What Microsoft announced at Ignite 2023 - Weekly Roundup - Issue #441
Plus: OpenAI is working on GPT-5; Google Gemini is delayed to Q1 2024; Nvidia launches H200; first CRISPR medicine approved in the UK; DeepMind's new AI for weather forecasting; and more!
This week, Microsoft hosted Ignite, an annual conference aimed at developers and IT professionals. It serves as a platform for Microsoft to announce new products, services, and updates, particularly focusing on cloud services, AI, enterprise tools, and software development. If there was a theme for this year’s Ignite, it would be Azure as the world's computer and copilots.
Here are the most important new products and services Microsoft announced this week.
Azure, “the world's computer”
Satya Nadella, the CEO of Microsoft, opened his keynote with the new products and services coming soon to Azure. Nadella highlighted improvements Microsoft has made to make Azure more efficient and greener, stating that Azure is on track to achieve the target of using 100% energy from zero-carbon sources by 2025.
One of the most important announcements of the whole keynote were Microsoft’s custom chips for Azure. The first one, named Cobalt, is an in-house designed 128-core ARM-based CPU optimized for performance, power efficiency, and cost-effectiveness for general-purpose workloads. The second one is Maia, Microsoft’s very own AI accelerator chip designed to run cloud-based training and inferencing for AI workloads. Both Cobalt and Maia are already being used internally in Azure and are powering some Microsoft services. Cobalt-powered Azure services will be available to customers next year.
Microsoft will also be adding the newest high-performance GPUs from both Nvidia and AMD to Azure, including Nvidia’s newest tensor core GPU, H200 (more on H200 later in this week’s roundup).
New AI models are coming to Azure, including OpenAI’s latest models GPT-4 Turbo, GPT-4 with Vision (GPT-4V) and DALL-E 3, announced last week. Nvidia’s models under Nvidia AI Foundry are coming to Azure, too. Microsoft also introduced Models as a Service which offers pay-as-you-go inference model APIs and fine tuning. With Models as a Service, developers can start using AI models straight away without worrying about getting GPUs and setting up the models to work. Models as a Service will start with Meta’s Llama-2 but there are plans to bring in Mistral’s premium models and G42’s Jais model (the world’s highest-quality Arabic language model). Open-source AI models are coming to Azure, too, including Nvidia’s Nemotron, Meta’s Code Llama, OpenAI’s Whisper, Microsoft’s Phi-2, and more. To manage all those models, Microsoft is offering Azure AI Studio which aims to make deploying, training and securing AI models easier.
Copilots, copilots everywhere
Microsoft sees copilots as the future of interacting with computers. As Nadella said, “We're at a tipping point, this is clearly the age of copilots”.
To fulfil this vision, Microsoft is launching Microsoft Copilot. It will replace Bing Chat to become a standalone AI assistant for everyday use and it seems it will be Microsoft’s version of ChatGPT. There will also be an enterprise version of Microsoft Copilot that will add enterprise data protection.
Copilot for Microsoft 365, a new AI-powered companion that provides contextual suggestions and actions across all Microsoft 365 applications, aims to streamline workflows and boost productivity. And the newly announced Copilot Studio will enable people to create their own copilots for Microsoft 365.
GPTs, smaller and tailored versions of ChatGPT announced last week at OpenAI’s first Dev Day, can be integrated with copilots with Copilot Studio to extend their functionality.
Almost every Microsoft product and service has or soon will get its own copilot. Github Copilot is already helping developers be more productive. Microsoft hopes that newly announced copilots can do the same for security, sales and customer service teams. Azure will also get its own copilot to help simplify day-to-day IT administration.
Microsoft, the copilot company
“We are the copilot company. We believe in a future where there will be a copilot for everyone and everything you do” - Satya Nadella at Microsoft Ignite 2023
At the very end of his keynote, Nadella presented two examples of how AI can be used together with mixed reality and quantum. The mixed reality example is interesting. It envisions a world in which everything can be a prompt, including what you are looking at through Hololens. It is a interesting vision that looks like something from a sci-fi movie and the video below does a good job explaining it.
The second example of AI enhancing human intelligence was the convergence of quantum computing and AI, particularly in simulating chemical reactions. Using Azure Quantum Elements, a chemist can now create new chemical compounds, a task that might have traditionally taken three years using conventional computational methods, in just 9 hours with this new approach.
Ignite 2023 clearly shows that Microsoft is now a company that fully embraced AI and offers AI products and services on all levels, from Azure services running the AI models to user-facing copilots. It sees its future in AI-powered copilots for everyone and everything.
I warmly welcome all new subscribers to the newsletter this week. I’m happy to have you here and I hope you’ll enjoy my work.
The best way to support the Humanity Redefined newsletter is by becoming a paid subscriber.
If you enjoy and find value in my writing, please hit the like button and share your thoughts in the comments. Additionally, please consider sharing this newsletter with others who might also find it valuable.
For those who prefer to make a one-off donation, you can 'buy me a coffee' via Ko-fi. Every coffee bought is a generous support towards the work put into this newsletter.
Your support, in any form, is deeply appreciated and goes a long way in keeping this newsletter alive and thriving.
🧠 Artificial Intelligence
OpenAI chief seeks new Microsoft funds to build ‘superintelligence’
In an interview with the Financial Times, Sam Altman stated that OpenAI's partnership with Microsoft is “working really well” and that he hopes “to raise a lot more over time” from the tech giant, among other investors. Altman emphasizes the high costs associated with developing sophisticated AI models, mentioning the company's unprofitability due to training expenses. He also revealed that OpenAI is working on GPT-5, but Altman did not provide any details about its release date. He only mentioned that the GPT-4 successor is currently in the training phase. We can expect GPT-5 to be more sophisticated than GPT-4, but it is difficult to predict the capabilities and skills of the new model, Altman said.
