Category: Artificial Intelligence

• Open-Source AI Model Performance is highlighted by the launch of Reflection 70B, which outperforms GPT-4o and Claude Sonnet 3.5, making it the leading open-source model.
• Developed by HyperWrite using a novel Reflection-Tuning technique, Reflection 70B can detect and correct its own mistakes before finalizing answers.
• The model shows significant improvements in benchmark tests such as MMLU, MATH, IFEval, and GSM8K, addressing issues like model hallucinations and enhancing reasoning accuracy.
• Reflection 70B outputs internal reasoning and final answers using specific tags, showcasing its advanced capabilities in understanding and correcting errors.
• An upcoming model, Reflection 405B, is expected to further elevate performance standards for language models globally, following the recent success of Alibaba’s Qwen2-VL, which also outperforms closed-source models.

https://analyticsindiamag.com/ai-news-updates/new-open-source-champion-reflection-70b-outperforms-gpt-4o-and-claude-sonnet-3-5/…

• Open-source solves OpenAI’s ‘strawberry’ problem with the introduction of Reflection 70B, a leading model utilizing reflection tuning to correct its own mistakes.
• Reflection 70B is based on Llama 3.1-70B Instruct and is gaining traction in the competitive landscape of commercial AI models.
• Developer Matt Schumer claims Reflection 70B is “the world’s best open-source AI model,” showcasing strong benchmark performance.
• The model’s reflection tuning technique allows it to identify and rectify errors before providing full answers, addressing issues present in OpenAI’s models.
• Meta has already reached out to the developers after the model’s launch, indicating its potential and demand in the AI community.

https://www.techzine.eu/news/applications/124161/open-source-already-solves-openais-strawberry-problem/…

• The new open source AI model, Reflection 70B, developed by HyperWrite, is based on Meta’s Llama 3.1-70B Instruct and features an innovative error self-correction technique.
• Matt Shumer, co-founder of HyperWrite, claims Reflection 70B is now the world’s top open-source AI model, outperforming others in various benchmarks.
• The model utilizes Reflection-Tuning, allowing it to identify and correct its own mistakes before finalizing responses, enhancing accuracy and reliability.
• Reflection 70B introduces special tokens for structured reasoning and error correction, making it easier for users to interact with and receive accurate outputs.
• A larger model, Reflection 405B, is set to be released soon, with expectations to outperform even the leading closed-source models like OpenAI’s GPT-4o.

https://venturebeat.com/ai/meet-the-new-most-powerful-open-source-ai-model-in-the-world-hyperwrites-reflection-70b/…

• Google is testing its “Ask Photos” AI assistant, which understands what’s in your pictures and allows users to explore their photo libraries through natural language queries.
• The feature is currently available to select Google Labs users in the US and can answer questions about past events and retrieve relevant images.
• Users can also utilize Ask Photos to summarize recent vacations or select the best family pictures for shared albums.
• The Photos app will continue to support a “classic search” option while enhancing search capabilities with natural language queries.
• A new Collection page has replaced the Library tab to streamline photo and video access, with more language support expected soon.

https://www.theverge.com/2024/9/5/24236817/google-ask-photos-ai-assistant-test…

• Apple AI gadgets innovation aims to transform the current underwhelming landscape of AI gadgets, with the company set to unveil new iPhones featuring Apple Intelligence.
• Major tech companies like Microsoft, Google, and Samsung have launched AI devices, but many have failed to impress consumers, often only offering incremental features rather than groundbreaking experiences.
• Apple Intelligence will enhance Siri and integrate with apps, but its success hinges on convincing millions of iPhone users to upgrade, especially amidst a post-pandemic sales slump.
• Consumer interest in AI-driven gadgets varies, with only 15% of U.S. respondents expressing a high inclination to purchase, compared to 43% in China, emphasizing the need for effective marketing.
• Regulatory challenges in markets like China and the EU may hinder Apple’s global rollout of Apple Intelligence, as the company navigates complex approval processes to expand its user base.

https://www.cnbc.com/2024/09/05/ai-gadgets-have-been-a-bust-so-far-apple-aims-to-change-that.html…

• iPhone 16 AI features will debut at Apple’s fall event on September 9, 2024, marking the first integration of AI in Apple smartphones, including both base and Pro models.
• Siri will receive a significant overhaul, featuring a modern design and improved natural language understanding, allowing for better interaction and follow-up commands.
• The new “Genmoji” feature will enable users to create personalized emojis using text prompts, enhancing self-expression with custom symbols.
• Safari will leverage machine learning for a redesigned browsing experience, offering highlights, summaries, and a distraction-free reading mode for articles.
• The iPhone 16 will introduce Voice Transcribe, converting voice recordings into text while ensuring ethical standards by notifying participants during call recordings.

