Oracle AI World 2025 | 主题演讲:Oracle愿景与战略
文章语言:
简
繁
EN
Share
Minutes
原文
会议摘要
AI's transformative potential in healthcare and agriculture is explored, emphasizing Oracle's contributions to AI infrastructure, secure data handling, and innovative projects like multimodal model training, biometric security, and autonomous drones. The focus is on leveraging AI to enhance efficiency, solve complex problems, and maintain data privacy across various sectors.
会议速览

The dialogue highlights Oracle's involvement in AI model training and its anticipation of AI's transformative impact on solving complex global challenges.

Discusses Oracle's pivotal role in enabling AI models to utilize private, high-value data for enhanced reasoning, moving beyond publicly available datasets. Highlights the importance of AI in augmenting human capabilities across various professions, and the significant investments being made in AI model development.

Discusses AI's significance over internet bubble myths, highlighting genuine tech advancements versus mere e-commerce mislabeling as AI, emphasizing AI's unparalleled value in technology.

The dialogue explores AI's sensory capabilities, such as smelling chemicals and diagnosing illnesses like cancer, akin to dogs' abilities. It highlights AI's potential in surgical precision, surpassing human doctors through microscopic vision and enhanced coordination, exemplified by robots learning from internet videos to perform tasks efficiently.

The dialogue explores the superior surgical capabilities of AI systems, emphasizing their enhanced hand-eye coordination and precision over human surgeons. It delves into the structural parallels between AI models and the human brain, particularly in visual processing and reasoning. The discussion highlights the energy efficiency of the human brain compared to AI's high computational demands, showcasing AI's rapid reasoning and data processing abilities.

Oracle is constructing the world's largest AI cluster for OpenAI in Texas, featuring over 450,000 Nvidia GPUs, powered by grid and on-site turbines. This sprawling project, spanning 1,000 acres, involves complex engineering feats including power generation and transmission, and is crucial for training multimodal AI models. More than 3,500 personnel work daily to meet unprecedented demand for AI infrastructure.

Discusses DeepMind's Nobel Prize-winning protein folding model and Elon's multimodal AI Grok and real-time Tesla models, highlighting their distinct applications and importance in fields ranging from molecular biology to autonomous driving, emphasizing the need for low latency in critical decision-making scenarios.

Oracle's AI Data Platform enables models to reason on private data securely, enhancing capabilities in coding, diagnostics, and drug design without compromising data privacy.

The dialogue highlights the capability of integrating real-time and private data into AI models using Oracle's AI database and Rag technology, enabling models to access current information such as today's stock prices.

Oracle's AI database leverages vectorization to make data from Oracle databases and external cloud storages accessible to AI models, enabling advanced reasoning capabilities beyond traditional data boundaries.

The dialogue explores the strategic use of AI in analyzing customer data to predict future product purchases and optimize sales efforts. By integrating customer relationship management with AI models, the initiative aims to identify potential buyers, tailor marketing messages, and guide sales teams effectively. This approach not only enhances customer engagement but also streamlines sales forecasting and strategy, showcasing the transformative power of AI in business operations.

The dialogue discusses the creation of AI agents for tasks like emailing prospective buyers, highlighting the concept of vibe coding and debating the use of English versus precise declarative languages for programming, with Oracle's Apex offering a custom solution.

Oracle leverages AI to generate stateless, secure, and scalable applications, rebuilding Cerner's codebase and expanding into hospital-specific systems. The goal is to automate the entire healthcare ecosystem, including patients, providers, payers, regulators, and financiers, for a truly modern and efficient system.

The dialogue discusses creating an AI agent to connect healthcare providers with payers, ensuring the best possible care is both effective and reimbursable. It highlights the challenge of prescribing treatments like Ozempic in the UK where NHS doesn't cover it, emphasizing the need for affordable, reimbursable alternatives. The AI model accesses medical literature and EHRs to assist doctors in making informed decisions, considering clinical trials and drug efficacy for personalized patient care.

The AI model is trained to understand and apply insurance policies, ensuring prescribed treatments are fully reimbursable. It navigates complex insurance landscapes, such as Medicare and Medicaid, to propose the best care aligned with the highest reimbursement levels, crucial in government-funded healthcare systems. The AI also adapts to policy changes, like recent updates on Ozempic eligibility for patients with specific BMI thresholds, enhancing care and reimbursement accuracy.

Discusses leveraging AI to automate healthcare financial processes, enabling banks to loan against anticipated reimbursements, and Oracle's unique role in combining AI technology with scaled enterprise applications to optimize healthcare operations.

Discusses advancements in AI code generation, focusing on scalable, secure, and reliable applications for modernizing industries and ecosystems, including healthcare and utilities. Highlights the automation of complex inter-company processes to enhance efficiency and reduce costs, emphasizing the role of AI agents in streamlining operations across different sectors.

