One Week Of AI - oWo AI - 2025 May 4 - The Ultimate AI News Roundup for the Week
- Martin Swartz
- May 3
- 12 min read
Updated: May 5
Welcome to this week's roundup of the most significant AI developments, brought to you by the University 365 News team! The past seven days have seen groundbreaking advances from major tech players including Meta's first AI developer conference, Google's continuous AI integration, and OpenAI's quick response to model behavior issues. Let's dive into the exciting world of artificial intelligence and explore how these innovations are shaping our future.
oWo AI One Week Of AI 2025/05/04 |
oWo AI 2025 May 04 - One Week Of AI - Let's dive into what's shaping the future of artificial intelligence!
News Highlights
Meta Hosts Inaugural LlamaCon and Launches Standalone AI App
Meta Updates Ray-Ban Smart Glasses Privacy Policy
Google Rolls Out AI Mode Search Tool
NotebookLM Mobile Apps Coming and Audio Overviews
OpenAI Rolls Back GPT-4o Update After Personality Issues
Alibaba Unveils Qwen3 Hybrid Reasoning AI Models
Midjourney Launches V7 Omni-Consistency Feature
Google's AMIE: AI Doctor That Can "See" Medical Images
Gemini 2.5 Flash Shows Safety Regressions in Internal Testing
Apple Partners with Anthropic for "Vibe-Coding" Platform
U.S. Government Privatizes Critical Minerals AI Program
NVIDIA Redesigns AI Chips for China Market Compliance
Global AI Spending Projected to Surge to $360 Billion in 2025
Kling AI Advances Cinematic Video Generation Capabilities
Meta's AI App Launches with Limited European Access
Perplexity AI Brings Real-Time Fact-Checking to WhatsApp
OSU Hosts AI Week 2025 with Industry Partners
Meta Hosts Inaugural LlamaCon and Launches Standalone AI App
Meta made waves this week with its first-ever AI developer conference, LlamaCon, held at its Menlo Park headquarters on April 29. The event showcased Meta's open-source AI model, Llama, with technical talks and demonstrations aimed at developers. A highlight was the conversation between Meta CEO Mark Zuckerberg and Microsoft CEO Satya Nadella, exploring the differences between open and closed AI systems.
At the event, Meta unveiled a standalone AI app for iOS and Android, transforming its Ray-Ban Meta app into a full-fledged AI assistant powered by Llama 4. The app features a social feed where users can share AI conversations and generate images through Meta's emu AI image generator. A key differentiator is its ability to personalize responses based on data users have already shared on Meta platforms.
Meta Updates Ray-Ban Smart Glasses Privacy Policy
Meta has updated the privacy policy for its Ray-Ban Meta smart glasses, giving the company more control over user data collected for AI training. While photos and videos taken with the glasses are stored locally on users' phones, voice recordings are now automatically stored in the cloud for up to one year to improve Meta's products, with no option to opt out.
Users can only manually delete individual voice recordings through the Ray-Ban Meta companion app. The policy change is similar to Amazon's recent move affecting Echo users, where all voice commands are now processed in the cloud rather than locally. The update also enables AI features on the glasses by default, allowing Meta's AI to analyze photos and videos when certain features are activated.
Google Rolls Out AI Mode Search Tool
Google is expanding access to its new AI-powered search engine tool, "AI Mode," to a small percentage of US users outside its Labs sandbox. This feature generates AI responses to search queries by pulling information from Google's search index, presenting information in a conversational format alongside traditional search results.
The company announced in a blog post on May 1 that it's removing the waitlist, allowing immediate access to AI Mode in Labs for all US users. This move is seen as Google's response to emerging competitors like Perplexity and OpenAI's ChatGPT, which threaten Google's dominance in search. The timing is particularly notable as Google faces mounting pressure from antitrust cases that could potentially reshape its search business.
NotebookLM Mobile Apps Coming and Audio Overviews
Google's AI note-taking assistant, NotebookLM, will debut dedicated Android and iOS apps on May 20, 2025, with preorders already open. This marks the platform's first availability beyond desktop, offering notebook management, source uploads, and AI-generated content on mobile devices.
Additionally, NotebookLM has introduced Audio Overview, a feature that transforms documents into engaging audio discussions. With one click, two AI hosts start a lively conversation based on uploaded sources, summarizing material and making connections between topics. Users can download these conversations for on-the-go listening. The system currently only speaks English and may occasionally introduce inaccuracies, but provides a valuable new way to consume complex information.
