustainable and hydrogen-powered alternatives, like its planned ZEROe aircraft. The team hopes to identify alternative materials to improve performance and reduce the production costs of fuel cells.
[AI] AWS rolls out generative AI service for healthcare documentation software. Amazon Web Services (AWS) has launched a new AI-enabled service called AWS HealthScribe, designed to help healthcare professionals with transcription and analysis of conversations with patients, thereby aiding clinical note creation. The service uses generative AI and speech recognition to transform patient-clinician conversations into preliminary clinical notes, extracting key medical terms and creating summaries to be integrated into electronic health record (EHR) systems. Powered by Amazon Bedrock, AWS HealthScribe aims to make it easier for healthcare software providers to integrate generative AI capabilities into their applications without the need to manage machine learning infrastructure or train their own healthcare-specific language models. Furthermore, AWS HealthScribe promotes responsible AI deployment by citing the source of each generated text line from the original transcript. AWS asserts that this service prioritizes data security and privacy, providing users with control over data storage and encryption. The company has also partnered with 3M Health Information Systems, Babylon Health, and ScribeEMR to deploy AWS HealthScribe. The service is expected to streamline clinical documentation processes and enhance EHR workflows.
[Blockchain] Blockchain bond platform LedgerEdge to close. Founded by R3 CEO. Blockchain bond solution LedgerEdge, co-founded by David Rutter, is shutting down due to struggles with regular adoption and funding issues. LedgerEdge was established in 2020 to allow buy and sell-side firms to share pre-trade data on a peer-to-peer basis using smart contracts, aiming to give users more control over their data. At launch, it had 45 firms in the UK Multilateral Trading Facility (MTF) and 70 globally using or onboarding it, but the company struggled to get firms to use it regularly. Additional time for traction would require more capital, which has become harder to acquire. The decision to use distributed ledger technology (DLT) allowed sell-side participants to own their data, but the nodes were centrally hosted, necessitating some level of trust. The LedgerEdge shutdown follows closures of other enterprise blockchain projects, including we.trade, Marco Polo, Tradelens, B3i, and the ASX’s CHESS settlement system.
[Metaverse] Pixar, Adobe, Apple, Autodesk, and NVIDIA form Alliance for OpenUSD. Pixar, Adobe, Apple, Autodesk, and NVIDIA, in collaboration with the Joint Development Foundation (JDF), have formed the Alliance for OpenUSD (AOUSD) to foster the evolution and adoption of Pixar’s Universal Scene Description (OpenUSD) technology. OpenUSD, created by Pixar, provides a robust interoperability across 3D tools, data, and workflows. The alliance aims to promote this technology to achieve standardization in the 3D ecosystem and to stimulate the development of a broader range of 3D-enabled products and services. The AOUSD will prepare written specifications detailing OpenUSD's features to encourage its wider adoption and integration. The JDF, an affiliate of the Linux Foundation, was selected to host the project, providing a pathway to recognition through the International Organization for Standardization (ISO). The alliance invites various companies and organizations to join and participate in shaping OpenUSD's future. OpenUSD is considered essential for platforms like Apple’s visionOS and the new Reality Composer Pro developer tool, among others.
[General technology] UK launches ￡40M fund to drive local IoT and 5G innovation. The UK government has launched a ￡40 million fund to improve digital connectivity and accelerate the development of advanced wireless technologies, including IoT and 5G. The fund aims to support innovation in sectors such as manufacturing, transport, agriculture, and public services. The primary goal is to establish 5G Innovation Regions that have strong potential to drive 5G adoption and development. Benefits of 5G-powered IoT technologies include new economic opportunities, enhanced public services, and improved living conditions. Potential applications include 5G-enabled drones for data-driven farming and 5G sensors for traffic management. Research suggests that widespread 5G adoption could yield productivity benefits of up to ￡159 billion by 2035. The fund also complements the UK Telecoms Innovation Network's nationwide campaign to promote collaborations among businesses, telecoms providers, and vendors interested in 5G services. The application deadline for the fund is September 3rd, with successful bidders expected to be announced later this year.
领英推荐
FEATURED: ‘ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs’
By: Yujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, Yankai Lin, Xin Cong, Xiangru Tang, Bill Qian, Sihan Zhao, Runchu Tian, Ruobing Xie, Jie Zhou, Mark Gerstein, Dahai Li, Zhiyuan Liu, Maosong Sun
“Despite the advancements of open-source large language models (LLMs) and their variants, e.g., LLaMA and Vicuna, they remain significantly limited in performing higher-level tasks, such as following human instructions to use external tools (APIs). This is because current instruction tuning largely focuses on basic language tasks instead of the tool-use domain. This is in contrast to state-of-the-art (SOTA) LLMs, e.g., ChatGPT, which have demonstrated excellent tool-use capabilities but are unfortunately closed source. To facilitate tool-use capabilities within open-source LLMs, we introduce ToolLLM, a general tool-use framework of data construction, model training and evaluation. We first present ToolBench, an instruction-tuning dataset for tool use, which is created automatically using ChatGPT. Specifically, we collect 16,464 real-world RESTful APIs spanning 49 categories from RapidAPI Hub, then prompt ChatGPT to generate diverse human instructions involving these APIs, covering both single-tool and multi-tool scenarios. Finally, we use ChatGPT to search for a valid solution path (chain of API calls) for each instruction. To make the searching process more efficient, we develop a novel depth-first search-based decision tree (DFSDT), enabling LLMs to evaluate multiple reasoning traces and expand the search space. We show that DFSDT significantly enhances the planning and reasoning capabilities of LLMs. For efficient tool-use assessment, we develop an automatic evaluator: ToolEval. We fine-tune LLaMA on ToolBench and obtain ToolLLaMA. Our ToolEval reveals that ToolLLaMA demonstrates a remarkable ability to execute complex instructions and generalize to unseen APIs, and exhibits comparable performance to ChatGPT. To make the pipeline more practical, we devise a neural API retriever to recommend appropriate APIs for each instruction, negating the need for manual API selection.
This work introduces how to elicit the tool-use capabilities within LLMs. We present an instruction tuning dataset, ToolBench, which covers 16k+ real-world APIs and various practical use-case scenarios including both single-tool and multi-tool tasks. The construction of ToolBench purely uses ChatGPT and requires minimal human supervision. Moreover, we propose DFSDT to reinforce the planning and reasoning ability of LLMs, enabling them to navigate through reasoning paths strategically. For efficient evaluation of tool learning, we devise an automatic evaluator ToolEval. By fine-tuning LLaMA on ToolBench, the obtained model ToolLLaMA matches the performance of ChatGPT and exhibits remarkable generalization ability to unseen APIs. Besides, we develop a neural API retriever to recommend relevant APIs for each instruction. The retriever can be integrated with ToolLLaMA as a more automated tool-use pipeline. In general, this work paves the way for future research in the intersection of instruction tuning and tool use for LLMs.”
Read the full paper: here.
If you enjoyed today’s QX Snapshots, I would love it if you subscribed for more. Can’t wait a full week? You can keep up with me here on LinkedIn for daily emerging technology content.