Meta Introduces Self-Supervised Vision Transformer Model Dinov2

By

-

April 20, 2023

Meta Self-Supervised Vision Transformer Model DinoV2 — Image Credits: Meta

Meta AI announced the release of Dinov2 two years after debuting DINO, a self-supervised vision transformer model. In contrast to other models of a similar type, like CLIP, this one performs exceptionally well and doesn’t need modifications.

Meta accomplished this by pretraining on a vast volume of unprocessed text utilizing pretext objectives like non-supervised word vectors or language modeling. The model is free source and has been pre-trained on 142 million photos in an unlabeled, self-supervised manner.

High-performance characteristics offered by DINOv2 can be used as direct inputs for basic linear classifiers. Due to its adaptability, DINOv2 may be used to build multifunctional backbones for a variety of computer vision jobs, according to a blog post from the company.

DinoV2 assists with tasks like depth estimation, image classification, semantic segmentation, and image retrieval without the requirement for expensive labelled data, which will save developers a great deal of time and resources.

According to Meta, the model uses self-supervised learning and produces results that are comparable to or better than those produced by the traditional approaches used in the relevant fields.

Self-supervised learning is the key appeal since it enables DINOv2 to create adaptable, general frameworks for a range of computer vision applications and tasks. Before using the model in various domains, no further fine-tuning is necessary.

Stability AI Announces StableLM, A New Open-source Large Language Model

By

Sahil Pawar

-

April 20, 2023

Stability AI announces StableLM — Image Credits: Stability AI

A group of open-source large language models (LLMs) referred to as StableLM have been made available by Stability AI, the firm that created the AI-powered Stable Diffusion image generator. In a post published on Wednesday, the business revealed that its models are now accessible for developers to use and modify on GitHub.

StableLM is made to efficiently create text and code, much as ChatGPT’s opponent. The open-source dataset called Pile, which includes data from a variety of sources such as Wikipedia, Stack Exchange, and PubMed, served as its training data.

According to Stability AI, StableLM models presently have parameters ranging from 3 billion and 7 billion, with models having 15 billion to 65 billion parameters coming later.

In addition to building on the open-source language models that Stability AI has already developed in partnership with the nonprofit EleutherAI, StableLM also advances the company’s goal of making AI tools more approachable, as it did with Stable Diffusion.

The optimized conversation model from StableLM is available for testing in a demo on Hugging Face. Although the datasets Stability AI employs should steer the base language models into safer distributions of text, the company issues a warning that “not all biases and toxicity can be mitigated through fine-tuning.”

Snapchat Releases its AI Chatbot ‘My AI’ to Everyone for Free

By

Sahil Pawar

-

April 20, 2023

Snapchat releases AI chatbot 'My AI' everyone free — Image Credits: TechCrunch

Less than two months after it was initially made available to the app’s over 3 million paying subscribers, Snap is making its “My AI” chatbot available to all 750 million monthly users of Snapchat for free.

Additionally, Snapchat is integrating My AI more and more to its initiatives. By mentioning it in group conversations with the @ sign, it can now be added. Snap will enable users to give their bot a unique Bitmoji avatar to change its name and appearance.

Additionally, My AI will now suggest locations to visit from the map tab of the Snapchat app or AR filters to use in the app’s camera. As an example, during the company’s annual conference today, a photo of tomatoes in a garden prompted the bot to answer with a generated image of gazpacho soup. Snap also hopes to soon enable users to visually message My AI and receive AI generated responses.

While Microsoft and Google race to incorporate generative AI into their search engines, Evan Spiegel, the CEO of Snap, views the technology as “an awesome creative tool.” In a recent interview, he gave specific instances of how he used My AI to arrange Miranda Kerr’s birthday itinerary and write bedtime stories for his kids. He claims that My AI is already facilitating more than 2 million chats each day.

Regarding My AI’s possible effects on Snap’s advertising business, which has had significant growth hurdles, Spiegel has been quiet. He acknowledges that using My AI’s interactions for ad targeting may be a chance, but he doesn’t go into further detail, leaving room for future advances to be hinted at.

Meta Opens Access to Its Social VR Space Horizon Worlds for Teens

By

Sahil Pawar

-

April 19, 2023

Meta opens access Horizon Worlds teens — Image Credits: TechCrunch

According to a Tuesday announcement by Meta, teens in the US and Canada will soon have access to the social VR platform Horizon Worlds. Meta will have certain extra features in place for teen players, defined as those between the ages of 13 and 17, to assist create a safer experience on the platform.

For instance, teen accounts will be private by default, and Meta will not display a teen’s online status unless they want to do so. Teens will also by default have access to the “voice mode” option, which converts unfamiliar voices into quiet, friendly sounds and muddles your own voice.

In the “people you might know” tab, Meta won’t display any adults who the teenager doesn’t know. Additionally, the business claims to have included its VR parental control toolkit to Horizon Worlds.