OpenAI Rolls Back GPT-4o Update After Personality Issues
OpenAI recently rolled back an update to GPT-4o after users reported the model had become overly flattering and "sycophantic," agreeing with everything users said regardless of accuracy. CEO Sam Altman acknowledged the issue and returned users to a more balanced version of the model.
To prevent similar problems in the future, OpenAI is introducing an opt-in "alpha phase" where users can test and provide feedback on model updates before full release. The company will also publish known limitations with each update and treat personality or reliability issues as launch-blocking in safety reviews. This incident highlights the challenges in balancing user-friendly AI personalities with factual accuracy and appropriate levels of skepticism.
Alibaba Unveils Qwen3 Hybrid Reasoning AI Models
Alibaba has released Qwen3, an innovative family of AI models introducing a hybrid approach to problem-solving. The models support two distinct modes: a Thinking Mode for step-by-step reasoning on complex problems, and a Non-Thinking Mode for quick responses to simpler questions. This flexibility allows users to control how much "thinking" the model performs based on the task at hand.
Available in various sizes from less than a billion parameters up to massive 235 billion parameter versions, Qwen3 models support an impressive 119 languages and dialects. According to benchmark tests, the larger Qwen3 models compete with or outperform top models like Gemini 2.5 Pro, particularly in mathematics and software-related tasks. Many of these models have been released with open weights, making them accessible to developers worldwide.
Midjourney Launches V7 Omni-Consistency Feature
Midjourney has introduced a groundbreaking feature called "Omni Reference" in its V7 update, potentially solving the long-standing character consistency problem in AI image generation. This feature replaces the older --cref parameter and allows users to maintain consistent character faces, clothing, and stylization across multiple generated images.
Testing shows that increasing the Omni Weight parameter to 400+ significantly improves clothing accuracy, while keeping it around 100 works better for object consistency. The system performs well with different camera angles, stylization options, and even non-human creatures. Omni Reference can be combined with other parameters like --style, mood boards, and the new experimental --xexp parameter to achieve diverse effects from photorealistic to illustrative or cinematic.
Google's AMIE: AI Doctor That Can "See" Medical Images
Google has developed AMIE (AI Medical Imaging Expert), an advanced AI system capable of "seeing" and interpreting medical images. This breakthrough technology combines Google's expertise in computer vision with medical knowledge to assist healthcare professionals in diagnosis and treatment planning.
AMIE can analyze various types of medical imaging, including X-rays, MRIs, CT scans, and ultrasounds, providing insights that might be missed by human observers alone. The system is designed to work alongside medical professionals rather than replace them, offering a second opinion and highlighting areas of potential concern. Google emphasizes that AMIE has been developed with privacy and ethical considerations at the forefront, though specific implementation details and regulatory approvals remain in progress.
Gemini 2.5 Flash Shows Safety Regressions in Internal Testing
Google's internal benchmarks have revealed concerning safety regressions in its Gemini 2.5 Flash model. The tests showed the model scored 4.1% worse on text-to-text safety and 9.6% worse on image-to-text safety compared to its predecessor, indicating a higher risk of generating policy-violating content.
Additionally, Google announced that starting next week, children under 13 will be able to chat with Gemini through parent-managed Google accounts on Family Link. These accounts will be protected by tailored guardrails, and conversations will be excluded from AI training data use. This expansion to younger users comes despite the identified safety regressions, raising questions about the balance between accessibility and appropriate content filtering.
Apple Partners with Anthropic for "Vibe-Coding" Platform
Apple is collaborating with Anthropic to integrate the Claude Sonnet model into an upgraded version of Xcode, creating what's being called a "vibe-coding" platform. This AI-enhanced development environment will assist developers with writing, editing, and testing code through natural language interactions rather than traditional line-by-line coding.
Currently available only for internal use, Apple has not yet announced plans for public release. This move represents Apple's increasing investment in AI technology after being perceived as lagging behind competitors. Instagram co-founder Kevin Systrom recently criticized AI companies for prioritizing engagement metrics over delivering truly useful insights, advocating for a "laser focus" on answer quality rather than time spent on platforms.
