IT News Review by Control F5 Software: Most people are afraid of the impact of AI, but they do not believe it will affect their jobs.
Most people fear the impact of AI, but do not believe it will affect their jobs
A new survey conducted by YouGov, in partnership with Udemy, which interviewed over 4,500 adults from the USA, UK, India, and Brazil, shows that most workers predict their jobs will be relatively unaffected by AI, even though they acknowledge that the technology will have major effects on society. In the UK, 70% of adults are concerned about the impact of AI on the economy, but only 39% are worried about the consequences for their own occupation. In the USA, the figures are 72% versus 47%.
This gap between general perception and personal impact reflects similar psychological trends from previous economic transitions, akin to the period of the 1990s when the internet was becoming popular. The YouGov/Udemy report also found that many workers feel they do not have the necessary AI skills for an AI-centered economy. In the UK, 55% of workers say they have not received any training in AI, and only 64% of British workers claim to be motivated to improve their career skills. Experts suggest that this hesitation is normal when it is unclear in which areas technology will be most transformative.
AI Data Centers and Renewable Energy
The world will spend $580 billion on data centers by 2025, $40 billion more than for discovering new oil sources, according to a report from the International Energy Agency. This major shift in the global economy raises questions about the impact of artificial intelligence on climate change. Data centers require huge amounts of energy and put pressure on already strained electrical grids.
However, there are also positive aspects, with solar energy being poised to power many of these new projects, which could create opportunities for startups developing innovative approaches to renewable energy. Companies like OpenAI have committed $1.4 trillion for building data centers, Meta $600 billion, and Anthropic has announced a $50 billion plan. Although there are ambitious plans, the question remains how many of these centers will actually be built, given the huge amounts of capital required.
Google fined €572 million for antitrust violations
A German court has ruled that Google abused its dominant position in the price comparison market and must pay a total of €572 million ($665.6 million) in damages to German price comparison companies Idealo and Producto. Idealo will receive approximately €465 million, while Producto will get €107 million. The decision comes after the European Court of Justice ruled in 2024 that Google favors its own price comparison service for online shopping, violating competition rules, and fined it approximately $2.7 billion.
Google announced that it intends to appeal both rulings, arguing that the changes made in 2017 are working well and that the number of price comparison sites using Shopping Unit has increased from seven to 1,550 currently. The ruling comes in the context of a recent EU investigation into how Google's anti-spam policy affects the search rankings of publishers and after the company was fined €2.95 billion for violating antitrust rules in advertising.
Police dismantles three cybercrime operations
An international coalition of law enforcement agencies, coordinated by Europol, has targeted and dismantled three cybercrime operations in the latest round of "Operation Endgame." The operation targeted the infostealer malware Rhadamanthys, a botnet named Elysium, and the remote access trojan VenomRAT, all playing key roles in international cybercrime. Police confiscated over 1,000 servers as part of the operation and arrested the main suspect behind VenomRAT in Greece.
The dismantled malware infrastructure consisted of hundreds of thousands of infected computers containing millions of stolen credentials. Rhadamanthys saw a dramatic increase after authorities dismantled the popular infostealer Lumma earlier this year, demonstrating that after such actions, criminals quickly adapt by using different tools. Experts recognize that this fight is a permanent game of cat and mouse, where other attackers take the place of those dismantled.
ChatGPT will follow instructions not to use long dashes
Sam Altman revealed that if you tell ChatGPT in custom instructions not to use long dashes, the chatbot will listen in the end. Previously, ChatGPT ignored this instruction and continued to use long dashes even when explicitly asked not to do so. The presence of long dashes, especially when used abundantly, is considered one of the most obvious signs that a text was written by language models.
It is unclear why generative AI models tend to pepper the generated text with long dashes. LLMs are trained on a vast number of books and online content, such as scientific papers, public forum posts, and articles. It is possible that this punctuation mark appears so frequently in training materials and is not flagged by AI trainers as something the model should avoid using.
Disney+ will allow AI-generated content
Disney CEO Bob Iger sparked negative reactions online after announcing, at the Q4 2025 earnings conference, that the company will allow the appearance of AI-generated content on Disney+. Iger stated that Disney is in the midst of launching the largest and most significant changes from a product and technology perspective since the service's launch in 2019. Creatives have been vocal about the existential threat that AI poses to creative fields.