The network, which has had trouble keeping users, might draw more players as a result of the teens-only expansion. However, the choice has received a lot of backlash. After learning about his intentions via The Wall Street Journal in February, Senators Ed Markey and Richard Blumenthal wrote to Mark Zuckerberg in March and pleaded with him to abandon them.

The senators wrote, “Meta’s plan to target young people with offerings in the metaverse is particularly concerning in light of your consistent failures to protect young users. With a documented track record of failure to protect children and teens, Meta has lost parents’, pediatricians’, policymakers’, and the public’s trust.”

By making Horizon Worlds more approachable to teenagers, Meta is increasing pressure on Horizon Worlds to compete with other metaverse-like platforms like Fortnite and Roblox, which also feature specific safeguards for younger users. Furthermore, Meta should soon make Horizon Worlds available on the web and mobile devices.

Meta Prepares for Another Round of Layoffs Across Facebook, WhatsApp, and Instagram

By

Sahil Pawar

-

April 19, 2023

Meta prepares for another round of layoffs — Image Credits: USNews

Meta Platforms has started with another round of layoffs across the board today as it reorganizes teams and pursues founder Mark Zuckerberg’s vision of higher productivity.

Facebook’s parent firm alerted managers in a note acquired by Bloomberg News that job cuts would be announced on Wednesday. It implies that the company’s virtual reality initiatives at Reality Labs, Facebook, WhatsApp, Instagram, and Quest hardware will all be impacted.

According to Zuckerberg’s announcement in March, the action is a part of a cost-cutting drive that will eventually result in the dismissal of 10,000 jobs at the corporation. In May, there will be yet another round of layoffs.

In November, Meta had laid off 11,000 employees, or around 13% of its staff. Additionally, it prolonged a hiring ban throughout the first quarter. According to Zuckerberg’s remarks, the corporation intends to cut costs and equalize the proportion of technologists and engineers to business and administrative workers.

According to the memo sent to managers, teams will be reorganized and a number of surviving individuals will be reassigned to work for new supervisors. Meta has requested that all North American workers who are able to work remotely to do so on Wednesday so that they can grapple with the news.

A representative for the company declined to comment, citing Zuckerberg’s “year of efficiency” announcement from March, which stated, “We expect to announce restructurings and layoffs in our tech groups in late April, and then our business groups in late May.”

Philips, IMES Create Enhanced Critical Care Data Set for AI in Healthcare

By

Sahil Pawar

-

April 19, 2023

Philips IMES Critical Care Data Set AI Healthcare — Image Credits: Regulatory Review

In order to advance machine learning and artificial intelligence in healthcare, Royal Philips, a provider of health technology, today announced a collaboration with MIT’s Institute for Medical Engineering and Science (IMES).

De-identified data from 200,000 critical care patients, including those affected by COVID-19, are included in the revised eICU Collaborative Research Database (eICU-CRD). The creation of solutions that enhance patient care and clinical outcomes will be aided by the larger and more clinically reliable data collection.

The COVID-19 pandemic resulted in a sharp rise in the number of patients in the eICU and critical care settings as well as particular difficulties in the delivery of care. This led Philips and IMES to expand the initial data set, first made public in 2016.

De-identified and comprehensive clinical data, including vital signs, prescription drug prescriptions, laboratory test results, diagnoses, and severity of disease scores, are contained in the new secure database. Comprehensive information on patient therapies, comorbidities, readmissions, and clinical outcomes is also provided by the dataset.

Researchers from across the world will have access to the data, thanks to researchers at Philips and the Laboratory of Computational Physiology at IMES. This will allow researchers to create cutting-edge algorithms and offer fresh perspectives on critical care.

The Laboratory of Computational Physiology will be the initiative’s academic research center, providing and maintaining access as well as assisting with database education and providing a forum for collaboration. To individuals who have the necessary credentials, have completed human subjects training, and have signed a data usage agreement, the database is available for use in medical research.

Press Hook Launches Free AI-Based Press Release Generator

By

Sahil Pawar

-

April 19, 2023

Press Hook Launches AI-Based Press Release Generator — Image Credits: Press Hook

A free press release generating tool has been launched by Press Hook, a provider of public relations platforms, enabling businesses to create automatic news releases. Deep learning will be used by the tool, which employs autoregressive language processing models and artificial intelligence to produce content.

Users need only fill out a basic form with the necessary information such as the name of the company or product, a quote, and a few brief descriptors, and the generator will instantly produce a press release. The release consists a news summary, bullet points, and website links.

The GPT (Generative Pretrained Transformer), a language processing AI model created by OpenAI, powers the generator.

Michelle Songy, the founder of Press Hook, said, “We’ve written thousands of press releases over the years, so we know how difficult and time-consuming it can be for the small businesses to write compelling stories that are optimized and media-ready.”

“We created this tool to assist businesses in producing highly readable, engaging material that is formatted and prepared for distribution, allowing them to concentrate on the content of their story instead of being bogged down in the mechanics of telling it,” she added.