https://www.indiatoday.in/technology/news/story/iphone-16-to-launch-on-september-9-5-ai-features-coming-to-base-models-2592728-2024-09-03…

• Gemini Extensions make Google’s chatbot smarter than ChatGPT by enabling it to perform useful real-world tasks and access personalized information.
• Google has transitioned to supporting Gemini as the default assistant on the Pixel 9 series, bridging the feature gap with extensions that enhance its functionality.
• Extensions allow Gemini to interact with various web services, providing accurate responses by consulting real-time data from sources like Google Maps and Gmail.
• Users can enable or disable Gemini Extensions individually, maintaining control over their personal data while enjoying features like trip planning and video summarization.
• Future updates may introduce more extensions, such as those for WhatsApp and smart home control, further expanding Gemini’s capabilities.

https://www.androidauthority.com/gemini-extensions-3477277/…

• Emotion AI business software is emerging as a trend in the AI landscape, aiming to help bots understand human emotions for better interaction in roles like customer service and sales.
• This technology utilizes multimodal inputs, including visual and audio sensors, combined with machine learning to interpret emotions during interactions.
• Major cloud providers, such as Microsoft and Amazon, offer Emotion AI capabilities, leading to a rise in startups focused on this technology.
• Despite its potential, there are concerns about the effectiveness of Emotion AI, as research suggests that human emotions cannot be accurately determined through facial expressions alone.
• Regulatory challenges, such as the EU’s AI Act, may hinder the development and application of Emotion AI in certain sectors, raising questions about its future viability.

https://techcrunch.com/2024/09/01/emotion-ai-could-be-the-next-trend-for-business-software-and-that-could-be-problematic/…

• Google AI search update includes new safeguards preventing AI from responding to election-related queries ahead of the U.S. presidential election.
• The company is implementing restrictions across its generative AI products, including Gemini and YouTube AI summaries, to combat misinformation.
• Features like content labels, digital watermarking, and polling place information are being introduced to provide authoritative election-related information.
• Google aims to help users find credible information on registering to vote and access resources from state election offices through its Search and YouTube platforms.
• These measures are intended to remain in place beyond the election to continue preventing misinformation on Google’s platforms.

https://www.tomsguide.com/ai/google-just-made-a-major-change-to-its-ai-search-engine-heres-what-you-need-to-know…

• ChatGPT active user statistics show over 200 million weekly active users, doubling since November 2023, with 92% of Fortune 500 companies adopting its products.
• Despite the impressive user growth, skepticism remains about the mass-market utility of generative AI tools, with critics questioning OpenAI’s long-term viability.
• Many users may not openly admit to using ChatGPT due to corporate restrictions and fears surrounding generative AI, leading to a culture of secrecy among employees.
• Social stigmas related to job loss, privacy, and ethical concerns are complicating public perceptions of generative AI technology.
• OpenAI reports increased usage of its tools in various sectors like healthcare and education, while competition in the generative AI market continues to intensify among major tech companies.

https://arstechnica.com/information-technology/2024/08/chatgpt-hits-200-million-active-weekly-users-but-how-many-will-admit-using-it/…

• Apple Intelligence AI release is expected to debut for iPhone users with iOS 18.1, iPadOS 18.1, and macOS Sequoia 15.1 at the end of October 2024.
• Major OS updates will be released in mid-to-late September 2024, but Apple Intelligence features will not be included until the subsequent update.
• Historically, Apple releases x.1 updates around the end of October, with the likely dates for iOS 18.1 being October 23, 24, or 28.
• Initial Apple Intelligence features will be available at launch, with more potentially arriving in future updates, including ChatGPT integration in iOS 18.2 by December.
• Not all devices will support Apple Intelligence; it will be limited to modern devices, including all iPhone 16 models, and features will be labeled as ‘beta’ post-launch.

https://9to5mac.com/apple-intelligence-release-date/…

• Gmail users on Android can now chat with Gemini about their emails through the new Gmail Q&A feature, which allows for interaction with Google’s AI assistant within the Gmail app.
• The feature, available to paying users, enables Gemini to summarize emails and search for specific details in the user’s inbox.
• Users can access Gmail Q&A by clicking the Gemini logo in the app, and it will also cite the source email in its responses.
• Currently, the feature is limited to emails, but Google plans to expand it to include files from Google Drive in the future.
• Google aims to encourage subscriptions to Gemini by integrating AI features across its products, with no plans for a free version of Gmail Q&A in the near term.

https://techcrunch.com/2024/08/29/gmail-users-on-android-can-now-chat-with-gemini-about-their-emails/…

