HOLLYWOOD

🎭 Hollywood talent agency explores AI clones for actors

The Creative Artists Agency (CAA), one of Hollywood's leading talent agencies, is exploring the possibilities offered by artificial intelligence through an initiative called CAA Vault. This project allows high-caliber stars to create digital duplicates of themselves, thus exploring new frontiers in digital entertainment and virtual events without the need for the physical presence of the actors.

This move is a response to concerns about the impact of automation and AI on jobs and the creative process in Hollywood. CAA is seeking ways for artists to benefit from these emerging technologies while adapting to industry changes. The idea of digital twins could potentially revolutionize how content is produced and consumed, offering new income and exposure opportunities for talents represented by the agency.

APPLE

📱 Apple steps up negotiations with OpenAI for generative AI features on iPhone

Apple is intensifying discussions with OpenAI to potentially integrate generative artificial intelligence features into its iPhone devices. The negotiations involve the use of OpenAI's AI technology to power new functionalities in iPhone software, possibly starting from iOS 18. This collaboration aims to incorporate functions that allow the creation of images and texts from simple commands, reflecting Apple's significant effort to expand its generative AI capabilities.

While the final specifications and terms of the agreement between Apple and OpenAI have not yet been decided, more details are expected to be revealed during Apple's annual developers conference in June. This partnership highlights Apple's strategy to strengthen its AI offerings in response to the rapid advancements of competitors like Google and Microsoft in the AI technology market.

META + RAY-BAN

🕶️ Meta and Ray-Ban launch new generation of smart glasses

Meta, in partnership with EssilorLuxottica, announced the latest collection of Ray-Ban Meta smart glasses. Completely redesigned from the ground up, these glasses enhance the core features of the first generation and introduce new capabilities. They allow for taking photos and videos, listening to music, making calls, and even livestreaming directly to Facebook and Instagram, all without the need for a smartphone. These glasses are available starting at $299.

In addition to technological advancements, the glasses incorporate the iconic design of Ray-Ban and are available in a limited edition in the exclusive Scuderia Ferrari color for the Miami 2024 event. Meta also highlighted the use of its new AI, built with Llama 3, which facilitates interaction and task execution through the glasses. This launch represents a significant step for Meta in integrating fashion, technology, and AI functionalities.

MICROSOFT

🧠 Microsoft introduces Phi-3: a powerful small language model.

Microsoft recently unveiled Phi-3, a new family of open-source artificial intelligence models developed to enhance the capabilities of small language models (SLMs). Phi-3 promises to be the most capable SLM model currently available, surpassing other models of similar sizes and even larger ones in language, reasoning, programming, and mathematics benchmarks.

The initial model in this series, Phi-3-mini, is already available and stands out for offering context lengths of 4K and 128K tokens. This means it can handle extensive context windows without losing quality, which is a novelty for models of this class. Additionally, Phi-3-mini has been optimized for practical use in generative AI applications, being available on various platforms such as Azure AI, Hugging Face, and Ollama. It is expected that more models from this family will be released, offering even more options for users.

ADOBE

🎨 Adobe launches the new Firefly Image 3 AI model

Adobe has revealed its latest innovation in generative artificial intelligence, the Firefly Image 3 model, aimed at empowering creativity across various fields. The new model is designed to be commercially safe, ensuring that the generated content respects copyrights and other intellectual properties. This advancement is part of Adobe's effort to develop AI responsibly.

Firefly Image 3 comes equipped with enhanced features that allow users to apply the structure of an existing image to new creations, making it easier to generate variations of images with the same layout. This capability is a significant advancement, eliminating the trial-and-error process in creating prompts to achieve desired results. Additionally, Adobe has integrated Firefly natively into its Creative Cloud applications, such as Photoshop and Illustrator, aiming to enhance creative workflows and provide new opportunities for creators to explore and monetize their talents.

SOCIETY

🧠 Artificial intelligence can predict political orientations through faces

A recent study revealed that facial recognition algorithms can predict individuals' political orientations with considerable accuracy, just by analyzing their photos. Conducted by Stanford University, the study used over one million face images collected from dating websites in the US, Canada, and the UK, as well as Facebook users in the United States. These individuals had identified their political leanings as part of website questionnaires, allowing the artificial intelligence to study statistical differences between conservatives and liberals.

Despite facial recognition techniques being relatively common, the study avoided simplifications of attributing political characteristics to obvious demographic traits such as age or skin color. Instead, the algorithm was trained to make subtle distinctions among individuals of the same age group, gender, and ethnicity, achieving accuracy of up to 73% in its predictions. This type of research raises serious privacy concerns, as seemingly neutral characteristics can inadvertently reveal sensitive personal information, which could be misused in oppressive regimes or other delicate situations.

RESEARCH

🧬 Profluent launches first open-source, AI-created gene editor

Profluent, an innovative protein design company, has launched OpenCRISPR-1, standing out as the first AI-created gene editor made available as open-source code. This tool, which integrates a Cas9-like protein and guide RNA, was developed using large-scale language models, which learned from a vast amount of sequences and biological contexts to generate CRISPR-like proteins that do not occur naturally.