NVIDIA launches H200 Tensor Core GPU
Nvidia has released the H200, a new product designed for AI and high-performance computing (HPC) workloads. The H200 utilizes the same silicon as the H100 but offers more memory and memory bandwidth. It features HBM3e memory, boasting 144GB of memory compared to the H100's 80GB, and achieves up to 4.8 TB/sec peak memory bandwidth, which is a 40% increase over the H100’s 3.35 TB/s. In terms of performance, Nvidia claims H200 is up to 2x faster than H100 in LLM inference tests and uses 50% less energy than H100. Major cloud service providers such as Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure will be the first to deploy H200-based instances starting next year Nvidia says the H200 will be available "from global system manufacturers and cloud service providers" starting in Q2 2024.
SAG-AFTRA Deal - Analysis Of AI & Streaming Fund
SAG-AFTRA has reached a tentative agreement with movie studios, ending one of the longest strikes in Hollywood’s history. Although not everyone is happy with the deal, it enables actors to receive success-based bonuses for streaming shows and regulates the use of AI in movie production. Movie studios are required to obtain consent from actors before creating their digital replicas, and actors must be compensated for this. The deal also sets rules and guidelines on the use of generative AI in movies and for creating background actors with AI. This article on Deadline provides a detailed explanation of the new rules for using AI in movies. For more in-depth information, here is an 18-page long summary of the deal.
Google Delays Release of Gemini AI That Aims to Compete With OpenAI
The Information reports that Gemini, Google’s highly anticipated next-generation AI model aimed at competing with OpenAI’s GPT-4, has been delayed to Q1 of 2024.
China May Stay Permanently Behind the US in Generative AI
In his piece on
Google DeepMind’s weather AI can forecast extreme weather faster and more accurately
After solving the protein-folding problem with AlphaFold, DeepMind has now turned its attention to weather forecasting in its ongoing quest to apply AI to advance science. This week, DeepMind released a paper describing GraphCast, a weather forecasting AI model. GraphCast proved its efficacy in September when it accurately predicted Hurricane Lee's landfall in Nova Scotia nine days ahead of time, three days earlier than previous models. Google DeepMind has made GraphCast open source. The introduction of GraphCast marks a revolutionary shift in weather forecasting, challenging both traditional models and other AI-driven models like Huawei’s Pangu-Weather and Nvidia’s FourcastNet.
If you're enjoying the insights and perspectives shared in the Humanity Redefined newsletter, why not spread the word?
🤖 Robotics
AI robotics’ ‘GPT moment’ is near
What will be the next major advancement in AI? According to a recent article, it will be robotics. AI-powered robots that learn to interact with the physical world have the potential to revolutionize repetitive tasks in various sectors, including logistics, retail, and healthcare. The article predicts a rapidly approaching 'GPT moment' for AI in robotics, where robotic applications, particularly those requiring precise object manipulation, are increasingly being deployed in real-world production environments. The growth trajectory of robotic foundation models is on the rise, and 2024 is expected to witness an exponential increase in commercially viable robotic applications on a large scale. It’s worth noting that DeepMind is making progress in the application of transformer models to robotics. While OpenAI has had some research projects in robotics, they do not appear to be a current priority. However, OpenAI has invested in 1X Technologies, a company focusing on bringing humanoid robots to the market.
🧬 Biotechnology
The First CRISPR Medicine Just Got Approved in the UK
The United Kingdom authorized the first medical treatment that uses CRISPR gene editing. This one-time therapy, branded as Casgevy, is designed for patients with sickle cell disease and a related blood disorder called beta thalassemia, both of which are inherited conditions. The UK Medicines and Healthcare Products Regulatory Agency's authorization applies to sickle cell patients experiencing recurrent pain crises and to those with the most severe form of beta thalassemia, aged 12 years and older. Vertex and CRISPR Therapeutics estimate that approximately 2,000 people in the UK are eligible for this therapy.
Scientists create artificial protein capable of degrading microplastics in bottles
Researchers from Spain have developed artificial proteins that can degrade PET microplastics and nanoplastics (PET is a plastic commonly used in packaging and bottles). This development is significant given the annual production of about 400 million tons of plastics worldwide, which contributes to environmental issues and climate change. This new protein, engineered with the aid of machine learning and supercomputers, can degrade PET more efficiently and at room temperature, unlike current methods that require higher temperatures, leading to high CO2 emissions.
The Rise of “Wet” Artificial Intelligence
AI is transforming every aspect of human activity, and science is one of them. This article introduces the concept of 'wet AI,' which refers to AI models that drive wet lab experiments, and highlights how AI, especially deep learning, understands intricate patterns in biological data. However, challenges remain, such as the need for more comprehensive datasets to improve AI performance in biology. “The most exciting aspect of wet AI, however, is that it promises to turn the molecular world into a programmable medium. That could have far reaching impacts, giving us the ability to imagine new categories of products based on engineered biology”, concludes the article.
Organoids Evolve from Academic Marvel to Industrial Tool
Organoids have come a long way, evolving from a cellular curiosity in the early 20th century to one of the most powerful research tools in science and medicine. This article focuses on the latest phase of that journey: transitioning from challenging-to-produce academic tools to mass-produced industrial tools for healthcare research and treatment development that will help us better understand human biology and diseases.
Thanks for reading. If you enjoyed this post, please click the ❤️ button or share it.
Humanity Redefined sheds light on the bleeding edge of technology and how advancements in AI, robotics, and biotech can usher in abundance, expand humanity's horizons, and redefine what it means to be human.
A big thank you to my paid subscribers, to my Patrons: whmr, Florian, dux, Eric, Preppikoma and Andrew, and to everyone who supports my work on Ko-Fi. Thank you for the support!