The changes will include games or at least game-like features on Disney+ through the agreement with Epic Games, as well as short AI-generated video content by users in the future. Iger mentioned that Disney has had productive conversations with unnamed AI companies, hoping to reach an agreement that reflects the need to protect intellectual property. Disney added 3.8 million subscribers to Disney+ and announced plans to invest $1 billion more in content in 2026 compared to 2025.
Google introduces a dedicated tab for inspirational images
Google is adding a new feature to its mobile app, which could compete with Pinterest to become the users' first choice for inspirational images: a dedicated tab for images. The company announced that it is introducing a new "Images" icon at the bottom of the Google Search app for iOS and Android devices, which will display personalized images according to the user's interests. The images will update daily, providing users with a way to organize their creative ideas or find visual inspiration for things like planning a trip, organizing a party, decorating an apartment, or discovering new fashion.
These use cases are, of course, among the main reasons users turn to Pinterest and other social apps like Instagram. Users can browse images in the new tab, search for other images, and save images they like in their collections. With this addition, Google is likely hoping to increase user engagement with its app, especially considering that AI companies are eroding its search and advertising business. More importantly, the functionality gives Google a new surface within its popular Search app, where it could display ads.
Disney and YouTube TV reach an agreement to end blackout
After a two-week blackout, YouTube TV and Disney announced on Friday that they have reached an agreement. Along with restoring Disney networks like ABC, ESPN, and FX on YouTube's TV streaming service, the agreement will make the new direct-to-consumer ESPN service available on YouTube TV at no additional cost. YouTube will also be able to sell certain Disney networks and the Disney+/Hulu package as part of various bundles.
YouTube TV is designed to be an alternative to cable television for those who cut the cord, but like most cable TV providers, the service has had its share of disputes with major content providers, usually related to the amounts they have to pay and the level of flexibility in offering content in different packages. The last time YouTube and Disney renegotiated, in 2022, a similar blackout occurred, but it was significantly shorter. This time, YouTube offered customers a $20 credit that can be applied to their next bill. A survey suggested that 24% of the more than 10 million subscribers said they had already canceled their subscription or planned to do so soon.
Documents shed light on how much OpenAI pays Microsoft
Documents obtained by tech blogger Ed Zitron provide a glimpse into OpenAI's finances, particularly regarding revenues and computing costs in recent years. Zitron reported that in 2024, Microsoft received $493.8 million in payments from revenues distributed by OpenAI. In the first three quarters of 2025, this figure jumped to $865.8 million. OpenAI distributes 20% of its revenues to Microsoft as part of a previous agreement, in which the software giant invested over $13 billion in the AI startup.
The situation is, however, more complicated, as Microsoft in turn distributes revenues to OpenAI, providing about 20% of revenues from Bing and Azure OpenAI Service. The disclosed payments refer to Microsoft's net revenue share, not the gross share, meaning they do not include amounts paid by Microsoft to OpenAI from Bing and Azure OpenAI royalties. Based on the 20% revenue distribution, reported on a large scale, it can be inferred that OpenAI's revenues were at least $2.5 billion in 2024 and $4.33 billion in the first three quarters of 2025. OpenAI would have spent approximately $3.8 billion on inference in 2024, with the figure rising to about $8.65 billion in the first nine months of 2025, suggesting that OpenAI may spend more on inference costs than it generates in revenue.
DOJ issues warrant for satellite systems linked to scam complexes
The U.S. Department of Justice has issued a seizure warrant to Starlink for satellite internet systems used in scam complexes. As these complexes ramp up their activities of stealing funds from victims worldwide, U.S. law enforcement has continued to target their operations. Cybercriminals running scam complexes in Myanmar and other parts of Southeast Asia are now using Starlink satellite internet terminals for connectivity, allowing them to reach victims worldwide.
In a warrant issued this week, U.S. Magistrate Judge G. Michael Harvey authorized the seizure of nine Starlink terminals and two Starlink accounts used in scam operations in Payathonzu, near the Myanmar-Thailand border. The intervention force has already begun the operation, with the DOJ claiming the seizure of over $400 million in stolen digital assets from victims of these schemes. U.S. officials claim that at least 79 Starlink antennas appear on the rooftops of buildings in the Tai Chang complex in Myanmar, with activities in the building linked to the Karen Benevolent Democratic Army, an armed group in Myanmar that was sanctioned by the U.S. government this week.
