IT News Review by Control F5 Software: Amazon set to test humanoid delivery robots
Amazon is preparing to test humanoid delivery robots
Amazon is developing software for humanoid robots that could, in the future, take over the jobs of delivery couriers, according to a report by The Information, which cites a person familiar with the project.
The company is currently finalizing the construction of a "humanoid robot park," an obstacle course set in one of Amazon's offices in San Francisco, California, where the robots are to be tested, the article added.
According to the source, Amazon is also developing the artificial intelligence-based software that will power these robots, noting that the company plans to use hardware from other firms for its tests for now.
The budget for AI-powered search ads could reach $26 billion by 2029
Spending on AI-powered search engine ads is on track to rise dramatically in the US, from just over $1 billion this year to nearly $26 billion by 2029, according to data published by Emarketer.
This growth is being fueled by the rapid adoption of technology and increasingly refined methods of targeting users.
Companies that still rely on traditional keyword-based ads could see revenue declines as AI-powered ads offer users a more convenient and engaging experience, according to the research firm.
EU 'trusted partner' for digital cooperation amid global tensions, says EU official
The European Union will work with international partners to strengthen its competitiveness and promote a rules-based global digital order, EU technology chief Henna Virkkunen said on Thursday.
The move aims to counter growing tensions in the relationship with the United States.
In recent months, the US administration has repeatedly criticized the European Commission over new regulations imposed on tech giants and threatened tariffs in response to fines against US companies.
Meta could invest more than $10 billion in Scale AI
Meta is in talks for a multibillion-dollar investment in Scale AI, according to Bloomberg. The deal could be worth more than $10 billion, which would make it the largest outside investment by parent company Facebook and one of the largest private financings ever.
Scale AI provides data labeling services for companies such as Microsoft and OpenAI to help them train their artificial intelligence models. Much of this work is done by collaborators, and the US Department of Labor recently closed an investigation into possible misclassification and underlabeled practices.
According to Bloomberg, the company had revenue of $870 million last year and estimates it will reach $2 billion this year.
Lawyers risk severe penalties for false citations generated using AI, warns UK court
The High Court of England and Wales warns that lawyers must take far more rigorous measures to prevent the misuse of artificial intelligence in their work.
In a judgment bringing together two recent cases, Judge Victoria Sharp wrote that generative AI tools, such as ChatGPT, "are not capable of carrying out reliable legal research.""These tools may provide seemingly coherent and plausible answers to queries, but those answers may be completely incorrect," the judge said. "They can make confident assertions that are simply false."
2025 will be a "watershed year" for augmented and virtual reality at Meta, says CTO
Andrew "Boz" Bosworth, Meta's CTO and one of the company's top 15 engineers, published a memo earlier this year saying that 2025 could be the year that Reality Labs (Meta's AR and VR division) reaches a peak of success. Or, conversely, it could be the year Metaverse becomes a "failed venture of legendary proportions."At present, Boz seems to be leaning toward the optimistic view, but says the market will decide the final verdict.
"We'll judge at the end of the decade, but I feel like this is the decisive year," he said Thursday in an interview with Bloomberg Tech.OpenAI is storing conversations deleted from ChatGPT because of the New York Times lawsuit
OpenAI said it is obligated to preserve "indefinitely" deleted ChatGPT conversations following a court ruling in a copyright lawsuit filed by The New York Times.
In a post published Thursday, chief operating officer Brad Lightcap said OpenAI is appealing the ruling, which he called an "abuse" that "abandons well-established privacy rules and weakens the protection of personal data."
Last month, a court ordered OpenAI to retain "all output log data that would otherwise have been deleted," even if a user requests it or if privacy laws require deletion. Until now, OpenAI's policy was to retain conversations for 30 days, then delete them permanently. That policy is now suspended pending a new court decision.
Anthropic launches new Claude service aimed at US military and intelligence services
Anthropic on Thursday announced Claude Gov, a product specifically designed for US defense and intelligence agencies. AI models dedicated to this service have more flexible user rules and are trained to more efficiently analyze classified information.
The company said that these models are already in use by agencies at the highest level in US national security and that access will be limited exclusively to government entities that handle classified information. Anthropic did not disclose how long the models have been in use.
Meta would negotiate exclusive content for its new VR headset
According to the Wall Street Journal, Meta is in talks with Disney and A24 to get exclusive video content for a new virtual reality device in development.
The newspaper notes that the device, codenamed "Loma," would resemble a pair of VR goggles and work in conjunction with an external, card-sized module that can be kept in your pocket.
This description is similar to details about the "Puffin" headset in development, which could be the next generation after the Quest 3 and 3S. The Puffin is described as an extremely lightweight device with an open field of view and external processing unit.