• Apple training blocked on several major publishing websites as companies opt out of using their data for AI training.
• Apple introduced a tool allowing companies to refuse AI data scraping, offering millions in exchange for access to their content.
• Significant companies like The New York Times, Facebook, and Instagram have chosen to block Applebot from training its AI.
• Studies indicate that 6% to 7% of high-traffic websites are blocking Applebot, while a quarter of all sites have opted out.
• Apple has faced scrutiny over its data collection methods, including accusations of using stolen YouTube captions, but maintains that no such data was used for training.

https://www.tomsguide.com/ai/apple-blocked-from-training-apple-intelligence-on-several-publishing-websites-heres-what-we-know…

• AI replicates playable Doom, as Google researchers create an algorithm-driven version of the classic game using neural networks and video clips of gameplay.
• The project, detailed in a paper titled “Diffusion models are real-time game engines,” demonstrates the ability to simulate complex game mechanics at 20fps.
• GameNGen, the neural model developed by the team, can perform tasks like tracking health and ammo, attacking enemies, and maintaining game state over time.
• While the output quality is comparable to the original Doom, human input is still necessary to provide gameplay footage for the AI to learn from.
• This milestone raises questions about the future of game development, with over 60% of developers already incorporating AI into their processes, highlighting both potential benefits and challenges.

https://www.eurogamer.net/google-reaches-grim-ai-milestone-replicates-playable-version-of-doom-entirely-by-algorithm…

• AI object removal is a new feature in Apple’s latest iOS update, allowing users to remove objects from photos seamlessly using AI technology.
• The feature, named Clean Up, enables users to select and delete unwanted objects with a single tap or by circling them.
• Apple’s AI system intelligently fills in the background, even addressing shadows and reflections from the removed object.
• This update is part of the iOS 18.1 developer beta, which also includes various Apple Intelligence features like writing tools and notification summaries.
• Currently, Apple Intelligence features are available only to English-speaking users in the U.S. region.

https://techcrunch.com/2024/08/28/apple-introduces-ai-powered-object-removal-with-the-latest-ios-update/…

• Midjourney AI image generator is venturing into hardware development and is seeking new team members for this initiative.
• The specific type of hardware being developed remains unclear, as the company typically focuses on AI image generation rather than hardware.
• Key personnel, including founder David Holz and Head of Hardware Ahmand Abbas, have backgrounds in virtual and augmented reality, suggesting potential AR/VR hardware projects.
• Midjourney has hinted at multiple hardware efforts and teased responses about their plans on social media, indicating a playful engagement with their audience.
• The company is bootstrapped and has not raised venture capital, but its ambitious hardware plans may lead to seeking funding in the future.

https://siliconangle.com/2024/08/28/ai-image-generator-midjourney-venturing-hardware-development/…

• Google’s Gemini will let users create AI images of people following a controversial history of generating inappropriate and misleading content.
• The company has launched an early access version of its upgraded image-generating feature for subscription service users, promising improved image quality.
• Google acknowledged past failures, including generating “absurdly woke” images and providing dangerous misinformation through its AI systems.
• The tech giant is committed to ongoing improvements by incorporating user feedback and enhancing its evaluation processes.
• Other tech companies, like Meta, have also faced criticism for errors in their AI systems, highlighting a broader industry challenge.

https://nypost.com/2024/08/28/business/google-to-let-users-generate-ai-images-of-people-on-gemini/…

• Google AI can generate images of humans again, implementing safeguards to avoid controversial outputs such as photorealistic identifiable individuals and violent or sexual content.
• The new Imagen 3 model prohibits generating images of prominent people, as demonstrated by its refusal to create an image of “President Biden playing basketball.”
• Imagen 3 has improved upon previous models by avoiding historically inaccurate racial representations, generating more contextually appropriate images for certain historical prompts.
• Some prompts still encounter restrictions, with requests for images of specific historical figures or events leading to error messages due to content policy violations.
• Google plans to gradually roll out the new model, aiming for broader availability and continued improvement based on user feedback.

https://arstechnica.com/ai/2024/08/months-after-controversy-google-ai-can-generate-images-of-humans-again/…

• AI makes Doom in its own game engine — Google’s GameNGen project utilizes Stable Diffusion to generate original gameplay through a neural network.
• The GameNGen engine processes player inputs and previous frames to create new visuals with notable fidelity and consistency.
• Players can engage in real-time gameplay mechanics like movement, shooting, and managing ammo, with the game running at 20 FPS.
• Google trained its AI agent on various skill levels in Doom to gather extensive training data, rewarding successful actions and penalizing failures.
• Despite some inconsistencies and visual artifacts, GameNGen represents a significant step in AI game development, potentially influencing future gaming technologies.

https://www.tomshardware.com/tech-industry/artificial-intelligence/ai-makes-doom-in-its-own-game-engine-googles-gamengen-project-uses-stable-diffusion-to-simulate-gameplay…

