Issue 70 17th September 2024

Highlights

OpenAI unveils a model that can fact-check itself

ChatGPT maker OpenAI has announced its next major product release: An generative AI model code-named Strawberry, officially called OpenAI o1. To be more precise, o1 is actually a family of models. Two are available today, o1-preview and o1 mini, in both chatbot form and OpenAI’s API.

MedFuzz: Exploring the robustness of LLMs on medical challenge problems

Large language models (LLMs) have achieved unprecedented accuracy on medical question-answering benchmarks, showcasing their potential to revolutionize healthcare by supporting clinicians and patients. However, these benchmarks often fail to capture the full complexity of real-world medical scenarios.

Real v fake: how the Harris-Trump debate laid out different takes on AI

One candidate revels in AI-generated images of cats and geese, while the other posts real photos of her grandparents In their first, and likely only debate, Kamala Harris and Donald Trump argued about artificial intelligence. They spoke of China, chips and “domestic innovation”. The country learned how Harris, Trump and their allies would – or intentionally wouldn’t – use artificial intelligence for their own ends. But the real lessons were in the aftermath. The online furor over the IRL confrontation revealed that Republicans use AI to illustrate their political points. Democrats do not.

Research

OpenAI o1: ChatGPT Supercharged!

TwoMinutePapers

Federated Multilingual Models for Medical Transcript Analysis

Federated Learning (FL) is a novel machine learning approach that allows the model trainer to access more data samples, by training the model across multiple decentralized data sources, while data access constraints are in place. Such trained models can achieve significantly higher performance beyond what can be done when trained on a single data source. As part of FL's promises, none of the training data is ever transmitted to any central location, ensuring that sensitive data remains local and private. These characteristics make FL perfectly suited for large-scale applications in healthcare, where a variety of compliance constraints restrict how data may be handled, processed, and stored. Despite the apparent benefits of federated learning, the heterogeneity in the local data distributions pose significant challenges, and such challenges are even more pronounced in the case of multilingual data providers. In this paper we present a federated learning system for training a large-scale multi-lingual model suitable for fine-tuning on downstream tasks such as medical entity tagging. Our work represents one of the first such production-scale systems, capable of training across multiple highly heterogeneous data providers, and achieving levels of accuracy that could not be otherwise achieved by using central training with public data. Finally, we show that the global model performance can be further improved by a training step performed locally.

Video

Episode 302 - Teams AI library with Ayça Baş

Join us in this episode where we talk with Ayça Baş, Senior Cloud Advocate at Microsoft about Teams AI library and building custom engine copilots for M365.

Abdul Rasheed Feroz Khan - Humans in AI

Dive into the future with Abdul Rasheed Feroz Khan, a Microsoft MVP on Azure and CEO of CodeSizzler! Discover how AI is revolutionizing industries from healthcare to logistics and adding immense value every day.

Articles

Mistral releases Pixtral, its first multimodal model

French AI startup Mistral has released its first model that can process images as well as text. Called Pixtral 12B, the 12-billion-parameter model is roughly 24GB size. (Parameters roughly correspond to a model’s problem-solving skills, and models with more parameters generally perform better than those with fewer parameters.) Available on GitHub as well as the AI and machine […]

Do you see blue or green? This viral test plays with color perception

A visual neuroscientist realized he saw green and blue differently to his wife. He designed an interactive site that has received over 1.5m visits It started with an argument over a blanket. “I’m a visual neuroscientist, and my wife, Dr Marissé Masis-Solano, is an ophthalmologist,” says Dr Patrick Mineault, designer of the viral web app ismy.blue. “We have this argument about a blanket in our house. I think it’s unambiguously green and she thinks it’s unambiguously blue.”

Notes on OpenAI's new o1 chain-of-thought models

OpenAI released two major new preview models today: o1-preview and o1-mini (that mini one is not a preview) - previously rumored as having the codename "strawberry". There's a lot to understand about these models - they're not as simple as the next step up from GPT-4o, instead introducing some major trade-offs in terms of cost and performance in exchange for improved "reasoning" capabilities.