Cloud Hosting

AI has, by now, proven its power and impact. The artificial intelligence space is constantly evolving and improving with every passing day. Tech companies and researchers are investing big in bringing out innovations due to the massive potential the impact of AI can hold on the worlds biggest problems. As we head towards the end of 2021, let us look back at some of the major AI innovations and incidents that took centre stage this year.

OpenAI released DALLE, a 12-billion parameter version of GPT-3 trained to generate images from text descriptions, using a dataset of text-image pairs. OpenAI said that DALLE is a transformer language model that receives both the text and the image as a single stream of data containing up to 1280 tokens. It added that DALL.E can render an image from scratch and also alter aspects of an image using text prompts.

For more information, refer here.

Google released TensorFlow 3D (TF 3D) that is designed to bring 3D deep learning capabilities into TensorFlow. The tech giant said that TF 3D comes with popular operations, loss functions, and data processing capabilities that will help to develop, train and deploy 3D scene understanding models.

For more details, refer here.

Meta AI released SEER (SElf-supERvised), a billion-parameter self-supervised computer vision model that can learn from any random group of images on the internet. It does not need the careful curation and labelling that most computer vision training models need. The company also said that SEER also outperformed state-of-the-art supervised models on downstream such as low-shot, object detection, segmentation, and image classification.

For more details, refer here.

The EUs executive branch European Commission published a proposal for a Regulation on Artificial Intelligence. This aims to place mechanisms and restrictions on the use of AI, its violations, AI regulatory requirements, among others.

For more details, refer here.

In May, at the Google I/O event, Google announced the general availability of Vertex AI. It is a managed machine learning platform that will allow companies to speed up the deployment and maintenance of AI models, Google claimed. Google also said that Vertex AI requires nearly 80% fewer lines of code to train a model compared to competitive platforms. It will help data scientists and ML engineers implement machine learning operations to build and manage ML projects throughout the development cycle.

For more details, refer here.

OpenAI and Microsofts GitHub Copilot is an AI pair programmer to write better code. GitHub Copilot works with different languages like Python, JavaScript, TypeScript, Ruby, Java, and Go. GitHub Copilot can be used as an extension on the desktop or in the cloud on GitHub Codespaces. The company said that with the Copilot, the programmer can look at alternative suggestions, choose what to accept and reject, and edit suggested code manually.

For more details, refer here.

The code of AlphaFold 2.0 was made open-sourced by DeepMind. This AI algorithm predicts the shape of proteins, which is a major challenge in life sciences. In 2018, AlphaFold 1.0 was released, though it proved to be not good enough to employ researchers in the field. After further improvement, AlphaFold 2.0 was released in December 2020 and has received much appreciation. By making the source code public, DeepMind aims to offer better research opportunities to the scientific community in areas like drug discovery.

For more details, refer here.

At the Tesla AI Day, Elon Musk announced that the company is working on a humanoid robot. He added that Tesla will build a robot in a human form that could perform repetitive tasks, and the prototype is likely to be ready by next year. The code name for the bot is Optimus. Teslas director Ganesh Venkataramanan showed the computer chip that Tesla uses to run its supercomputer, Dojo. It contains 7nm technology and is packed with 362 teraflops of processing power.

For more details, refer here.

Toshiba Corporation unveiled Visual Question Answering (VQA) AI that can recognize people and objects, colours, shapes, appearances, and background details in images. It said that this mechanism solves the issue of answering questions on the positioning and appearance of people and objects. It can learn the information required to handle a wide range of questions and answers and can find applications to a diverse range of applications without any customization requirements.

For more details, refer here.

In a major move, CEO Mark Zuckerberg announced that Facebook has changed its name to Meta at Connect 2021 held recently. The reach of Meta will be much beyond just social media. The metaverse will give the feeling of a hybrid structure of online social experiences expanded into the physical world.

For more details, refer here.

NVIDIA has also jumped on the metaverse bandwagon. It announced Omniverse VR, where creators, designers, researchers, and engineers can connect major design tools and assets to collaborate in a shared virtual space. It also revealed Omniverse Avatar, a new platform for creating interactive AI avatars with the help of computer vision, NLP, and simulation technologies. The company also showed us the NVIDIA Omniverse Replicator, a synthetic data generation engine to train deep neural networks. NVIDIA also announced that Omniverse Enterprise is now generally available.

For more details, refer here.

Alphabet announced the launch of Isomorphic Labs that aims to accelerate the drug discovery process. Demis Hassabis has taken over as the founder and CEO of Isomorphic Labs. He posted on Twitter that the goal is to reimagine the drug discovery process from first principles with an AI-first approach.For more details, refer here.

Original post:
AI Innovations That Made Headlines In 2021 - Analytics India Magazine