This Week I Learned - Week #49 2023
This Week I Learned -
* Streamlit is an open-source app framework for Machine Learning and Data Science teams
* FoodData Central is an integrated data system that provides expanded nutrient profile data and links to related agricultural and experimental research. It provides nutrition data of ~500K branded food products. The FoodData Central API provides REST access to FoodData Central (FDC). USDA FoodData Central data are in the public domain and they are not copyrighted. They are published under CC0 1.0 Universal (CC0 1.0)
* Gemini succeeds PaLM 2, the current foundation model from Google.
* Internal tests showed that Q, Amazon's AI-powered assistant that enables employees to query documents and corporate systems, could leak confidential information from Amazon such as internal discount programs, unreleased features, and locations of AWS data centers. Amazon spokespeople called such scenarios hypothetical and denied that Q had leaked such information. Amazon is not the only major AI company whose chatbot has leaked private information. Google researchers recently found that they could prompt OpenAI’s ChatGPT to divulge personal information found in its training data. For Amazon, issues with a newly released system are a bump in the road to competing effectively against competitors like Microsoft Copilot and ChatGPT Enterprise. For developers, it’s a sobering reminder that when you move fast, what breaks may be your own product. - The Batch
* Big picture of the modern GenAI stack. Almost every GenAI-related service available from AWS, GCP, and Azure can be mapped to this. - Janakiram MSV
* Azure is the new AI Operating System, and Copilots are the new Apps. When Microsoft creates a platform, it leads to a new ecosystem of independent software vendors and solution providers, helping enterprises leverage it. With AI, Microsoft wants to repeat the magic of creating a brand new platform that results in a thriving ecosystem of developers, ISVs, system integrators, enterprises and consumers. This season, Azure becomes the operating system, providing the runtime and platform services, while the apps are the AI assistants that Microsoft calls copilots. So, Azure is the new Windows and copilots are the new applications.
* Azure ML, Microsoft’s ML PaaS offers model-as-a-service to consume foundation models as an API without the need to provision GPU infrastructure. Azure ML now supports additional open source foundation models, including Llama, Code Llama, Mistral 7B, Stable Diffusion, Whisper V3, BLIP, CLIP, Flacon and NVIDIA Nemotron. - Forbes
* Vector Database Feature Matrix by Dhruv Anand
* The LVM (large vision model) revolution is coming a little after the LLM (large language model) one, and will transform how we process images. But there’s an important difference between LLMs and LVMs:
- Internet text is similar enough to proprietary text documents that an LLM trained on internet text can understand your documents.
- But internet images – such as Instagram pictures – contain a lot of pictures of people, pets, landmarks, and everyday objects. Many practical vision applications (manufacturing, aerial imagery, life sciences, etc.) use images that look nothing like most internet images. So a generic LVM trained on internet images fares poorly at picking out the most salient features of images in many specialized domains.
That’s why domain specific LVMs – ones adapted to images of a particular domain (such as semiconductor manufacturing, or pathology) – do much better. At Landing AI, by using ~100K unlabeled images to adapt an LVM to a specific domain, we see significantly improved results, for example where only 10-30% as much labeled data is now needed to achieve a certain level of performance.
For companies with large sets of images that look nothing like internet images, I think domain specific LVMs can be a way to unlock considerable value from their data. Dan Maloney and I share more details in the video. Andrew Ng on LinkedIn
* Al Sweigart, the author of Automate the Boring Stuff with Python is working on the third edition of the book and removing some (but not all) of the RoboCop references. He is trying to make the book easier to translate to languages and cultures outside of English.
* Although the BBC is politically independent, its chairperson is appointed by the government. As BBC chairman, a three-day-a-week role with an annual salary of 160,000 pounds, 71-year-old Dr Samir Shah, an India-born media executive with over 40 years of experience in TV production and journalism will be responsible for upholding and protecting the taxpayer-funded licence fee-operated public broadcaster and ensuring it fulfils its mission to “inform, educate and entertain”.
* Adele Goldberg (born July 22, 1945) is one of the co-developers of the programming language Smalltalk.
* How to Have an American Baby, a documentary directed by Leslie Tai, investigates the complex shadow economy of birth tourism. Chinese-based companies rent out maternity hotels in the U.S., where expecting Chinese mothers are looked after as they carry their pregnancies to term. Because the babies are born on American soil, they're granted U.S. citizenship.
* Forbes has been keeping track of the world’s billionaires since 1987. Earlier this year, we found 2,640 of them for our annual list.
* Urban (urban centres with over 100,000 population) and metropolitan markets (population of over one million) account for 80% of term deposits, indicating the skewed nature of the deposit mobilisation pattern in the country.
* The third Telangana Assembly has 50 newcomers of 119 elected members, with 15 doctors, four of them orthopaedicians.
* "The failures of capitalism are not failures but rather capitalism working as intended." - Al Sweigart
* "I know only one thing: that I know nothing." — attributed to Socrates
* "We don’t learn by putting information into our heads; we learn by trying to retrieve information from our heads." - Trey Hunner
Comments
Post a Comment