• iOS 18.1 beta notification summaries are now available for all apps, expanding the Apple Intelligence feature previously limited to two Apple apps.
• The latest beta 3 update for iOS 18.1, iPadOS 18.1, and macOS Sequoia 15.1 introduces new capabilities, enhancing the notification experience.
• Summaries utilize AI to intelligently condense messages, emails, and news headlines into more useful notifications.
• Users can enable or disable notification summaries for individual apps through the Settings menu, allowing for personalized notification preferences.
• While generally beneficial, the summaries may occasionally contain inaccuracies, but overall they have proven to be a valuable addition across devices.

https://9to5mac.com/2024/08/28/ios-181-beta-3-expands-apple-intelligence-notification-summaries-beyond-messages-and-mail/…

• Google AI image generator is being reintroduced after backlash over controversial image generation of people, particularly in unrealistic settings.
• The updated tool, Gemini Advanced, will feature Google’s latest model, Imagen 3, and aims to improve user experience while generating images.
• Google had previously disabled the feature due to criticism over biased and inappropriate outputs, which sparked significant public and media backlash.
• The reintroduction comes amid a competitive landscape in AI, with other tech giants like Apple making strides in marketing their AI tools effectively.
• Google plans to limit the tool’s capabilities to prevent abuse, such as generating images of identifiable individuals or excessive violence, while continuing to seek user feedback for improvements.

https://finance.yahoo.com/news/google-revamps-controversial-ai-image-generation-tool-after-backlash-194238918.html…

• Google has rolled out three new ‘stronger’ Gemini models: Gemini 1.5 Flash-8B, improved Gemini 1.5 Flash, and stronger Gemini 1.5 Pro, aimed at enhancing performance and gathering user feedback.
• The models can handle complex information from over 10 million tokens and support various inputs, including documents, video, and audio, accessible via Google AI Studio and the Gemini API.
• Internal tests showed significant improvements for Gemini 1.5 Flash, which climbed from 23rd to 6th place in performance rankings, while Gemini 1.5 Pro excelled in math, coding, and complex prompts.
• Users have noted solid enhancements in image analysis and speed, while some expressed a desire for a more substantial update like Gemini 2.0 and criticized performance on longer outputs and coding tasks.
• Google plans to automatically transition users to the new models by September 3, 2024, and will continue refining the models based on user feedback in future updates.

https://gagadget.com/en/ai/496453-google-rolls-out-three-new-stronger-gemini-models/…

• Midjourney AI hardware announcement reveals the company is officially entering the hardware market, seeking to build a new team in San Francisco.
• The generative AI art maker, launched in 2022, recently made its image creation tools accessible for free, albeit with limitations.
• Midjourney has hinted at multiple hardware projects in development but has not disclosed specific details or timelines for product releases.
• The company has appointed a former Apple Vision Pro engineer as its head of hardware, sparking speculation about potential hardware offerings.
• Midjourney’s social media engagement reflects public curiosity, with the company responding to inquiries while maintaining a degree of secrecy about its plans.

https://www.neowin.net/news/the-midjourney-ai-image-maker-company-now-says-it-is-getting-into-hardware/…

• The NotePin is the latest pointless AI wearable app that should simply function as an app instead of requiring additional hardware.
• Plaud’s NotePin mimics existing iPhone capabilities, offering a wearable device for note-taking that can be worn in various ways.
• Targeted at productivity enthusiasts, it allows users to record audio for later review, but its utility is questioned given the iPhone’s existing features.
• The device costs $169, with an optional $79 annual fee for additional recording minutes and features, raising concerns over its value.
• Overall, the trend of creating AI hardware that replicates smartphone functions is seen as unnecessary, suggesting that such innovations would be better suited as apps.

https://9to5mac.com/2024/08/28/notepin-pointless-ai-wearable/…

• The Plaud NotePin is one of the most practical AI gadgets of 2024, designed to simply record audio with a button press without overcomplicating the user experience.
• It can be worn in various styles, such as a pin, necklace, or wristband, making it versatile for different settings.
• The device utilizes OpenAI’s GPT-4o model for voice transcription and summarization, with recordings accessible via a mobile app and web dashboard.
• Users can choose between a free plan offering 300 minutes of recording monthly and a Pro plan for $6.60 a month, which includes additional features and 1,200 minutes of recording.
• The NotePin supports 59 languages and will be available for $169, with preorders starting soon, while competitors like Apple and Google continue to expand their AI capabilities.

https://www.digitaltrends.com/mobile/plaud-notepin-may-be-one-of-2024s-most-practical-ai-gadgets/…