AI is poised to dramatically improve life quality by enhancing security, healthcare, and living standards. Innovations include biometric identity theft prevention, IoT medical devices for home monitoring, and connected ambulances for seamless patient care. These advancements aim to reduce hospital stays, combat fraud, and create a safer, more efficient world.

Recalling a motorcycle accident, the speaker highlights how an MRI only focused on broken ribs, ignoring other data. They predict AI will uncover overlooked details in scans, offering greater precision and accuracy.

A discussion on the transformative impact of biometric authentication and AI in enhancing security, reducing fraud, and improving patient monitoring and medical diagnostics, advocating for a future where technology ensures privacy and efficiency.

A groundbreaking discussion highlights AI's capability to accurately measure fetal development and diagnose cancers from tumor biopsy slides within minutes, offering immediate health outcomes. The dialogue also introduces a metagenomic testing device, promising swift responses and improved patient outcomes, eliminating weeks of uncertainty and delaying treatment.

Current pathogen identification relies on PCR testing, which can miss unknown viruses requiring genomic testing after culturing. A new sensor promises to sequence all genetic material in a sample, including circulating tumor DNA, revolutionizing disease detection and diagnosis.

AI distinguishes false positives from serious cancers, enabling early treatment. Metagenomic sequencers identify pathogens, antibiotic resistance, and could have provided early warnings for pandemics like COVID-19, emphasizing the need for widespread low-cost devices in hospitals.

The dialogue discusses the necessity of mass-producing medical devices, specifically metagenomic sequencers, in robot factories to ensure reliability and affordability, emphasizing the importance of cost-effective production methods for global healthcare accessibility.

The dialogue highlights the necessity of efficient food production, focusing on robotic greenhouses that reduce water usage by 90%, save space, and minimize transportation CO2 emissions. These innovative structures, suitable for urban proximity, enhance food freshness, nutrition, and taste, while also serving as potential Martian habitats. Their air pressure design and ETFE roofing material enable easy transport and deployment, even to Mars, showcasing a sustainable future for agriculture.

The dialogue discusses how AI can be used to enhance food production by genetically modifying crops like wheat to produce more grain per acre and absorb more CO2, which can then be converted into calcium carbonate, effectively removing it from the atmosphere. Additionally, the potential to engineer plants to extract nitrogen directly from the atmosphere is explored as a sustainable alternative to harmful nitrogen fertilizers, promising a significant reduction in environmental pollution while increasing agricultural efficiency.

An air traffic control system for drones has been developed to deliver blood samples from clinics to testing laboratories. An RFID specimen VA with a BYD tag ensures personal privacy by anonymizing the samples during transit, while test results are securely stored in the cloud for retrieval by doctors and patients. This system also prevents errors in sample handling, enhancing privacy and efficiency in healthcare.

Drones equipped with infrared cameras can instantly detect forest fires and identify arsonists, improving response times and safety. They also offer a safer alternative to high-speed police chases by tracking vehicles from a distance.