Samsung raises memory chip prices by up to 60%
Samsung Electronics has raised prices for certain memory chips this month, now in deficit due to the global race to build AI data centers, by up to 60% compared to September, two people familiar with these increases said. The move follows the decision of the world's largest memory chip manufacturer to delay a formal announcement of prices for supply contracts in October. Rising prices for these memory chips, primarily used in servers, will likely add pressure on large companies building data infrastructure.
Samsung's contractual prices for 32-gigabyte DDR5 memory chip modules have risen to $239 in November, up from $149 in September. Samsung has also raised prices for 16 GB and 128 GB DDR5 chips by about 50%, to $135 and $1,194, respectively. Prices for 64 GB and 96 GB DDR5 chips have increased by more than 30%. The chip crisis has been so severe that it has triggered panic buying from some customers. The deficit is, however, an advantage for Samsung, which has fallen behind rivals in offering advanced AI chips and has not seen its profits rise nearly as much until recently.
Apple must pay Masimo $634 million in smartwatch patent case
A federal jury in California ruled on Friday that Apple owes medical monitoring technology company Masimo $634 million for infringing a patent related to blood oxygen reading technology. The jury agreed with Masimo that the training mode of the Apple Watch and heart rate notification features infringed Masimo's patent rights. An Apple spokesperson stated that the company disagrees with the verdict and will appeal, noting that in the last six years, Masimo has sued Apple in multiple courts and invoked over 25 patents, most of which are considered invalid.
The California trial is a branch of a controversial, multi-front battle between Apple and Masimo, which has accused Apple of hiring its employees and stealing pulse oximetry technology to use in the Apple Watch. The dispute led a U.S. trade court to block imports of the Apple Watch Series 9 and Ultra 2 in 2023 after finding that Apple's technology infringed Masimo's patents. Apple removed the blood oxygen reading technology from its watches to avoid the ban and reintroduced an updated version of the technology in August, with the approval of the U.S. Customs and Border Protection. A California judge declared the trade secrets lawsuit filed by Masimo against Apple in 2023 null and void.
Google offers EU concessions for adtech
Google has presented a proposal to European Union regulators, offering changes to its advertising technology operations in an effort to resolve a major antitrust case without selling off part of the business. The company stated that its plan would make it easier for publishers and advertising agencies to use competing online advertising tools, addressing the European Commission's concerns about conflicts of interest in Google's adtech stack. Regulators fined the company €2.95 billion ($3.4 billion) in September for favoring its own display technology and reinforcing its dominance in the AdX exchange market.
In a blog post published on Friday, Google announced immediate product updates, such as allowing publishers to set different minimum prices for various bidders in Google Ad Manager. The company also plans to increase the interoperability of its tools to provide users with more flexibility and options. The European Commission has urged Google to consider divesting parts of its adtech operations, a step the company has called disruptive and potentially harmful to the thousands of publishers that rely on its systems.
AI incidents monitored by OECD
The OECD is monitoring AI incidents and dangers to help governments manage risks. The AI Incident and Hazard Monitor (AIM) has been initiated and is being developed by the OECD.AI expert group on AI incidents, with the support of the Patrick J. McGovern Foundation. The AIM's goal is to track, in real-time, actual AI incidents and hazards as defined by the OECD, and to provide the evidence base to inform related AI policy discussions. AIM detects AI incidents and hazards from reputable international news sources, monitoring over 150,000 news articles daily.
Between 2022 and 2023, the number of reported AI incidents increased by approximately 1,278%, coinciding with the mainstreaming of generative AI. The OECD has developed a reporting framework for incidents that includes clear definitions. An AI incident is defined as an event in which the development, use, or malfunction of an AI system directly or indirectly causes harm such as injuries, disruption of critical infrastructure, violations of human rights, or material damage. These definitions are essential for managing and preventing these risks.
European mobile operators seek to gain key spectrum in the fight against the Wi-Fi industry
Telecommunications operators in Europe are seeking to gain access to most of a key spectrum, also coveted by internet providers, with the support of an advisory body of the European Commission. The issue pits Deutsche Telekom, Orange, TIM, Vodafone, and other telecom companies against tech giants like Amazon, Apple, Meta Platforms, Microsoft, and others. The former want more spectrum for mobile services, warning that Europe could fall behind the U.S. in implementing future 6G if it does not gain access to the upper 6 GHz band, one of the few remaining large blocks of mid-band spectrum available.
