Sponsored by

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each Thursday, we aggregate the major developments in artificial intelligence; we pass along the news, useful resources, tools or services, and exciting projects in open source. Even if you aren’t an engineer, we’ll keep you in touch with what’s going on in AI.

• Microsoft is rebranding pretty much everything AI in their suite under ‘Copilot’. Microsoft is integrating Copilot functionalities across Bing, Edge, and Windows in an attempt to unify the AI experience. That means Bing Chat will soon just be called Microsoft Copilot, starting December 1st.

• Microsoft unveiled their own new advanced chips: Azure Maia, an AI Accelerator for cloud-based AI workloads, and Azure Cobalt, optimized for general-purpose workloads. Both custom silicon chips are designed to power Microsoft’s Azure data centers and ready the company and its enterprise customers for a future full of AI.

• New AI tools are being released: Copilot Studio (A low-code tool for customizing Copilot integration), Copilot for Service (transforming customer service with AI), Copilot in Microsoft Dynamics 365 Guides (combining AI with mixed reality for frontline workers), Microsoft Fabric (a unified AI-powered platform for data management), Azure AI Studio (exploring, building, testing, and deploying AI apps), Vector Search in Azure AI Search.

• ‘Model-as-a-Service’ will enable developers to integrate open source AI models as API endpoints to their applications, in addition to the OpenAI API.

Some news from OpenAI this week: CEO Sam Altman confirmed in an interview that GPT-5 is in development, although a timeline was not offered on its potential release. The company is also hoping to raise even more capital from Microsoft to support its efforts to develop superintelligence - on top of the $10B Microsoft has already invested this year.

Altman also announced via Twitter that the OpenAI would be temporarily halting ChatGPT Plus sign-ups amid high demand following DevDay. It’s unclear how long users might have to wait before being able to subscribe again. The chatbot also suffered a DDoS attack last week that led to users being unable to access the platform for about 90 minutes. Russian hacktivist group Anonymous Sudan claimed responsibility for the attack.

CarePods are currently being deployed in malls, gyms, and offices in the San Francisco Bay Area, New York, Chicago, and Philadelphia. The company released a video for the CarePod, which will give you an idea of some of the apps and services you can get access to.

Health data and personalized plans will be continuously updated and managed in the Forward mobile app. Forward clinicians ultimately make all care decisions, ensuring an actual medical team is with you. Memberships start at $99/mo.

Research from Stanford and the University of Illinois has revealed another significant vulnerability in the safety mechanisms of large language models. The study demonstrated that the protections from Reinforcement Learning with Human Feedback (RLHF), which are critical to preventing harmful LLM outputs, can be effectively removed through fine-tuning.

Fine-tuning is a common process used to adapt models for specific tasks or preferences; but this research shows that it can also be used to negate safety protocols designed to censor harmful content. The ability to generate detailed instructions for illegal or unethical activities, usually restricted by these models through RLHF, becomes worryingly accessible. The researchers measured the success rate of the fine-tuned models on generating harmful content - and the fine-tuned versions of GPT-4 achieved up to a 95% harmful content rate, compared to only 7% for the base version of GPT-4.

Google DeepMind has just introduced the world’s most accurate weather forecaster. GraphCast, a groundbreaking and open source AI model, can offer 10-day forecasts with unparalleled accuracy in less than a minute. The model outperforms the industry's gold-standard High Resolution Forecast (HRES) system in both speed and precision. Utilizing deep learning and neural networks, GraphCast is trained on decades of historical weather data, enabling it to understand complex cause and effect relationships in weather patterns. This approach differs from traditional Numerical Weather Prediction, which relies on physical equations and supercomputers, and often requires hours for similar forecasts. It can also predict extreme weather events earlier than previous systems.

More in AI this week:

• Introducing ‘How Do You Use ChatGPT?’ - how the most interesting people in the world use ChatGPT to think, create, and grow

• Why the Godfather of AI fears what he’s built

• (Google DeepMind) Lyria: transforming the future of music creation

• (Bill Gates Blog) AI is about to completely change how you use computers

• Why AI-assisted bioterrorism became a top concern for OpenAI and Anthropic

• OpenAI’s six-member board will decide ‘when we’ve attained AGI’

Artificial Intelligence online short course from MIT

Study artificial intelligence and gain the knowledge to support its integration into your organization. If you're looking to gain a competitive edge in today's business world, then this artificial intelligence online course may be the perfect option for you.

On completion of the MIT Artificial Intelligence: Implications for Business Strategy online short course, you’ll gain:

• Key AI management and leadership insights to support informed, strategic decision making.

• A practical grounding in AI and its business applications, helping you to transform your organization into a future-forward business.

• A road map for the strategic implementation of AI technologies in a business context.

Get More Information

We’ll see you next Thursday - thanks for readin