Discusses deploying drones during the dry season in Palisades for rescuing lost hikers, highlighting their portability and the establishment of charging stations for safe landing and operation.
要点回答
Q:What are the two big phases of AI technology according to the speech?
A:The two big phases of AI technology mentioned in the speech are the dawning of the AI era, where companies build enormous AI models, and the use of these models to solve very important problems, such as early diagnosis of cancer and precise and accurate surgery.
Q:What is the role of Oracle in the context of AI?
A:Oracle is a major participant in building data centers for AI training. However, the much larger opportunity for Oracle is to enable the use of remarkable electronic brains, or AI models, to solve humanity's most difficult and enduring problems by making private data available to these models.
Q:How can private data enhance the value of AI models?
A:Private data can enhance the value of AI models by allowing them to reason not just on public data but also on private data. This is important because high value data is often in an Oracle database, and making this data available to AI models can lead to more accurate and personalized outcomes.
Q:What are the potential applications of AI beyond just solving problems?
A:Beyond solving problems, AI can assist humans in various endeavors, including scientific research, engineering, teaching, cooking, and even performing surgeries. It will make people better at these professions by providing assistance and enhancing their capabilities.
Q:Why is the investment in AI considered significant?
A:The investment in AI is considered significant because the smartest people are investing fortunes in building and training AI models, indicating the importance and extraordinary capabilities of these technologies.
Q:What is the connection between the internet and AI?
A:The internet is foundational to computing and is critical for AI, as it enables the vast amounts of data necessary for training AI models. The internet allows for the creation of new technologies and facilitates the spread of information, which is essential for advancements in AI.
Q:How do AI models perceive and interact with the world?
A:AI models have various functionalities similar to human perception; they can hear, see, and even smell (as in the case of the 'Dog's Nose' project). They can read, recognize patterns, understand language, and reason logically. AI can perceive its environment and interact with users through language and mathematical understanding.
Q:What advantages do AI robots have over human surgeons?
A:AI robots have superior hand-eye coordination and precision, allowing them to perform surgeries with perfect results. They can see at a microscopic level, which is beyond human capability, enabling them to operate more accurately and with less damage to surrounding tissues compared to the best human surgeons.
Q:How does the structure of AI models compare to the human brain?
A:AI models are structured similarly to the human brain, with multiple neural networks performing different functions like vision, language, and reasoning. Convolutional neural networks simulate the visual cortex and there are transformer networks for language and holistic understanding. AI models are specialized just like the human brain's specialized centers.
Q:What is the significance of the '20 W meat computer' analogy?
A:The '20 W meat computer' analogy compares the human brain's energy consumption and computational capabilities to that of a 20-watt computer. This highlights the remarkable efficiency and power of the human brain, which operates with 20 watts and supports complex functions like sensation, recognition, reasoning, and creativity.
Q:How many people work on site at Oracle's AI cluster daily?
A:More than 3500 people work on site each day to deliver capacity at an unprecedented rate.
Q:What is the significance of the AI models trained by OpenAI?
A:The AI models trained by OpenAI, like Grok and Tesla, are significant because they are multimodal, trained on diverse data types, and are designed for real-time applications requiring immediate decision-making, such as self-driving cars. These models incorporate complex reasoning capabilities, perform multi-step reasoning, and can answer questions and generate code.
Q:What differentiates real-time AI models from other types of AI models?
A:Real-time AI models, such as those used in Tesla vehicles, are designed to operate within milliseconds, making immediate decisions without latency. They are tailored for applications where rapid response is critical, as opposed to models that can afford to process information over a network or take more time to generate responses.
Q:How does Oracle's AI database facilitate the use of private data with AI models?
A:Oracle's AI database facilitates the use of private data with AI models by allowing users to add their private data to the models' library of information and knowledge while keeping it secure. This enables the models to reason across both public and private data without exposing the private data to external parties. The database uses a Rag capability to incorporate new, unseen data and make it accessible to the AI models.
Q:What is the 'Rag' capability in Oracle's AI database?
A:The 'Rag' capability in Oracle's AI database stands for 'Retrieve and Generate'. It allows the database to incorporate new data that the AI model has not been trained on, such as today's stock prices or recent news, and make that information accessible to the model for use in reasoning and decision-making.
Q:How does the new Oracle AI database handle data from various sources?
A:The new Oracle AI database can handle data from various sources by vectorizing the data from Oracle databases, different clouds, and object stores like Oci Object Store or Amazon S3. It creates vector indexes to data in other cloud storages, making the data accessible for reasoning by the multimodal AI models. This means that data from different sources can be made easily accessible to AI models without being restricted to a single location.
Q:How did Oracle use its customer data to predict customer behavior?
A:Oracle used its customer data by vectorizing it and making it accessible to a multimodal AI model via the 'Rag' capability. This allowed the model to analyze the customer data and predict which customers are likely to buy another Oracle product in the next six months. An AI agent was then created to identify the best customer references for potential buyers, encouraging them to purchase the product. This application demonstrated the model's ability to quickly solve complex business problems and inform sales strategies.
Q:What is Vibe coding and how does it facilitate generating computer programs?
A:Vibe coding is a term used to describe a process where one can generate computer programs directly from English without needing to think too hard or get too precise, allowing for a more intuitive approach to programming.
Q:Why might some engineers debate using English as a programming language?
A:Engineers might debate using English as a programming language because English is notorious for its imprecision, and there could be discussions on whether it's better to create a custom, highly precise, declarative language for programming.