The Radio Spectrum Policy Group, which met in Brussels, is expected to issue its opinion soon, proposing that 540 MHz of the 6 GHz band go to mobile operators. The remaining 160 MHz, which could return to Wi-Fi companies, will be "frozen" until the World Radiocommunication Conference 2027, organized by the International Telecommunication Union (ITU), which will provide a roadmap for spectrum bands intended for future mobile technologies. Experts expect 6G to be commercially launched in the 2030s. The United States, Canada, South Korea, and a few other countries have already allocated portions of the 6 GHz band for Wi-Fi use.
Cisco raises annual forecast as AI boom drives demand for networking equipment
Cisco Systems raised its profit and revenue forecast for the entire year on Wednesday, betting on robust demand for its networking equipment driven by multi-billion dollar data center expansions in the context of the AI boom. The company's shares, a supplier for cloud, enterprises, and communication service providers, rose by more than 7% in extended trading. Companies like Cisco have benefited from the fact that businesses are rapidly migrating workloads to the cloud and investing billions in upgrading campus infrastructure in response to the growing adoption of AI models.
Cisco expects $3 billion in revenue from AI infrastructure from hyperscalers in fiscal 2026, CEO Chuck Robbins said in a post-earnings call. The company has already secured AI orders of over $2 billion, almost all from hyperscalers, in fiscal 2025. Cisco, based in San Jose, California, expects revenues to be between $56 billion and $56.5 billion for fiscal 2025, compared to a previous forecast of $55.3 billion to $56.3 billion. Top tech companies like Alphabet, Microsoft, Meta, and Amazon are increasing capital expenditures, investing heavily in data centers and advanced chips.
WhatsApp to launch interoperability chats in Europe soon
Meta is preparing to launch interoperability chat integration on WhatsApp in Europe, the company announced on Friday. The tech giant notes that the EU's Digital Markets Act (DMA) requires it to provide users in Europe with the option to talk to people using interoperable messaging services that have chosen to do so. Following successful small-scale tests in recent months, the option for WhatsApp users to talk to users of the BirdyChat and Haiket messaging apps directly through interoperability chats will soon be launched across Europe.
Users in Europe who opt for interoperability will be able to share messages, images, voice messages, videos, and files. The option to create groups with foreign users will be launched once Meta's partners are ready to accept it. In the coming months, users in the European Region will start to see a notification in the "Settings" tab, explaining the opt-in process for connecting with people on foreign apps. Interoperability will only work on Android and iOS, not on desktop, web, or tablets. Of course, connecting with people on other messaging apps is optional, and users can enable or disable interoperability chats at any time.
TikTok users can now share tracks from Amazon Music
TikTok is launching a new feature that will allow users to share content from Amazon Music to the social network. With a new "Share to TikTok" button, Amazon Music users can share songs, albums, playlists, and important listening moments from Insights on Amazon Music directly to TikTok. Amazon Music Insights is a new personalized feature that provides listeners with a detailed look at their music habits and statistics, including "Monthly Recaps" that highlight artists, songs, podcasts, and total listening time.
TikTok already supports "Share to TikTok" functionality for both Spotify and Apple Music. The update builds on the existing partnership with Amazon Music, through the "Add to Music App" functionality, which allows users to save songs they discover on TikTok directly in Amazon Music and other top streaming services. Thursday's announcement makes sense for TikTok, an app known for influencing songs that reach the charts. By adding Amazon Music to its "Share to TikTok" options, TikTok is enhancing its ability to connect users with their favorite tracks and artists.
Google's SIMA 2 agent uses Gemini to reason and act in virtual worlds
Google DeepMind unveiled a research preview of SIMA 2, the next generation of its generalist AI agent, which integrates the language and reasoning capabilities of the Gemini model to go beyond simply following instructions and to understand and interact with its environment. SIMA 2 is powered by the Gemini 2.5 Flash-Lite model and represents a major shift from SIMA 1, which was revealed in March 2024 and had a success rate of only 31% in completing complex tasks, compared to 71% for humans.
