In-context learning happens when a machine learning model can perform a task successfully after a short number of attempts, even though the information was not included in model's training data.Researchers at MIT, Stanford University, and Google conducted experiments to find out more about the process, and were able to show how machine learning language models inherently create smaller models within themselves when performing in-content learning.
GPTZero is a new service that purports to allow users to detect output generated by ChatGPT, using AI, just like ChatGPT. GPTZero are currently looking for users in education to help them develop a written-by-AI detector.
Researchers at Carnegie Mellon University have developed a WiFi based human pose sensing system that can identify multiple people and their positions in a room using the phase and amplitude of the WiFi signal, filtered using machine learning models. The system can map the UV coordinates of body regions and is comparable to image based recognition systems.
Hourups is a collection of 165 free tech skills tutorials on YouTube that are grouped by skill and domain types, such as Python, Javascript, front-end dev, digital marketing, data analysis, communication, & devops.
Google released a new Google Sheets extension called SimpleML, which allows users to predict gaps in sheet data, and export trained models on sheets data, using a no-code interface.
The recently released chatGPT general AI text model has been sending shockwaves around the internet. Jonas Degrave from Deep Mind wrote about how he used chatGPT to create a virtual machine, access the chatGPT website from the command line within the VM, and query chatGPT running in the VM, running in chatGPT.
Intel launched a deepfake video detector called FakeCatcher. The deep learning system boasts 96% accuracy by detecting and modelling the blood flow in the face.