Q:How does Oracle ensure the applications it generates are stateless and reliable?
A:Oracle ensures the applications are stateless and reliable by designing them so that they don't store state information and can pick up where they left off in a different data center seamlessly, ensuring no data is lost and no single point of failure.
Q:What is the difference between low code application programming languages and Oracle's AI-generated applications?
A:Low code application programming languages are designed for creating applications that may only serve a few users and can slow down when scaling to millions of users. In contrast, Oracle's AI-generated applications are designed from the start for scalability and to handle large user loads efficiently.
Q:How does Oracle's AI contribute to modernizing the healthcare industry?
A:Oracle's AI contributes to modernizing healthcare by rebuilding entire systems such as Cerner, and creating AI-generated applications that automate not just hospitals and clinics but also the entire healthcare ecosystem, including accounting, HR, and banking systems for hospitals.
Q:Why is it necessary to automate the entire healthcare ecosystem, not just the hospitals and clinics?
A:Automating only hospitals and clinics is insufficient because the entire ecosystem, including payers, regulators, pharma companies, banks, and government entities that interact with hospitals and clinics, must also be automated to achieve a truly modern, efficient healthcare system.
Q:How does the AI agent developed by Oracle assist in connecting providers to payers?
A:The AI agent developed by Oracle assists in connecting providers to payers by determining the best possible care that is also fully reimbursable, taking into account the medical literature, test results, EHR data, and insurance policies to optimize care delivery.
Q:How does Oracle Cloud support AI development and application in different industries?
A:Oracle Cloud supports AI development and application by providing AI models, AI code generators, and a platform to develop and deploy scalable AI applications tailored to specific industries, such as healthcare and utilities.
Q:What role do AI agents play in business interactions?
A:AI agents automate processes not only within a company but also between companies, facilitating communication and transactions.
Q:How does the speaker describe the future impact of AI on living standards?
A:The speaker believes that with AI, people will live much better, healthier, and longer lives, eat better food, and live in better houses, leading to a much better world.
Q:What is the proposed solution to prevent identity theft?
A:To prevent identity theft, biometric methods such as face, voice, and fingerprint recognition are being worked on, eliminating the need for passwords.
Q:How can monitoring systems in hospitals and ambulances benefit patients and medical staff?
A:IoT medical devices can monitor patients at home or in a hospital, even during transport in an ambulance, providing real-time data to medical staff.
Q:What is the advantage of using AI in medical imaging?
A:AI can accurately measure and identify medical conditions from scans, providing precise information that can lead to better diagnoses.
Q:How will AI change password security?
A:AI will eliminate the need for passwords by using biometric authentication, which is more secure and convenient.
Q:How can AI reduce credit card fraud and improve privacy?
A:AI can provide a more secure and private method of payment by using biometric credit cards that are difficult to imitate, reducing fraud.
Q:What is the envisioned future for patient data privacy and medical devices?
A:Patients will have control over their health data, with medical devices connected to the internet and secure databases, allowing doctors to monitor and receive critical information without compromising patient privacy.
Q:How can AI assist in early cancer diagnosis?
A:AI can analyze circulating tumor DNA to detect early cancer stages accurately, distinguishing between false positives and serious conditions, leading to early treatment and better outcomes.
Q:What is the goal with regards to the production of medical devices?
A:The goal is to mass-produce affordable, high-quality medical devices in robot factories, making them widely available to improve healthcare globally.
Q:What are the advantages of growing food in greenhouses and moving plants around?
A:Growing food in greenhouses and moving plants around allows for greater efficiency in water usage, space utilization, and habitat preservation. It enables plants to grow in confined areas before they are moved to a harvesting area, significantly reducing water consumption and Co2 output.
Q:How can indoor agriculture near urban centers reduce transportation costs and Co2 emissions?
A:By placing indoor farms, such as greenhouses, near urban centers, the distance and Co2 emissions involved in transporting the food to population centers are greatly reduced, resulting in fresher and potentially more cost-effective produce.
Q:What are the benefits of fresher produce from greenhouses?
A:Produce grown in greenhouses can be harvested daily and is delivered to grocery stores the same afternoon, ensuring that the food is much fresher, which also increases its nutritional value, taste, and overall quality.
Q:What is the significance of robotic greenhouses and their potential use on Mars?
A:Robotic greenhouses, or Martian habitats, are large, positive air pressure buildings with an advanced infrastructure that includes a rail system for moving plants. They are designed to be efficient in Co2 usage and inhospitable to humans but ideal for plant growth. They could be constructed on Mars and used as such Martian habitats.
Q:How can the modification of wheat increase grain production and remove Co2 from the atmosphere?
A:Wheat can be genetically modified to produce 20% more grain per acre, which consumes more CO2 through photosynthesis. This increased grain production can then be used to extract Co2 and convert it into calcium carbonate, a process similar to how coral reefs build themselves, effectively removing Co2 from the atmosphere.
Q:What are the potential benefits of removing Co2 from the atmosphere using agricultural crops?
A:Removing Co2 from the atmosphere using crops could help manage the Co2 level in the atmosphere to a desired range, such as from 440 to 400 parts per million. This could be done by having crops like wheat, corn, and soybeans convert Co2 into calcium carbonate, a process that is natural and requires no additional cost.
Q:How can AI and engineered plants help reduce environmental damage from agriculture?
A:AI can help engineer plants to take nitrogen directly from the atmosphere instead of relying on nitrogen fertilizers, which can lead to environmental damage like nitrogen runoff into rivers and oceans. This innovation can reduce the cost and environmental impact of agriculture and is being researched for crops like corn.
Q:What applications have drones been put to beyond military purposes?
A:Drones have been utilized for delivering blood samples from clinics to testing laboratories, helping to maintain patient privacy, and for immediate detection of forest fires, including identifying arsonists. They also assist in police operations by tracking down suspects without the risks associated with high-speed chases.

Oracle Corp.
Follow