SIMA 2 has doubled the performance of its predecessor, combining the advanced language and reasoning capabilities of Gemini with embodied skills developed through training. The agent can describe its surroundings, determine the next steps by recognizing and interacting with objects, and uses Gemini for internal reasoning. For example, when asked to go to the house that is the color of a ripe tomato, the agent explains its thought process and finds the red house. Being powered by Gemini also means that SIMA 2 can follow emoji-based instructions. SIMA 2 can navigate photorealistic worlds newly generated by Genie, DeepMind's world model, correctly identifying and interacting with objects like benches, trees, and butterflies.
Google expands AI shopping with conversational search and a feature that calls stores
Google is launching a suite of AI updates for shopping just before the holiday season. The company revealed on Thursday a series of new tools and functionalities, including conversational shopping in Google Search, new shopping features in the Gemini app, agentic checkout, and even an AI tool that can call local stores to find out if a desired product is available. An update will allow consumers to ask shopping questions in AI Mode, Google's conversational search functionality, which allows natural language queries in a chatbot-like interface.
Google is launching agentic checkout in Google Search in the U.S., including in AI Mode, which is compatible with merchants like Wayfair, Chewy, Quince, and certain Shopify stores. To use agentic checkout, you can start by tracking the price of an item to be notified if the price drops. You can then opt for Google to purchase the item for you on the merchant's site using Google Pay. Another AI functionality can call local businesses in your name to find out if a store has a specific product, how much it costs, and if there are any promotions. The functionality is built on Google's Duplex technology, introduced in 2018, as well as on Shopping Graph and payment infrastructure. The AI will call local stores and ask questions about the product, then return with a summary of its findings.
ElevenLabs signs deals with celebrities to create AI audio
ElevenLabs has signed an agreement with actors Michael Caine and Matthew McConaughey to generate, with the help of AI, versions of their voices. The company also announced this week that it is launching a marketplace that will allow brands to use licensed AI-generated voices of celebrities, which will include Caine and other names like Liza Minnelli and Dr. Maya Angelou. ElevenLabs is one of the most popular AI unicorn companies and has supporters like a16z and ICONIQ. McConaughey, an investor in the company, will use the AI voice to turn his newsletter into an audio edition in Spanish.
This move prepares the company for rapid growth in generative audio and for normalizing a practice viewed with caution in Hollywood since AI became a point of conflict during last year's strikes. The marketplace will also feature voices like Caine's and licensed voices of deceased legends, such as Dr. Maya Angelou, with opt-in voices from public figures who have ceded their rights for commercial use. ElevenLabs' Iconic Voice marketplace operates as a bilateral platform, allowing companies to request access to talent for campaigns, partnerships, and creative initiatives. Through this marketplace, ElevenLabs connects interested companies with relevant brand owners or intellectual property rights holders.
iOS 26.2 beta adds Liquid Glass customization and alarms in the Reminders app
Apple released iOS 26.2 beta 1 last week, introducing several updates, such as the option to customize the Liquid Glass appearance of the clock on the iPhone's lock screen, the ability to set alarms in the Reminders app, and real-time translation features for users in the EU. The public release of iOS 26.2 is expected in December. The Liquid Glass interface, the updated design for Apple devices, is still in the process of refinement. The recent update includes a slider that allows users to adjust the opacity of the clock displayed on the Lock Screen, enabling them to make the clock nearly completely transparent or give it a more matte look.
Another change is the ability to set alarms for reminders. When creating a reminder, choosing a specific time and activating the "Urgent" option will trigger an alarm at that time. Additionally, Apple revealed that as part of the iOS 26.2 launch, the real-time translation functionality via AirPods is expanding into the EU. If someone speaks to you in another language, Siri can provide translations through AirPods. Furthermore, the Apple News app will include a new "Following" tab in the bottom navigation bar, which will bring together favorites, saved stories, and history.
Mac Pro is no longer important for Apple
A new report claims that Apple has "largely abandoned the Mac Pro," but before you get upset, it is important to understand why tower Mac desktops no longer make sense and haven't for years. It all comes down to the architecture of Apple's M-series processors. The Mac Pro series is a line of next-generation modular desktop workstations built for professionals who need extreme performance for demanding workflows. These computers are commonly used by film editors, 3D animators, VFX artists, music producers, and software developers.
