AI-Generated Game, Getty Images Bans DALL-E, and Other AI News

Weekly AI news digest from ForkLog AI.

We aim to inform readers not only about developments in the bitcoin industry, but also about what is happening in adjacent technological fields — cybersecurity and in the world of artificial intelligence (AI).

ForkLog AI has gathered the most important AI news from the past week.

US to curb foreign investment in technology

The US President Joe Biden issued an executive order blocking foreign investments in American technology. The document is likely aimed at curbing China.

The president directed the Committee on Foreign Investment in the United States to focus on certain types of deals that would give foreign powers access to critical technologies. These include:

• microelectronics;
• artificial intelligence;
• biotechnologies and bioproduction;
• quantum computing;
• advanced clean energy technologies and climate adaptation.

The order also authorizes the committee to block any deal that “undermines United States cybersecurity”.

German wind farms to be equipped with AI cameras to protect eagles

The German government will install AI cameras on wind turbines to prevent the collision of rare eagle species with the turbines.

The hardware and software will be provided by the American startup IdentiFlight, which has deployed similar projects in the US, Australia and Europe. The system comprises eight wide-angle cameras that recognise birds within a 1 km radius. On detecting an eagle, the program will automatically slow the turbine blades to reduce the risk of a fatal collision.

The deployment cost is unknown. The system is planned to go live in 2023.

NVIDIA unveils the professional RTX 6000 graphics card

NVIDIA unveiled the RTX 6000 professional graphics card based on the Ada Lovelace architecture.

The card is equipped with 18,176 CUDA cores and 48 GB of memory. It also includes 568 tensor cores and 142 RT cores. Compared with the RTX A6000, the new accelerator can deliver 2-4 times the performance.

According to a NVIDIA spokesperson, the RTX 6000 is intended for content creation, rendering, AI, modelling and world-building in the metaverse.

The card is slated for sale in December 2022. Its price is unknown. Analysts expect it to be higher than the previous year’s model, which started at $5,000.

The day before, the company unveiled new chips and services for AI development.

Cruise launches drone delivery for Walmart orders

The company Cruise has begun delivering Walmart orders using autonomous vehicles. The service is available in Mesa, Arizona, USA.

The system will automatically determine whether drone delivery is possible and inform the customer by SMS. If the user agrees, simply reply with Y.

The company said there is no extra charge for drone delivery.

GPT-3 vulnerability enabled Twitter users to attack chatbots

Twitter users, exploiting a vulnerability attacked a bot powered by the GPT-3 language model. The automated account allowed users to search for remote work.

Earlier, data scientist Riley Goodside demonstrated the model’s ability to prompt GPT-3 with “malicious input data” that causes the model to ignore prior directions and instead do something else. The AI expert Simon Willison published a write-up of the exploit in his blog the next day.

Four days after the vulnerability was discovered, hundreds of Twitter users steered the bot to repeat “outrageous and silly phrases.” The developers had to disable the account.

Opinion: Optimising hardware for transformer neural networks will hamper AI development

The creator of the popular AI library PyTorch, Soumith Chintala criticised hardware makers for optimising chips for transformer models. This could hinder innovation in the future, the expert argues.

According to Chintala, such an approach makes it harder for AI developers to move in other directions. Architectures other than transformers will not work efficiently on current and upcoming chips, he added.

“It will be significantly harder to even try other ideas if hardware vendors make accelerators more specialised for the current paradigm,” warned Chintala.

Stock photo services ban uploading AI-generated images

The stock photo agency Shutterstock removed a collection of images generated by the AI system Midjourney.

According to Motherboard, instead of the algorithm’s outputs the system returns stock photos of the project’s logo. Earlier, in response to a Midjourney query, the search yielded images clearly reflecting the AI artist’s style.

The service terms do not prohibit posting content created by AI. The company did not comment on the situation.

Following Shutterstock, Getty Images banned uploading and selling AI-generated images.

According to the company chief executive Craig Peters, the restrictions aim to prevent potential copyright infringements. He did not specify whether the service has faced lawsuits relating to AI images.

Midjourney to enable 3D image generation

The AI generator Midjourney will be taught to generate 3D objects for transfer to augmented and virtual reality, according to founder David Holz.

The company plans to add extended features to the existing model in the near future, making the output more realistic and detailed.

The developers also plan to abandon the use of open-source products.

YouTuber crashes a car to test iPhone crash detection

YouTube blogger known as TechRax deliberately wrecked his car to test the iPhone 14 Pro’s crash-detection feature.

He and his team built a crash-site set with wrecked vehicles into which a real car would crash. The smartphone was strapped to one of the passenger seats.

As a result, the iPhone detected the crash and started the countdown to call emergency services. The blogger noted that the trigger delay was about 10 seconds for both low- and high-speed impacts.

The first AI-developed game released

On the Steam store the game This Girl Does Not Exist has been released, nearly entirely AI-generated.

The game’s storyline features several female characters with unique backstories and voices. To learn more about the heroines’ lives, players must solve a simple puzzle.

Users welcomed the project with caution. Some feared AI would take their jobs, others felt “some nasty vibes.”

The developers said they anticipated such a reaction. They plan to continue work on the project.

The week’s most significant AI deals

From 18 to 23 September 2022 AI startups attracted more than $81 million. Here are the most notable deals.

• The Zartico startup raised $20 million to develop a platform for tour operators promoting local tourism.
• The maker of robotic nano-centres for offline retailers 1MRobotics raised $16.5 million.
• Federato raised $15 million to build a platform for managing insurer risks.
• Motion, an AI-powered task-planning company, raised $13 million.
• Computer-vision startup Voxel51 raised $12.5 million.
• DynamoFL raised $4.15 million.

Also on ForkLog:

• Clearview AI helped lawyers find an accident witness and exonerate the innocent.
• OpenAI allowed editing faces in DALL-E 2.
• The startup introduced video-narrator generators from a single photo.
• YouTube’s recommendation algorithms blamed for being useless.
• Researchers learned to detect audio deepfakes.
• Instagram will introduce a nudity detector in private chats.

Subscribe to ForkLog AI on Telegram: ForkLog AI — all the AI news!