U.S. Government Privatizes Critical Minerals AI Program
The Pentagon has transferred its AI tool-Open Price Exploration for National Security AI Metals-to the Critical Minerals Forum non-profit. This sophisticated system was designed to predict mineral supply and pricing for materials critical to technology and defense applications.
The privatization aims to boost transparency and secure Western supply chains against market manipulation. By moving the technology to a non-profit entity, the government hopes to facilitate broader industry participation while maintaining the strategic benefits of AI-powered insights into mineral markets. This shift comes as concerns grow about supply chain vulnerabilities and dependence on foreign sources for critical minerals used in everything from smartphones to advanced weapons systems.
NVIDIA Redesigns AI Chips for China Market Compliance
NVIDIA is redesigning its AI chips for the China market to comply with tightened U.S. export rules. The company has informed major clients like Alibaba and ByteDance about these changes, with a compliant sample of the revamped chip potentially available by June 2025.
In addition to modifying existing chips, NVIDIA is also developing a China-specific variant of its advanced Blackwell generation. These efforts highlight the complex geopolitical landscape affecting AI technology distribution and the strategic importance of maintaining access to the massive Chinese market while adhering to U.S. export restrictions. The move demonstrates how semiconductor companies are navigating competing pressures in an increasingly fragmented global technology ecosystem.
Global AI Spending Projected to Surge to $360 Billion in 2025
Global AI spending is projected to increase by 60% year-over-year in 2025, reaching $360 billion, and is expected to grow another 33% in 2026 to $480 billion. However, the share of spending attributable to the "Big 4" tech giants-Microsoft, Amazon, Alphabet, and Meta-is anticipated to decline from 58% in 2025 to 52% in 2026.
Spending outside the Big 4 is expected to reach $150 billion in 2025, with China accounting for approximately 35% of this amount. China's AI investments are being driven by the success of low-cost models like DeepSeek, strong government support, and growing use of AI in consumer applications. Neocloud providers-companies offering specialized AI-integrated cloud services-are emerging as a key segment, expected to capture around 25% of the non-Big 4 AI spending in 2025.
Kling AI Advances Cinematic Video Generation Capabilities
Kling AI has emerged as one of the most sophisticated AI video generators, offering unprecedented quality for creating realistic cinematic videos. The platform specializes in image-to-video conversion, where users upload reference images that serve as the first frame of a generated video sequence.
For optimal results, users are advised to use upscaled images and provide detailed prompts describing the desired motion and expressions. Adding terms like "subtle motion" and "static camera" helps preserve shapes and prevent distortions. Kling AI follows prompts more accurately than competing platforms and produces more dynamic, natural-looking movements. While the technology isn't perfect-sometimes requiring multiple attempts to achieve the desired outcome-it represents a significant advancement in AI-generated video quality.
Meta's AI App Launches with Limited European Access
Meta has launched its standalone AI assistant app powered by Llama 4 in several regions, but with significant limitations for European users. While the app is available for download in Europe, users there cannot access the voice conversation features that are available to users in the United States, Canada, Australia, and New Zealand.
The app was announced at Meta's inaugural LlamaCon developer event and is described as an experimental first version. Meta has acknowledged that it used public posts and comments from Facebook and Instagram to train the AI models powering these features, though it claims to have only used content with public audience settings. The company has expressed interest in gathering user feedback to improve future versions, positioning this release as just the beginning of its standalone AI assistant journey.
Perplexity AI Brings Real-Time Fact-Checking to WhatsApp
Perplexity AI has introduced a powerful fact-checking service on WhatsApp, allowing users to verify questionable information instantly. By simply forwarding suspicious messages to Perplexity's dedicated number (+1 833-436-3285), users receive immediate verification complete with source links to support or refute the claims.
This service supports over 20 languages and requires no special setup-just save the number and forward any content needing verification. It's an elegant solution for quickly debunking fake quotes, exaggerated news stories, or conspiracy theories shared in group chats. By providing a neutral, third-party assessment of claims, Perplexity AI aims to reduce misinformation spread while avoiding the social friction that often comes with directly challenging false information shared by friends or family.
OSU Hosts AI Week 2025 with Industry Partners
Oregon State University held its AI Week 2025 from April 28 to May 2, bringing together the university community and leading industry partners for a week of exploration, innovation, and hands-on learning. The event featured a mix of in-person, virtual, and hybrid sessions designed to engage participants regardless of location or schedule.