• Google Meet AI meeting notes feature, called “Take notes for me,” is rolling out to Google Workspace users, utilizing Gemini AI for automatic note-taking.
• Initially available only to users with specific Google Workspace plans, the tool aims to enhance focus during meetings by allowing users to concentrate on discussions.
• The feature captures notes in a Google Docs file, saves them to the meeting owner’s Google Drive, and attaches the document to the calendar event for easy access by colleagues.
• After meetings, participants receive an email link to the notes, and late joiners can view a “summary so far” to catch up on discussions.
• Currently, the feature is limited to Google Meet on computers, requires spoken English, and is part of a broader suite of AI tools designed to improve workplace productivity.

https://www.techradar.com/pro/google-meet-calls-will-now-automatically-take-notes-for-you…

• AI Doomers reflect on their current perspectives, noting a significant increase in interest and concern about AI risks since 2016, when the community was much smaller and less influential.
• The release of ChatGPT in 2022 brought AI risk experts into the spotlight, allowing them to communicate their concerns more broadly, but public interest in AI dangers has since waned.
• Key figures like Eliezer Yudkowsky have expressed despair over the lack of effective governance in AI development, particularly following the tumultuous events at OpenAI.
• The AI-risk community has learned that initial commitments to safety and cooperation with regulators often falter under financial pressures and corporate interests.
• Yudkowsky advocates for a more controlled approach to AI research, suggesting it should occur within a national-security framework to better manage existential risks, although he acknowledges the unpredictability of technological evolution.

https://www.theatlantic.com/technology/archive/2024/08/helen-toner-interview-doomers/679624/…

• Hobbyists are discovering how to insert custom fonts into AI-generated images using the Flux AI image synthesis model, which effectively renders custom-trained typefaces.
• The technique allows users to create accurate text representations in AI images, enhancing the realism of scenes like chalkboard menus or business cards.
• Previously, most AI image generators struggled with text accuracy, but Flux excels in rendering “in-world text,” surpassing models like Stable Diffusion and DALL-E 3.
• Two custom typeface LoRAs have already been developed, including a Y2K style font and a font inspired by Cyberpunk 2077, showcasing the potential for unique styles in AI imagery.
• While concerns about environmental impact and ethical use of AI persist, the community is rapidly exploring new applications, indicating that typeface LoRAs could become a significant aspect of AI image synthesis.

https://arstechnica.com/information-technology/2024/08/hobbyists-discover-how-to-insert-custom-fonts-into-ai-generated-images/…

• The Apple September Event 2023 is scheduled for September 9, with a potential focus on Apple Intelligence alongside the iPhone 16 lineup.
• The event invite features a “It’s Glowtime” design that may hint at the introduction of Apple Intelligence and its new generative AI features.
• Apple Intelligence, which includes updates to Siri and creative tools like GenMoji, is expected to be highlighted during the event.
• The new features are anticipated to be available on the iPhone 16 and other devices with M1 processors, including recent iPads and Macs.
• Speculations suggest that the iPhone 16 may feature bolder colors, contrasting with the more muted tones of the iPhone 15 models.

https://www.cnet.com/tech/mobile/apple-september-event-could-the-invite-hint-at-more-apple-intelligence/…

• Apple may be working on an AI personality to replace Siri, utilizing generative AI for a new human-like interface in upcoming robotic devices.
• Mark Gurman from Bloomberg reports that this AI assistant could be integrated into a tabletop robot and other future Apple robotics, which are still in early development stages.
• The envisioned tabletop robot is expected to be priced under $1,000, drawing comparisons to Amazon’s Astro.
• Apple is enhancing its generative AI capabilities across its devices, with features like text creation and image generation, largely powered by ChatGPT for the time being.
• While there is potential in Apple’s robotics projects, it remains uncertain how committed the company is to launching an actual home robot.

https://www.techradar.com/computing/artificial-intelligence/apple-may-be-working-on-an-ai-personality-to-replace-siri-on-its-robots…

• Apple Intelligence iPhone features include a ChatGPT-powered Siri and various AI tools set to launch with iOS 18, promising to enhance user interactions and streamline tasks.
• The new AI capabilities will debut with the iPhone 16 models in September, while the full features will arrive in the iOS 18.1 update expected in mid- to late October.
• Apple emphasizes data privacy, claiming that its integration of ChatGPT will anonymize and encrypt user data, providing an extra layer of protection compared to direct use of ChatGPT.
• Users will have access to new functionalities like writing tools, email summarization, and phone call transcription, but older devices will not support these features.
• While Apple aims for high-quality AI performance, there are concerns about potential inaccuracies and the need for access to personal data, leaving users with the choice to opt-in or out of Apple Intelligence.

https://www.theguardian.com/technology/article/2024/aug/24/apple-intelligence-iphone-ios-18-siri-chat-gpt-launch…

