Site icon

The neural network “fixed” the Little Mermaid and other AI news

We strive to inform readers not only about events in the bitcoin industry, but also to talk about what is happening in related technological areas – cybersecurity and in the world of artificial intelligence (AI).

Cryplogger AI collected the most important news from the world of AI over the past week.

  • “Yandex” has closed a deal to sell the “News” and “Zen” services to the VK group.
  • Apple has released iOS 16 and watchOS 9 with AI features.
  • Meta singled out the developer of the PyTorch machine learning framework into a separate structure.
  • Google AI introduced a visual-text neural network with 17 billion parameters, working with 100 languages.
  • Twitter has suspended the account of a user who, using a neural network, “recolored” the skin color of the actress who played the Little Mermaid in the 2023 movie of the same name.
  • The most important transactions in the field of AI.

Media resources of “Yandex” came under the control of VK

Company “Yandex” closed a deal to sell Zen and Novosti media resources to VK for an undisclosed amount.

The parties received the approval of the Federal Antimonopoly Service of the Russian Federation.

According to Yandex, from September 12, users will begin to notice changes related to the sale of media resources.

Apple releases iOS 16 and watchOS 9 with AI features

September 12 Apple released software updates for iPhone and Apple Watch.

Demonstration of the function of cutting objects from a photo. Data: The Guardian.

New AI features in iOS 16 and watchOS 9 include:

  • focus filters to hide distracting content;
  • support unlocking your smartphone or hidden photo folders and logging into websites, apps and services with your face or fingerprint;
  • Face ID work in landscape orientation;
  • improved search in the Mail app, as well as reminders to attach a file or add a recipient;
  • Live Text for video text recognition and context processing;
  • AI function for cutting out objects from a photo;
  • functions for translating inscriptions using the camera and searching for text in a photo;
  • support for voice input of words and emoji;
  • a drug tracking system with label scanning using a camera and a smart watch.

Android is getting AI noise reduction for phone calls

Google has launched a test of the Clear Calling AI noise reduction feature for phone calls. The option is available in Android 13 beta.

The system reduces background noise during phone calls. According to Twitter user Mishaal Rahman, Clear Calling works on most mobile networks, but is not available when calling over Wi-Fi.

Company Meta will form from the development team of the PyTorch AI framework to a separate organization, the Pytorch Foundation.

The division will leave the direct control of the tech giant and become a subsidiary of the Linux Foundation consortium. The fund will be led by a separate board of representatives from Meta, AMD, AWS, Google Cloud, Microsoft Azure, and Nvidia.

The work of the Pytorch Foundation will be based on four principles:

  • openness;
  • supporting neutral branding;
  • honesty;
  • building a strong technical identity.

According to Meta CEO Mark Zuckerberg, the company will continue to use PyTorch as its main research platform and provide it with appropriate financial support. However, he added that the corporation would maintain “a clear separation between business and technical management” of the fund.

Intel will equip mobile processors with a neuroaccelerator

Intel will equip the upcoming 13th Gen Raptor Lake Mobile Core chips with a new VPU co-processor to accelerate AI tasks. The company plans to fully integrate the technology into meteor lake.

According to the company, only leading OEMs will be able to use the VPU as part of select notebook models. The coprocessor capabilities will be made available to independent developers in preparation for the full integration of the technology into Meteor Lake.

The company did not disclose details about the VPU. However, Intel noted that the coprocessor will be able to offer a number of new features, such as improving the quality of AI effects as part of Windows. In 2023, the technology will be launched as a completely separate functional block as part of an SoC tile on Meteor Lake.

Probably Intel designed the VPU to compete with the Neural Engine on Apple processors.

In the United States expressed concern about the popularity of TikTok

Heads of American companies, politicians and media expressed concern about the growing popularity of TikTok.

At the Code Conference, they noted that the video service could pose a threat to the country’s national security. They fear the power of the social network and its potential for surveillance by China’s ByteDance.

According to Snap CEO Evan Spiegel, this growth is due to the scale of investment in TikTok. ByteDance has invested billions of dollars in user acquisition that no other startup could afford, he added.

Spiegel noted that the social network’s recommendation algorithm has become strong thanks to investments and a huge user base.

Google CEO Sundar Pichai also called TikTok one of YouTube’s biggest competitors.

Axel Springer CEO Matthias Dopfner called for the platform to be banned.

Google AI introduced the PaLI transformer neural network

Google’s AI division presented unified model

text-image PaLI with 17 billion parameters. The algorithm is capable of doing

many tasks in more than 100 languages.

Schematic representation of the PaLI algorithm. Data: Google AI.

To train the transformer neural network, the researchers developed the WebLI dataset. It includes 10 billion images and 12 billion of their annotations. The scientists also used the Cloud Vision API for text recognition in pictures to expand the dataset.

Google AI said that PaLI can create image captions, provide visual answers to questions, classify pictures, detect objects, recognize characters, parse text, and perform a number of other similar tasks.

Scientists have developed an AI fitness mirror

Researchers from the University of Brescia have developed a smart mirror that can improve the efficiency of fitness training.

The computer vision system includes a deep learning algorithm for video gesture recognition, an Nvidia Jetson Nano single board computer, and two fisheye cameras. The model evaluates the movements of the human body, detects posture errors and analyzes fitness exercises.

So far, the system can only detect simple “variables” like the angle of the arms or their position, the scientists say. In the future, the algorithm will be able to evaluate more complex movements, they added.

AI Dungeon has added an image generator to illustrate stories

Latitude Company expanded text-based AI Dungeon role-playing game with an image generator.

The developers have added the Stable Diffusion neural network so that gamers can create illustrations for their stories. According to them, the algorithm generates images quickly and cheaply, which allows access to the function to every player.

Illustration for the story in AI Dungeon. Data: Latitude.

Latitude said they chose Stable Diffusion due to the lack of restrictions in creating pictures. The algorithm can generate artwork, architectural concepts, pornography, and celebrity deepfakes.

Twitter Suspends User Account Who Changed The Little Mermaid’s Skin Color

Company Twitter blocked The account of the user who, with the help of AI, “transformed” the black actress from the movie The Little Mermaid.
Comparison of the original and the work of the neural network. Data: vandalibm.

A user under the nickname vandalibm posted a video where he whitewashed the heroine with neural networks, changed her facial features and dyed her hair red. At the same time, the author did not change the voice of Ariel from the film.

According to vandalibm, he and a friend created the video for educational purposes only and should not be construed as racist.

Twitter representatives have not yet commented on the blocking of the account.

The most important AI deals of the last week

From September 11 to September 17, 2022, AI startups raised over $107.6 million. Here are the most interesting deals.

  • IoT startup Litmus Automation attracted $30 million
  • Diveplane collected $25 million for the development of an MLOps platform for enterprises.
  • Nyshex received $25 million to expand the service that connects shippers with sea carriers.
  • Winn AI attracted $17 million to support an AI assistant for sales teams.
  • Synthetic Data Creation Platform Developers collected $10.6 million

Also on Cryplogger:

Subscribe to Cryplogger news in Telegram: Cryplogger AI – all the news from the world of AI!

Found a mistake in the text? Select it and press CTRL+ENTER

Exit mobile version