We strive to inform readers not only about events in the bitcoin industry, but also to tell about what is happening in related technological areas – cybersecurity and in the world of artificial intelligence (AI).
ForkLog AI has collected the most important news from the world of AI over the past week.
- Since July 2021, 392 self-driving cars have been involved in accidents.
- Tesla and Samsung have entered into a multi-billion dollar deal to supply video cameras for unmanned driving.
- Adobe has developed a tool for restoring damaged photos in Photoshop.
- Facebook collects data on people seeking abortion-related services.
- An artificial algorithm is used to analyze images from the James Webb telescope.
- Researchers have taught the algorithm to create videos based on text descriptions.
- The most important transactions in the field of AI.
About 400 robo-cars have been involved in accidents in the past year
National Highway Traffic Safety Administration (NHTSA) has published a report on accidents involving robo-cars and cars with driver assistance systems.
According to the agency, from July 1, 2021 to May 15, 2022, 392 accidents are related to drones. In 108 cases another vehicle was involved, in another 11 cyclists and pedestrians. Of the 98 fatal crashes, 11 involved serious injuries.
273 collisions involved Tesla robots using Autopilot or Full Self-Driving. Five incidents ended in death.
Honda vehicles have been linked to 90 crashes, Waymo to 62, Transdev Alternative Services to 34, Cruise to 23, Subaru to 10. Drones from other automakers like Ford, General Motors, Volkswagen and Toyota have been involved in five or fewer accidents.
Tesla to buy robo-car cameras from Samsung
Tesla has partnered with Samsung to supply robo-car cameras.
According to Electrek, a tech giant, has won a $3.2-$4 billion tender. Samsung will supply 5-megapixel version 4.0 cameras that provide five times the sharpness of 3.0 modules.
Tesla will integrate video devices into all robotic vehicles over the next few years, according to the source.
Adobe introduced a tool for restoring old photos
Adobe developed a neurofilter to restore old damaged photos in Photoshop.
The system is powered by Sensei AI. The neurofilter cleans images of dirt spots, scratches, wrinkles and tears. The technology smoothes the skin of a person in a portrait, removing unwanted textures or noise and adding tones.
The filter can work in conjunction with other tools like Content-Aware Fill to restore missing parts of a photo.
The company also announced a free web-based photo editor with basic features.
Meta has created an algorithm for modeling 3D avatars from selfies
Meta will present a method for creating realistic 3D avatars using a smartphone at Siggraph 2022.
3D model of a face created with a smartphone. Data: Meta.
The approach allows you to digitize the head in four minutes, and then generate its three-dimensional model with a wide range of emotions. To do this, the user will need to “scan” the face with the smartphone's front camera, recreating 65 specific expressions.
According to scientists, the process of generating a 3D avatar on a high-performance computer will take about six hours. You can also use the cloud for rendering.
The method does not work with glasses and renders long hair poorly, the researchers added.
Study: Facebook tracks people seeking abortion services
Facebook advertising tools track people seeking abortion services. The system is supposed to filter out sensitive health data, but instead it collects it, Reveal said.
Researchers have found that hundreds of maternity crisis centers use the Meta Pixel analytics tool. The system allows organizations to track visits to sites to set up targeted advertising on Facebook.
According to the report, often pregnancy crisis centers are not licensed medical facilities and are not affiliated with HIPAA. They are mostly run by religious organizations that encourage people to stop having abortions.
Meta Pixel collects “extremely” sensitive information, including scheduled appointments, according to Reveal. Third-party anti-abortion marketing companies could also access the information collected by the tool even if their websites were not visited.
ByteDance to invest in VR content development
TikTok owner ByteDance plans to invest “a lot of money” in VR content development.
According to Protocol, most of the funding will go towards hiring staff for the company's VR headset maker Pico. Most of the job openings are in the content-focused department of Pico Studios.
Microsoft Teams Leverages AI to Improve Audio in Video Conferencing
Microsoft Leverages AI to Improve Call Quality in Teams Video Chats.
The company said the new algorithms are able to minimize echo and reduce the chance of interruptions by eliminating overlap associated with sound reflections. All processing is done locally on client devices, Microsoft added.
The developers created the technology using 100,000 simulated rooms and trained it on 30,000 hours of speech, according to The Verge. The tech giant also paid Teams users to record their voices to train AI to recognize thousands of different devices.
The feature comes along with other updates, including AI improvements to limit bandwidth during video calls and text rendering optimizations.
AI will study space images from the James Webb Space Telescope
Artificial intelligence will be used to study the first images from the James Webb Space Telescope (JWST) . Images will be published on July 12, 2022.
A team of 50 researchers will use the Morpheus AI system, trained to analyze pixels, find blurry droplets in space photos, and determine if these structures are galaxies and what type. The model will also help to study the development of dark matter and the beginning of star formation.
Morpheus has been trained on 7,600 images of galaxies taken by NASA's Hubble Space Telescope. However, the researchers intend to retrain the algorithm to work with JWST data.
Researchers taught a neural network to create a video from a text description
Researchers from Tsinghua University (China) have developed a CogVideo transformer with 9 billion parameters for converting text into video. The neural network only works with Chinese.
Result work transformer CogVideo. Data: research.
The scientists trained the algorithm by inheriting a pre-trained CogView2 word query image generator.
The model generates a 4-second, 480×480 pixel video at 8 frames per second.
New Voicemod filter lets you speak in Morgan Freeman's voice
Voicemod has used AI to use Morgan Freeman's voice in real time. The beta version of the Voicemod AI Voices converter also includes seven more new filters.
The developers brought in English-speaking professional voice actors with the necessary speech characteristics to create a training dataset. The sound engineers then processed the data using traditional methods such as adding background music and dynamic effects.
Voicemod AI Voices is available on PC. Filters can be used on Twitch streams or Discord calls.
Last week's most important AI deals
Since 12 As of June 18, 2022, AI startups have raised over $237 million. Here are the most interesting deals.
- Invoca received $83 million to develop software to analyze the work of call center operators on calls.
- Charm Therapeutics raises $50 million to develop AI platform for modeling complex protein interactions.
- Lightning AI raises $40 million to expand team of AI tools developers.
- Design platform developers building structures Alice Technologies raises $30 million
- Able raises $20 million to help banks and other lenders process customer data faster
- Ahana, a managed service developer based on the Presto query engine, raises $7.2 million
- Workstream raises $7 million to develop a platform to streamline enterprise analytics workflows.
Also on ForkLog:
- Google has suspended an engineer who claimed that AI has consciousness.
- The Chinese government used the anti-covid infrastructure to prevent protest.
- Experts questioned the “birth of intelligence” of the Google algorithm.
- Tech giants have joined the fight against disinformation in the EU.