• Pixel 9 Pro AI review reveals the phone is “exceptional” yet “gimmicky,” showcasing a sleek design and advanced AI capabilities, particularly in imagery.
• The Pixel 9 Pro XL features a 6.8-inch display, powerful Tensor G4 chip, and impressive camera system, though it has drawbacks like high cost and lack of expandable storage.
• AI-generated reviews of the Pixel 9 Pro XL varied significantly, with some highlighting poor macro photography and inconsistent battery life, reflecting the unpredictable nature of AI responses.
• Factual inaccuracies were noted in AI reviews, including incorrect references to the Tensor G3 processor and other key specifications, raising concerns about the reliability of AI-generated information.
• The experience emphasizes the importance of trusting expert opinions over AI-generated content, advocating for critical thinking and common sense when considering AI advice.

https://www.cnet.com/tech/mobile/i-asked-the-pixel-9-pros-ai-to-write-a-review-of-itself-heres-what-it-said/…

• Google Pixel 9 Pro features a significant shift towards AI integration, particularly with its new assistant, Gemini, which enhances user interaction through a conversational mode.
• Users have reported mixed experiences with Gemini, encountering issues such as unresponsiveness and inaccurate information, indicating that the technology is still evolving.
• The Pixel 9 Pro’s design is improved with a flat-sided aesthetic and a faster under-display fingerprint reader, but it still faces challenges with low-light photo quality.
• While the Pixel 9 Pro offers high-end features and a starting price of $999, it raises concerns about storage limitations compared to more affordable alternatives.
• The Pixel 9 series marks a turning point for Google, focusing more on AI capabilities than on showcasing the latest Android version, signaling a new direction for the brand.

https://www.cnet.com/tech/mobile/google-pixel-9-pro-goodbye-pixel-hello-gemini-ai/…

• DeepMind workers protest contracts related to Google’s defense engagements, expressing concerns over military partnerships and their impact on ethical AI leadership.
• Over 200 employees signed a letter in May highlighting issues with Google’s contracts supplying AI services to the Israeli military.
• The letter emphasizes that such military involvement contradicts DeepMind’s mission statement and AI Principles established by Google.
• This internal dissent suggests a cultural clash between DeepMind and Google, particularly following Google’s previous commitments against military use of their technology.
• The protest reflects a growing concern among tech workers about the ethical implications of collaborating with military organizations.

https://techcrunch.com/2024/08/22/deepmind-workers-sign-letter-in-protest-of-googles-defense-contracts/…

• Google Keep will now tap into Gemini to help you create all kinds of lists, enhancing its note-taking capabilities with AI features.
• The “Help me create a list” feature, powered by Gemini, will assist users in generating notes like grocery lists and is accessible to all users, not just Google One subscribers.
• This integration signifies the transition from Google Assistant to Gemini, which has been gaining popularity since its launch.
• Google Keep’s latest version allows users to toggle AI assistance on and off, making it user-friendly and adaptable.
• Despite its strengths, Google Keep faces competition from other note-taking apps like Apple Notes and Microsoft OneNote, which may offer more focused features.

https://www.androidpolice.com/google-help-me-note-gemini-keep/…

• Google Keep AI list creation feature, “Help me create a list,” is now rolling out widely on Android after months of testing.
• The feature utilizes Gemini technology and was previously part of Workspace Labs, allowing users to generate lists from simple prompts.
• Users can create lists for various tasks, like grocery shopping, with automatic formatting applied to the generated items.
• The rollout is available to all Android users, regardless of device or Workspace Labs subscription status, and works on multiple Google accounts.
• A newer version of Keep includes an option to disable the AI feature, but it is not necessary for using the list creation capability.

https://9to5google.com/2024/08/22/google-keep-help-me-create-list-ai-rollout/…

• The Pixel 9 Pro XL review highlights Google’s push towards an AI-driven smartphone experience, building on the foundation laid by the Pixel 8 series.
• The device features a slightly larger and bulkier design with a 6.8-inch display and retains the camera specifications of its predecessor, delivering dependable photography performance.
• Key AI enhancements include generative AI tools like Magic Eraser, Best Shot, and the new Add Me feature for improved photo editing and composition.
• Powered by the Tensor G4 chipset, the Pixel 9 Pro XL focuses on AI efficiency over raw performance, providing a smooth user experience for most applications.
• With a 5060 mAh battery and improved charging capabilities, the Pixel 9 Pro XL emphasizes longevity and practicality, positioning itself as a technology demonstrator for Google’s future AI ambitions.

https://www.forbes.com/sites/ewanspence/2024/08/22/google-pixel-9-pro-xl-review-android-gemini-ai-smartphone-review/…