The initiative aims to democratize access to gene editing, offering this technology at no cost for ethical research and commercial use. In addition to expanding the known families of CRISPR, Profluent hopes that this AI-based approach can facilitate the development of personalized and more accessible genetic treatments, accelerating the creation of cures for thousands of currently incurable diseases. This approach represents a significant step in personalized medicine and could transform the treatment of genetic diseases such as sickle cell anemia.

• 💰 Canada allocates $50M from federal budget to address the challenge of "job-stealing robots." The fund aims to mitigate the effects of artificial intelligence on the workforce by promoting adaptation and the creation of new technological jobs. Learn more

• 📜 Amazon faces accusations from a former AI executive for allegedly encouraging copyright infringement and discrimination. Dr. Ghaderi, an engineer who worked with large-scale language models at Amazon, claims she was instructed to ignore internal copyright policies to improve Alexa's search results, despite her objections and internal legal advice. Learn more.

• ✈️ United Airlines is using artificial intelligence to enhance the flight experience, making the "skies friendlier." The initiative includes delay predictions, frequent passenger recognition, and even baggage optimization. Learn more.

• 🎬Deepfakes of Bollywood stars criticizing Prime Minister Narendra Modi and endorsing opposition parties circulated, raising concerns about AI interference in elections. This practice underscores the need for regulation and checks to combat AI-driven misinformation. Learn more.

• 🧠 Using AI to analyze the gut microbiome, researchers have discovered new connections to Alzheimer's disease. The technology allowed the identification of specific patterns linking changes in the microbiota to cognitive deficits, opening avenues for future therapies based on microbiome modifications. Learn more.

• 🎨 Blockade Labs has improved AI-generated 3D art for 360-degree applications, achieving 8K resolution. Ideal for use with Apple Vision Pro, the Skybox AI update promises to revolutionize digital environment creation, expanding applicability in entertainment and education. Learn more.

• 🌍 Groq and Earth Wind & Power launch an AI computing center in Norway, promising global-scale innovations. The center will focus on AI inference computing, offering efficient solutions in compliance with GDPR for European developers and government entities. Learn more.

• 🌐 Worldcoin, a cryptocurrency project that uses iris scanning, co-founded by OpenAI CEO Sam Altman, is considering a partnership with OpenAI. This collaboration aims to integrate more artificial intelligence technologies into the digital identification process, increasing both security and efficiency of cryptocurrency distribution. Learn more.

• 📹 Adobe revolutionizes video quality with its new AI technology, VideoGigaGAN, capable of transforming blurry videos into images up to 8 times sharper. The innovation promises to be a valuable tool for restoring historical footage and improving low-resolution video content, although there is no defined date for its public release. Learn more.

• 📈 Microsoft announces impressive results for the quarter, with revenue of $61.9 billion and a 17% increase over the previous year, driven by high demand for AI in the cloud. Learn more.

• 🌐 Huawei intensifies competition with Nvidia in developing artificial intelligence chips, significantly advancing with its new Ascend 910B chip. This move highlights Huawei's growing capability as a competitor in the semiconductor industry, challenging Nvidia's dominance amid trade tensions between the US and China. Learn more.

• 🐕 Throwflame introduced the Thermonator, a flame-throwing robotic dog designed for various applications, including fire control and agricultural management. Equipped with advanced technology, it can be operated remotely and is legally unregulated in the US, being sold for $9,420 with free shipping. Learn more.

The video of the week features Mustafa Suleyman's TED Talk, co-founder of DeepMind and current CEO of Microsoft AI. Mustafa argues that artificial intelligence should be understood as a new digital species, with the potential to become complex and integrated into human society, requiring care and planning to mitigate risks and maximize benefits. Click here or on the image above to watch.

• 🎭 Synthesia Avatars: AI video creation tool that allows users to turn text into narrated videos, using over 160 avatars with emotion capabilities. The tool supports video creation in over 120 languages, ideal for applications in corporate training, marketing, and entertainment. Link.

• 🖌️ Adobe Firefly: AI generative image creation tool. Embedded in Adobe Photoshop (beta), Firefly allows users to add, remove, and expand content, creating realistic images from scratch. Specifically designed for cloud use, this tool is available to Creative Cloud users. Link.

• 📝 Sonnet AI: complete assistant for meetings and notes. This platform offers audio recording, automatic meeting entry notifications, transcription, custom notes, action items, and CRM auto-update. Link.

• 💡 Open Agent Studio: tool that allows the creation of no-code co-pilots. This platform enables solutions impossible in other tools to become accessible, offering users an early advantage in markets yet to be explored by AI. Link.

🌄 Create images from a reference

In this tutorial, you will learn to generate new images based on a reference image, using Adobe Firefly for free. Step by step:

• Go to Adobe Firefly (you need to create an account).

• Click Generate to create any image and enter edit mode.

• In edit mode, upload the image that will be used as a reference in the Structure section on the left side.

• With the reference image loaded, create a prompt and click the blue button to generate.

• You can adjust the Intensity control on the left side to define how much the generated image should follow the reference. Firefly will create 4 examples of images for you to choose from.