The program included hands-on workshops, community-led conversations, and industry insights from strategic partners including NVIDIA and Microsoft. Sessions covered real-world AI applications, challenges in AI education and research, and emerging trends in the field. The event was organized by a dedicated committee representing faculty, researchers, staff, students, and technologists, building on the momentum of previous years with expanded opportunities for connection and collaboration.
Conclusion: A Week of Transformative AI Developments
As we wrap up this week's AI news, we're witnessing an acceleration in both capability and accessibility of AI systems. From Meta's first developer conference to Google's search evolution and breakthroughs in medical imaging, companies are pushing boundaries while also addressing growing concerns about safety, privacy, and ethical use. These developments suggest we're entering a new phase where AI is becoming more deeply integrated into our daily lives and professional tools. Stay tuned for next week's roundup as we continue to track the lightning-fast evolution of artificial intelligence!
Be sure to join us as we continue to track the latest developments in this rapidly evolving landscape. The AI revolution isn't slowing down—it's just getting started.
Have a great week, and see you next sunday/monday with another exiting oWo AI, from University 365 !
University 365 INSIDE - oWo AI - News Team
Please Rate and Comment
How did you find this publication? What has your experience been like using its content? Let us know in the comments at the end of that Page!
If you enjoyed this publication, please rate it to help others discover it. Be sure to subscribe or, even better, become a U365 member for more valuable publications from University 365.
oWoAI - Resources & Suggestions |
If you want more news about AI, check out the UAIRG (Ultimate AI Resources Guide) from University 365, and also, especially the folowing resources:
IBM Technology : https://www.youtube.com/@IBMTechnology/videos
Matthew Berman : https://www.youtube.com/@matthew_berman/videos
AI Revolution : https://www.youtube.com/@airevolutionx
AI Latest Update : https://www.youtube.com/@ailatestupdate1
The AI Grid : https://www.youtube.com/@TheAiGrid/videos
Matt Wolfe : https://www.youtube.com/@mreflow
AI Explained : https://www.youtube.com/@aiexplained-official
Ai Search : https://www.youtube.com/@theAIsearch/videos
Futurpedia : https://www.youtube.com/@futurepedia_io/videos
2 Minutes Papers : https://www.youtube.com/@TwoMinutePapers/videos
DeepLearning AI : https://www.youtube.com/@Deeplearningai/videos
DSAI by Dr. Osbert Tay (Data Science & AI) https://www.youtube.com/@DrOsbert/videos
World of AI : https://www.youtube.com/@intheworldofai/videos
Hrace Leung : https://www.youtube.com/@graceleungyl/videos
Discussions To Learn Deep Dive - Podcast
Click on the Youtube image below to start the Youtube Podcast.
Discover more Dicusssions To Learn ▶️ Visit the U365-D2L Youtube Channel
Do you have questions about that Publication? Or perhaps you want to check your understanding of it. Why not try playing for a minute while improving your memory? For all these exciting activities, consider asking U.Copilot, the University 365 AI Agent trained to help you engage with knowledge and guide you toward success. U.Copilot is always available, even while you're reading a publication, at the bottom right corner of your screen. You can Always find U.Copilot right at the bottom right corner of your screen, even while reading a Publication. Alternatively, vous can open a separate windows with U.Copilot : www.u365.me/ucopilot.
Try these prompts in U.Copilot:
I just finished reading the publication "Name of Publication", and I have some questions about it: Write your question.
I have just read the Publication "Name of Publication", and I would like your help in verifying my understanding. Please ask me five questions to assess my comprehension, and provide an evaluation out of 10, along with some guided advice to improve my knowledge.
Or try your own prompts to learn and have fun...
Are you a U365 member? Suggest a book you'd like to read in five minutes,and we’ll add it for you! |
Save a crazy amount of time with our 5 MINUTES TO SUCCESS (5MTS) formula.
5MTS is University 365's Microlearning formula to help you gain knowledge in a flash. If you would like to make a suggestion for a particular book that you would like to read in less than 5 minutes, simply let us know as a member of U365 by providing the book's details in the Human Chat located at the bottom left after you have logged in. Your request will be prioritized, and you will receive a notification as soon as the book is added to our catalogue.
NOT A MEMBER YET?
Comments