• OpenAI GPT-4 fine-tuning has been launched, allowing developers to enhance performance and accuracy for their applications using custom datasets.
• The feature, available to all developers on paid tiers, enables customization of response structure and tone, as well as adherence to complex instructions.
• Developers can achieve strong results with minimal training examples, demonstrating the effectiveness of the fine-tuning process.
• Fine-tuned models provide full ownership of business data, ensuring that inputs and outputs are not shared or used for training other models.
• OpenAI is also rolling out an advanced voice feature for ChatGPT, utilizing GPT-4o for realistic audio responses and real-time conversations.

https://www.pymnts.com/news/artificial-intelligence/2024/openai-launches-fine-tuning-feature-gpt-4o/…

• The iOS design app Procreate has pledged to never introduce AI, emphasizing that creativity should be made, not generated.
• CEO James Cuda criticizes generative AI, claiming it detracts from the humanity of creative work and is built on theft.
• Procreate has gained user support for its stance against AI, which aligns with concerns from many creative professionals.
• The app is priced at $12.99 and avoids a subscription model, offering features like Animation Assist and QuickShape.
• Procreate’s commitment to a non-AI approach distinguishes it from other software companies increasingly integrating AI into their products.

https://www.zdnet.com/article/at-least-one-ios-design-app-has-pledged-to-never-introduce-generative-ai/…

• Procreate AI controversy response highlights the company’s firm stance against generative AI, emphasizing that “creativity is made, not generated.”
• The Procreate team argues that generative AI is detrimental to human creativity and is built on a foundation of theft.
• Procreate’s CEO explicitly expressed his disdain for generative AI, stating, “I really f*cking hate generative AI,” resonating with many digital artists.
• The company’s clear opposition to AI contrasts with other digital art software companies, which are increasingly adopting AI features.
• Procreate’s approach has garnered positive reactions from the creative community, reinforcing its commitment to empowering artists rather than exploiting them.

https://www.creativebloq.com/ai/procreate-comes-out-swinging-against-ai-and-creatives-are-on-board…

• Google’s Gemini AI features show promise for enhancing productivity on mobile devices, particularly with the Pixel 9 models.
• The multimodal capabilities of Gemini allow it to recognize text, images, code, and video, streamlining tasks like setting reminders and compiling notes.
• Gemini’s research functionality enables users to delegate information gathering while maintaining control over the process, enhancing efficiency.
• The Gemini Live feature offers a natural-sounding voice assistant that facilitates brainstorming and informal conversations, improving user interaction.
• Despite lingering skepticism about AI on phones, Gemini demonstrates practical applications that go beyond superficial features, making it a tool for increased productivity.

https://www.tomsguide.com/ai/ive-been-skeptical-about-ai-features-on-phones-but-googles-gemini-might-change-my-mind…

• Swiss tech firm FinalSpark launches an AI human brain cells rental service, offering cloud access to biocomputers made from human brain cells for $500 a month.
• These organoids, or clumps of human brain cells, can live for about 100 days, after which the AI system would need to be transitioned or would cease to function.
• FinalSpark aims to provide a highly efficient, low-energy alternative for developing AI models, claiming organic biocomputers could be up to 100,000 times more efficient than traditional silicon-based technology.
• Training AI models using brain cells requires unique reinforcement techniques, such as administering dopamine for positive reinforcement and electrical signals for negative reinforcement.
• The potential exists for future AI systems to have longer lifespans and possibly develop blockchain technologies using organic material, aligning AI experiences more closely with human experiences.

https://cointelegraph.com/news/swiss-tech-firm-launches-ai-made-human-brain-cells-rental-service…

• Google Gemini Live launch introduces a conversational AI voice assistant to compete with OpenAI’s Advanced Voice Mode, amid a showcase of new Pixel 9 phones and AI-powered tools.
• Google unveiled the Pixel 9 Fold Pro priced at $1,799, along with the Pixel Buds Pro 2 featuring Gemini AI capabilities.
• Epic Games launched its rival iOS app store in the EU, featuring popular games like Fortnite and Rocket League, following Apple’s removal of Fortnite from its store.
• X (formerly Twitter) released Grok-2 and Grok-2 mini in beta, enhancing reasoning and image generation capabilities, though with limited access for users.
• CrowdStrike received the “Most Epic Fail” award at Def Con after a software update caused a global IT meltdown, while Waymo began testing autonomous robotaxis on freeways in California.

https://techcrunch.com/2024/08/17/google-takes-on-openai-with-gemini-live/…

• Midjourney AI image editor offers expanded capabilities for users to enhance, modify, and reimagine both AI-generated and real-world images.
• To use the editor, log in, generate an image with your prompt, and access the editor from the bottom-right corner.
• Users can edit the original prompt, reposition images, and utilize tools like erase and restore for precise modifications.
• The scale tool allows for zooming in on details or zooming out for an overview, while aspect ratio options help alter image composition.
• After making edits, users can submit their changes to generate the newly edited image, reflecting the ongoing advancements in AI technology.