Press Hook has developed a platform where businesses can get the tools they need to take control of their media relations, media outlets can proactively find sources, and PR can maximize client visibility all at once by prioritizing affordability and openness.

Finance Minister Sitharaman Says Crypto Issue Requires Immediate Attention

By

Sahil Pawar

-

April 19, 2023

FM Sitharaman says crypto issue requires immediate attention — Image Credits: NDTV

According to Union Finance Minister Nirmala Sitharaman, issues relating to crypto assets demand immediate attention. She said that the G20’s response on the same must make sure that people do not lose any potential benefits while safeguarding economies from harm.

On April 14, Sitharaman took part in a brainstorming session on the “Macrofinancial Implications of Crypto Assets” with G20 finance ministers and governors of central banks. India presently holds the G20 countries’ yearly rotating chairmanship.

Crypto-related issues have been a hot topic of conversation among G20 members, and all agree that this industry has to be regulated immediately. There were international experts present for the brainstorming session to discuss this topic.

In her remarks, Sitharaman stated that the G20 recognises the efforts of the Financial Stability Board (FSB) and the International Monetary Fund (IMF) in identifying important components of the policy and regulatory framework.

In addition, she stated that a synthesis work that incorporates macroeconomic and regulatory perspectives on crypto assets is necessary. The Minister remarked that there is an agreement among G20 countries to have a globally coordinated policy response on digital currency that accounts for all potential risks, including those unique to developing and emerging nations.

After the recent shocks in the cryptocurrency market, Sitharaman stated earlier this week that India’s G20 Presidency wants to create a uniform framework for all governments to deal with the risks involved with cryptocurrencies.

AI Training Dataset Market to Reach $9.3 Billion by 2031, Says Allied Market Research

By

Sahil Pawar

-

April 18, 2023

AI Training Dataset Market reach $9.3 Billion by 2031 — Image Credits: Info

Allied Market Research released a report titled “AI Training Dataset Market by Type (Text, Audio, Image/Video), by End User (IT and Telecom, BFSI, Automotive, Healthcare, Government and Defense, Retail, Others): Global Opportunity Analysis and Industry Forecast, 2021-2031.”

The report estimates that the global market for artificial intelligence training datasets generated $1.4 billion in 2021 and will grow at a CAGR of 21.6% from 2022 to 2031 to reach $9.3 billion. The research provides a thorough analysis of the industry’s shifting trends, leading market segments, important investment areas, value chains, regional environments, and competitive scenarios.

By type, text held the lion’s share in 2021, accounting for over two-fifths of the revenue generated by the worldwide AI training dataset market. By 2031, it is anticipated that text will rule the roost. On the other hand, the image/video segment will exhibit the fastest CAGR of 23.3% during the predicted period. This is a result of important players putting more effort into launching new datasets and applications.

In terms of end users, the IT and telecom segment held about one-fourth of the global AI training dataset market share in the year 2021 and is anticipated to take the lead by 2031. On the other side, the healthcare segment will experience the fastest CAGR of 24.9% during the predicted period. This is due to the wide range of therapeutic potential provided by AI in the healthcare industry, including management of lifestyle and wellness, diagnostics, virtual assistants, and wearable technology.

By region, North America had the biggest share in 2021, accounting for almost one-third of the market’s revenue for AI training datasets worldwide. By 2031, it is expected to continue to maintain the top spot. During the predicted period, the Asia-Pacific area would simultaneously exhibit the fastest CAGR of 24.1%, as a result of the broad availability of new datasets that will hasten the use of artificial intelligence technology in developing sectors.

Government School from Raipur to Offer Classes in AI, ML and Robotics

By

Sahil Pawar

-

April 18, 2023

Government school Raipur offer classes AI ML Robotics — Image Credits: RD School

As a part of the “Jigyasa” project, R. D. Tiwari Government English School from Raipur has become the tenth government school in India to establish a robotics lab. 500 students from various classes can learn about building models and technological equipment in this lab.

In addition to Raipur, the Jigyasa initiative also covers Rajasthan, Maharashtra, Madhya Pradesh, and Gujarat. In the lab, students will have the opportunity to learn how to construct robots, technical models, electronic gadgets, and more.

While those in grades 6 through 10 will be trained on how to construct electrical models, students in grades 4 and 5 will learn about models without electronic components. In the robotics lab, kids will also study computer programming.

Exposing the students to the creation of models, coding, and robots can enhance their creative ability. Anuj Kumar Dwivedi, headmaster of Pt. R. D. Tiwari Government English School, says that the main objective of this project is to increase students’ interest in crucial subjects like physics and mathematics. He added that everything in the laboratory was purchased from Sweden.

The headmaster also mentioned that a mentor had been recruited to train kids in robotics and programming. He believes that the lab will help the kids get better jobs in the future. According to Saurabh Verma, the lab’s mentor, children in the primary level will study designing through an interactive session.

Particular scientific principles and load balance will be taught to the students. The students would be trained in the robotics lab to develop models that can address problems affecting common people, such as the design of traffic lights and the software that makes them function.