Galactica - Novel AI Model for Scientific Assistance

On November 15 Meta unveiled a new large language model called Galactica.

Galactica is the first step towards a scientific assistant. This tool might be useful in the future in augmenting scientific work. The model is expected to help researchers who are buried under a mass of papers, increasingly unable to distinguish between the meaningful and the inconsequential. It can perform scientific NLP tasks at a high level, as well as tasks such as citation prediction, mathematical reasoning, molecular property prediction, and protein annotation. Meta promoted its model as a shortcut for researchers and students. In the company’s words, Galactica “can summarize academic papers, solve math problems, generate Wiki articles, write scientific code, annotate molecules and proteins, and more.”

The Galactica language model is an illustration of NLP for science and is trained on 48 million examples of scientific articles, websites, textbooks, lecture notes, and encyclopedias. The training data is a high-quality scientific dataset called NatureBook, making the models capable of working with scientific terminology, math, and chemical formulas as well as source codes.

It was introduced by Meta AI through the release of this Paper and this GitHub Repository. With a few adjustments to its current limitations. Galactica is expected to revolutionize scientific fields such as Machine Learning, Mathematics, Computer Science, Biology, Physics, and Chemistry. Galactica demonstrates the potential for language models as a new interface for science and has been open sourced.

Enroll at HURU School to learn more about Artificial Intelligence and NLP Models.

What is Pytorch?

PyTorch is a deep learning library built on top of the Torch library.

It provides a powerful and flexible set of tools for building and training neural networks, with a focus on ease of use and flexibility. It is one of the most popular AI frameworks.

PyTorch is an open-source machine learning library for Python, based on the Torch library. It was developed by Facebook's AI Research group led by Soumith Chintala. and was released in January 2017. PyTorch provides a wide range of tools for deep learning, including support for advanced neural network architectures such as convolutional and recurrent neural networks.

It also includes support for GPU acceleration, making it a popular choice for training large, complex models. Some popular models that have been created using PyTorch include ResNet, DenseNet, and Generative Adversarial Networks (GANs). PyTorch has also been used to develop natural language processing (NLP) models such as BERT and GPT-2.

PyTorch includes a variety of pre-built modules for common tasks such as image classification and natural language processing, as well as the ability to define and train custom models. It also includes support for dynamic computation graphs, which allows

for more efficient and flexible model design.

Additionally, PyTorch has a large and active community that contributes to the development and maintenance of the library, making it a popular choice for deep learning research and development.

This Article was written by Abednego Emonena from the HURU School AI Class

OWL-ViT Algorithm from Google AI

The OWL-ViT (short for Vision Transformer for Open-World Localization) was proposed in Simple Open-Vocabulary Object Detection with Vision Transformers by Matthias Minderer among others. It is a recent model and not many citations have been made on it. It can be used to query an image with one or multiple text queries to search for and detect target objects described in the text.

Its practical applications include:

• Pedestrian detection,

• People counting

• Face detection,

• Text detection,

• Pose detection,

• Number-plate recognition.

This has significant implications for downstream edge devices and applications.

This article was written by Danson Waweru of the AI Class of 2022. Learn more about Artificial Intelligence at HURU School by Enrolling here.