https://www.tomsguide.com/ai/ai-image-video/how-to-access-and-use-midjourneys-new-ai-image-editor…

• AI image generation advancements have seen significant updates this week from major players like Midjourney, Google, and Grok, showcasing the evolving technology.
• Midjourney introduced a new web editor that consolidates various image manipulation tools for a more seamless editing experience, moving away from its original Discord-based platform.
• Grok-2, launched by xAI, features impressive image generation capabilities but has raised concerns over its lack of guidelines regarding content moderation and intellectual property.
• Google’s Imagen 3 model claims to offer the highest quality text-to-image generation with improved detail and rendering, currently available in a closed beta.
• Other advancements include OpenAI’s release of the GPT-4o mini model, which enhances accessibility for developers, and Stability AI’s updated licensing for Stable Diffusion, expanding free usage.

https://www.digitaltrends.com/computing/midjourney-new-web-editor/…

• Humans simulating AI glitches has become a viral trend, with users on TikTok and Bilibili creating videos that mock the imperfections of AI-generated content.
• The trend features exaggerated mimicry of bizarre visual non-sequiturs, where users position their bodies to replicate the oddities seen in AI videos.
• These humorous parodies have gained popularity on X (formerly Twitter), with many viewers finding them entertaining and creative.
• Despite the trend, tech companies like OpenAI and Runway are working to improve AI-generated video realism through enhanced training data and computational methods.
• The phenomenon reflects a broader cultural engagement with AI, where even celebrities like Will Smith have participated in mocking flawed AI outputs.

https://arstechnica.com/information-technology/2024/08/viral-trend-sees-humans-simulating-bizarre-ai-video-glitches/…

• Google is bringing AI overviews to India, Brazil, Japan, the UK, Indonesia, and Mexico, following their initial launch in the U.S.
• The expansion includes local language support and features tailored specifically for Indian users, such as toggling between Hindi and English results.
• Google plans to enhance the display of source material links by adding site icons above AI overviews on both desktop and mobile.
• The company aims to drive more traffic to external sites, reporting that users engaging with AI overviews tend to visit a wider variety of websites.
• Google is testing new methods for displaying relevant links within the text of AI overviews to improve user experience and source visibility.

https://techcrunch.com/2024/08/16/google-is-bringing-ai-overviews-to-india-brazil-japan-uk-indonesia-and-mexico/…

• AI scraping search dominance by Google creates a dilemma for publishers, who risk losing traffic if they block Google’s AI from summarizing their content.
• A recent US federal court ruling deemed Google an illegal monopoly, intensifying concerns over its unfair advantage in the evolving AI landscape.
• Publishers are caught between the need to partner with Google for visibility and the threat of their content becoming obsolete due to AI summarization.
• Startups seeking to compete with Google face significant barriers, including the high costs of crawling websites and licensing content, which Google does not incur.
• The ongoing antitrust scrutiny of Google may lead to changes that could force the company to share its search data or index with competitors, impacting the future of search and AI.

https://www.thestar.com.my/tech/tech-news/2024/08/16/googles-search-dominance-leaves-sites-little-choice-on-ai-scraping…

• AI scraping search visibility presents a dilemma for online publishers: allow Google to use their content for AI-generated answers or lose search visibility altogether.
• The recent ruling against Google has intensified concerns for publishers already struggling with declining ad revenue and search traffic.
• Blocking Googlebot to protect content means publishers risk losing traffic and potential revenue, while allowing scraping may lead to fewer site visits.
• Google benefits from free training data from publishers, unlike smaller AI startups, which must pay for access to compete.
• The Justice Department’s potential actions against Google could reshape the landscape for online publishers and their relationship with the search engine.

https://www.engadget.com/ai/online-publishers-face-a-dilemma-allow-ai-scraping-from-google-or-lose-search-visibility-202246891.html…

• Google’s upgraded AI image generator, Imagen 3, is now available in the US, offering improved detail, richer lighting, and fewer artifacts than previous models.
• Users can access Imagen 3 through Google’s AI Test Kitchen and edit images by highlighting specific areas and describing desired changes.
• The tool has restrictions, declining to generate images of public figures and weapons, while allowing for creative interpretations of copyrighted characters.
• Reddit users began experimenting with Imagen 3 recently, and Google published a research paper detailing the tool’s capabilities.
• Unlike Grok, the AI image generator on Elon Musk’s X platform, Imagen 3 maintains stricter content guidelines, although it has faced its own challenges with generating accurate historical images.

https://www.theverge.com/2024/8/15/24221218/google-ai-image-generator-imagen-3-available…