Reddit is suing Anthropic, accusing it of accessing its platform more than 100,000 times
Reddit on Wednesday filed a lawsuit against Anthropic in a San Francisco court, alleging that rival OpenAI has accessed its platform more than 100,000 times since July 2024, despite its claim that it had locked down its bots.
In the complaint filed, Reddit describes Anthropic as a "late-blooming AI company that presents itself as the white knight of the AI industry," but which, they claim, "is far from it."
US-UAE deal for billion-dollar AI campus not close to completion
A billion-dollar deal to build one of the world's largest data centers in the United Arab Emirates using American technology is far from being finalized because of lingering security concerns, according to sources cited by Reuters.
The US and the Gulf state unveiled the project during a two-day visit by former President Donald Trump to Abu Dhabi last month.
Sam Altman brings his eye-scanning identity verification startup to the UK
Worldcoin, the biometric identity verification project co-founded by OpenAI CEO Sam Altman, will launch in the UK this week.
The project uses a spherical device called the Orb that scans people's eyes. It will be available in London from Thursday, with cities such as Manchester, Birmingham, Cardiff, Belfast and Glasgow to follow in the coming months.
The aim is to authenticate human identity and prevent technological abuses, such as deepfakes, by scanning a person's face and iris, generating a unique code that confirms they are a human being, not an AI entity.
ChatGPT gets smarter: now connects to Google Drive and Dropbox
OpenAI has announced important new updates for ChatGPT users: the AI assistant can now connect directly to cloud storage platforms such as Google Drive, Dropbox, Box, SharePoint and OneDrive. This means users can ask questions related to their documents and files stored in these services - for example, "What was my income in Q1 last year?" or "How many times did I take the ferry on vacation in Italy?" - and they will receive structured, clear answers backed up with sources from the files.
In addition, OpenAI has introduced a new feature called "recording mode" that allows users to capture notes from meetings, brainstorming sessions or even thoughts spoken aloud. These recordings are automatically transcribed with time quotes and action suggestions, turning spontaneous thoughts into structured ideas.
DeviantArt founder launches Layer - a $22,000 luxury display for generative digital art
Angelo Sotira, founder of DeviantArt, the platform that helped shape one of the earliest online creative communities, returns to the forefront with a project set to once again revolutionize the way we experience digital art - this time, in the physical space.
On Thursday, Sotira unveiled Layer, a deluxe "digital canvas" designed specifically to showcase generative art at its highest level of quality. Created to blend harmoniously into elegant spaces, Layer is no ordinary display.
"A canvas should behave differently than a simple display - enriching, not disrupting its environment," Sotira told TechCrunch.
Oversight Council criticizes Meta for inaction in the face of celebrity deepfake scams
Meta is once again in the spotlight - this time criticized by its own Supervisory Board itself - for not doing enough to stop scams based on deepfake celebrity videos distributed on its platforms.
In a new decision, the council accuses Meta of allowing this type of content to proliferate and failing to enforce its own rules. "Meta likely tolerates significant amounts of fraudulent content in order to avoid over-checking real celebrity posts," the decision says.
According to the council, content reviewers working at scale do not have the tools and authority to effectively block posts that imitate celebrities for fraudulent purposes.Google's Gemini assistant now allows task scheduling
Google is enhancing Gemini's capabilities, positioning it as an even more useful AI assistant. The company has begun rolling out a new feature called "scheduled actions," available to subscribers of its AI Pro and AI Ultra plans.
With this feature, users can ask Gemini to perform tasks at set times - for example, deliver a daily calendar summary at the end of the day or generate blog ideas every Monday. The function is also available for one-off actions, such as summarizing an event (such as an award or conference) the day after it has taken place.
"Just tell it what you want and when, and Gemini takes care of the rest," Google explained in an official announcement.
EleutherAI launches one of the largest AI datasets licensed for training
AI AI research organization EleutherAI has released a huge new dataset designed to train AI models using only licensed or public domain content. The dataset, called Common Pile v0.1, is an impressive 8 terabytes in size and is the result of nearly two years of work in collaboration with AI startups such as Poolside and Hugging Face, as well as academic institutions such as the University of Toronto.
Common Pile v0.1 was used to train two new 7 billion-parameter AI models - Comma v0.1-1T and Comma v0.1-2T - which EleutherAI says offer comparable performance to the top models that were trained on copyrighted data.
Through this endeavor, the researchers want to demonstrate that it is possible to build high-performing and ethical AI without infringing copyright.
Summary made with the help of a monitoring feed provided by Control F5 Software.ȘTIRI PE ACELEAȘI SUBIECTE