The problem is that Apple can no longer build a Mac with an extremely expandable internal architecture. The latest report shows that Apple will not launch an updated Mac Pro before the end of 2025, as was expected, nor in 2026. Additionally, it seems that Apple has lost interest in this line. The Mac Studio has become the coveted Mac, being, according to many, the best Mac ever created. In comparison, the Mac Pro was the last Mac to transition to Apple Silicon and was so limited by severely restricted upgrade possibilities that it seemed a strange machine for Apple to launch.
Tim Cook may resign as Apple CEO "as early as next year"
Apple is preparing for Tim Cook to resign as CEO of the company "as early as next year," according to the Financial Times. The company's board of directors and senior executives "have recently intensified preparations for Cook to hand over the reins." Although the report suggests it is unlikely that Apple will appoint a new CEO before the next earnings report at the end of January, it mentions that an announcement early in the year would allow for a smooth leadership transition before Apple's annual WWDC conference in June and the iPhone event in September. No final decisions have been made, and the timeline may change.
Cook has been Apple's CEO since August 2011 and turned 65 this year, an age typically associated with retirement in the United States, but he has not yet publicly announced any plans to resign. John Ternus, Senior Vice President of Hardware Engineering at Apple, is widely seen as the most likely successor to Cook when the time comes. The report is not related to the company's current performance, as Apple reported record revenues in the quarter ending in September and is expecting the December quarter to be the best in the company's history.
Co-founder of an AI transcription company claims there were just two men pretending to be AI
According to a report, an AI transcription company that claimed to offer automated transcriptions of meetings was actually powered by just two people manually writing down what was said. The company's co-founder, Nabeel Hyatt, admitted this in an interview, explaining that in the early days of the company, he and his co-founder manually created the transcriptions to understand how the process works and what their technology should do. This occurred before the firm developed real AI technology for transcription.
This strategy, known as "Wizard of Oz," is actually a common practice in the industry, where companies simulate AI functionality while building the real technology. Hyatt explained that this helped them learn what functionalities are important to users and how the product should work. The company later transitioned to using real AI technology for transcriptions. The revelation raised questions about the transparency of AI startups and how much of their technology is truly automated compared to what is actually hidden human labor.
Airbnb wants to ensure your rental has a well-stocked fridge
Airbnb is working on a new feature that would help hosts provide a better experience for guests by ensuring that properties have well-stocked refrigerators and other essential amenities. The company is exploring ways to standardize the experience for guests and ensure that rental properties meet certain quality standards. This initiative is part of a broader effort to improve service consistency on the Airbnb platform.
The functionality could include guides for hosts on what products to keep in the fridge, recommended supply lists, and possibly partnerships with food delivery services to make it easier for hosts to keep properties well-stocked. Airbnb hopes that this initiative will improve reviews and guest satisfaction, differentiating the platform from traditional hotels by offering a more home-like experience but with the consistency of professional service.
Chinese hackers targeted Anthropic Claude
State-sponsored hackers from China have targeted the AI company Anthropic and its model Claude, according to recent reports. The attack is part of a broader campaign of cyber espionage aimed at Western tech companies and their intellectual property in the field of artificial intelligence. Anthropic, considered one of OpenAI's main competitors, has become an attractive target for state actors interested in cutting-edge AI technology.
Specific details of the attack have not been fully disclosed, but the incident underscores the increasing vulnerability of AI companies to sophisticated cyberattacks. Tech companies are investing more and more in cybersecurity measures to protect their AI models and training data from theft and sabotage. This incident highlights the importance of security in the AI industry and the intense global competition for dominance in artificial intelligence technology.
Google NotebookLM adds Gemini Deep Research
Google has added the Gemini Deep Research functionality to NotebookLM, its AI-powered research tool. This new feature allows users to conduct deeper research on complex topics using the advanced reasoning capabilities of the Gemini model. Deep Research can analyze multiple sources, synthesize information, and generate detailed research reports that go beyond simple summaries.
The functionality is designed for researchers, students, and professionals who need to explore topics in depth. NotebookLM with Deep Research can identify connections between different sources, highlight gaps in research, and suggest additional directions for exploration. The integration represents a significant step in the evolution of AI research tools, moving from simple information retrieval to real analysis and synthesis of knowledge.
Sam Altman praises Kosmos, the AI scientist who discovers clues to new diseases
Sam Altman praised Kosmos, an "AI scientist" that can find clues about new diseases and can do months of work in a single day. The AI system is capable of analyzing massive amounts of medical and scientific data to identify patterns and connections that might escape human researchers. Kosmos represents a new generation of AI tools specifically designed for scientific discoveries and medical research.
The system uses advanced automated learning techniques to process medical literature, clinical study data, and genomic information to generate new hypotheses about diseases and potential treatments. Its ability to radically accelerate the scientific discovery process could revolutionize how medical research is conducted. Altman suggested that tools like Kosmos could be essential for solving complex medical problems and accelerating the development of new treatments.
ChatGPT launches pilot group chats in Japan, New Zealand, South Korea, and Taiwan
OpenAI is testing a new group chat functionality in ChatGPT, launching it in a pilot regime in Japan, New Zealand, South Korea, and Taiwan. The functionality allows multiple users to interact simultaneously with ChatGPT in a shared conversation, facilitating collaboration and brainstorming in groups. This feature marks a significant evolution from the traditional one-on-one interaction model with chatbots.
Group chats could be particularly useful for teams working together on projects, for students collaborating on assignments, or for groups planning events. The functionality maintains the context of the conversation as more users contribute, allowing ChatGPT to respond appropriately to multiple questions and to make connections between different discussion threads. OpenAI is seeking feedback from these pilot markets before deciding on a broader launch of the functionality.
Anthropic demonstrates new capabilities of Claude
Anthropic has demonstrated the improved new capabilities of its Claude model, showcasing advanced reasoning and problem-solving functionalities. The demonstrations highlighted Claude's ability to handle complex multi-step tasks, maintain context in long conversations, and provide detailed explanations for its reasoning. The company emphasized its commitment to safety and transparency in AI.
The new capabilities include improved performance in programming tasks, better data analysis, and extended natural language understanding capabilities. Claude can now more efficiently handle long documents and extract relevant information from multiple sources. Anthropic has stressed its responsible approach to AI development, ensuring that capacity improvements are accompanied by appropriate safety measures to prevent misuse.
Companies turn to OpenAI and Google for AI innovations
Numerous companies are turning to OpenAI, Google, and other major AI providers to fuel their own innovations, according to recent reports. The trend reflects the increasing concentration of power in the AI industry, with a few large companies providing the core technology that many others use to build solutions. This dynamic raises questions about competition, diversity, and dependence in the AI ecosystem.
While smaller companies benefit from access to next-generation AI models without having to develop the technology from scratch, there are concerns about dependence on a few major providers. Some companies are exploring alternatives, including open-source models and specialized AI solutions for specific industries. The AI landscape continues to evolve rapidly, with new players entering the market and emerging technologies challenging established players.
Corporate real estate investments in AI pilots grow, ROI still elusive
Companies are rapidly increasing investments in AI pilot projects for corporate real estate, but return on investment (ROI) remains difficult to quantify, according to a recent report. Although there is significant enthusiasm around AI applications in property management, building automation, and space optimization, many companies struggle to translate these investments into tangible, easily measurable savings or revenues.
The report found that while AI pilots show promise in areas such as predictive maintenance, energy management, and space utilization optimization, many projects remain in the experimental phase. Companies cite challenges such as integrating AI systems with existing infrastructure, a lack of qualified personnel to manage these systems, and the difficulty of establishing clear benchmarks for success. However, industry experts remain optimistic that as technology matures and companies develop more expertise, AI applications in corporate real estate will provide substantial benefits in the future.
AI testing for ChatGPT capabilities may be exaggerated, study claims
A new study suggests that the tests used to evaluate the capabilities of artificial intelligence, such as ChatGPT, may be fundamentally flawed and may exaggerate the actual performance of these systems. Researchers found that many popular benchmarks used to measure AI "intelligence" fail to capture significant limitations and can create misleading impressions that these systems are more capable than they are in reality.
The study highlighted several issues with current testing methods, including overfitting to specific datasets, a lack of assessment of truly general reasoning, and a tendency to favor tasks that align well with how AI models are trained. Researchers advocate for more rigorous testing standards that better evaluate higher-level capabilities, such as abstract reasoning, adaptability to new situations, and contextual understanding. The findings have important implications for how the industry evaluates and reports progress in AI, suggesting the need for greater caution and nuance in assessing the capabilities of artificial intelligence systems.
Synthesis made with the help of a monitoring flow provided by Control F5 Software
