Adobe recently released Firefly 3, their new AI image generation model, which has improved in terms of realism. While the images generated by Firefly 3 are still not quite on par with those from models like Midjourney or Stable Diffusion 3, the new model is a step in the right direction.

The more exciting announcement from Adobe, however, is the integration of Firefly 3 and other new AI features directly into the Photoshop beta. These features include:

• The ability to generate entire images, not just do inpainting and outpainting
• A "generate similar" option to create variations of an image
• An adjustment brush
• The ability to generate backgrounds for e-commerce shots
• An "enhanced detail" feature

While Firefly 3 still feels like it's trying to catch up to the competition, the fact that these powerful AI tools are now available within Photoshop is a significant upgrade. Users can now create new images, refine existing ones, and leverage AI-powered features directly within the familiar Photoshop interface.

Adobe announced a new research project called Video Giga Gan, which focuses on video upscaling. This technology can take low-quality, blurry videos and significantly improve their quality, making them clearer and more detailed.

Some examples showcased include:

• A stack of pancakes being drizzled with sauce, transforming a low-res, blurry video into a high-quality, clear one.
• A blurry waterfall becoming a crisp, high-definition video.
• A man with a beard, where the low-quality original is upgraded to a clear, high-res version.

This Video Giga Gan research is not yet publicly available, but Adobe is likely to integrate it into future versions of their video editing software, such as Adobe Premiere. The detailed research paper is available for those interested in learning more about the technical aspects of this video upscaling technology.

Midjourney, the popular AI image generation tool, has recently added a new feature that allows users to generate images with random styles. This feature was showcased by Ally JWS, a prominent Midjourney user, who demonstrated the capabilities.

To use this feature, users can simply add the --srf random syntax at the end of their prompt. This will generate the same image concept with a variety of random styles applied. Ally shared several examples, showcasing how the same prompt of "a young woman with freckles wearing a yellow hat closeup photo" can result in drastically different stylized outputs when the --srf random option is used.

This new feature provides Midjourney users with even more creative possibilities, allowing them to explore a wide range of artistic styles for a single image concept. The ability to quickly generate multiple variations with random styles can be a valuable tool for artists, designers, and anyone looking to experiment with unique visual aesthetics.

Blockade Labs, a popular 360° image creation app, recently received a significant upgrade. The app now generates images in stunning 8K quality.

We can see some impressive examples of the new 8K capabilities. When viewed in full-screen mode, the images showcase an incredible level of detail and clarity. Users can freely navigate and explore the 360° scenes, zooming in and out without any loss of quality.

The prompt that generated this high-resolution image is displayed on the screen. Feel free to pause the video and take a closer look at the prompt if you're interested.

This new 8K model from Blockade Labs represents a significant advancement in the world of 360° image generation. The level of detail and immersion provided by these 8K images is truly remarkable, offering users a captivating and visually stunning experience.

The Rabbit R1 is a new handheld device that uses voice commands to access a large language model. Some key features of the Rabbit R1 include:

• Voice-controlled commands: Users can press a button on the device and issue voice commands to interact with the AI assistant.
• Quick response times: The Rabbit R1 is reported to be faster than the Humanoid AI Pen, with responses coming within 2-3 seconds.
• Image generation: The device can generate images directly within the Midjourney AI model.
• Productivity features: It can perform tasks like ordering Uber rides, controlling Spotify, and taking audio notes that are then summarized.

While initial reviews from AI enthusiasts have been positive, the author is waiting for more mainstream reviews to get a balanced perspective on the Rabbit R1's capabilities and usability. The author plans to share their own thoughts on the device once they receive a unit, likely in the coming months.

Anyone can have a website online within minutes with Hostinger's AI website builder. Simply head over to Hostinger.com/mtWolf, click on "Claim Deal", and select the "Business Website Builder" plan to access the powerful AI tools.

To create a new website, start by entering a brand name, such as "Matt Wolf's Blog". Choose "Blog" as the website type, and use the microphone to describe the site as "The personal blog of Matt Wolf where he shares tips, tricks, and tutorials on how to effectively use AI in your business."

Next, select a personalization style - "Minimal" is a great choice for a clean, elegant look. Hostinger's AI will then generate a fully-designed website for you, complete with blog posts and images.

All you need to do is customize the content to fit your needs. Hostinger's AI-powered website builder makes it easy to get your online presence up and running in no time. Head to Hostinger.com/mtWolf, use the coupon code "Matt Wolf" for an additional 10% discount, and start building your AI-powered website today.

Pretty much all of the big AI companies this week came together to essentially commit to prevent the use of AI in sexual harm against children. This is really one of the first times these companies have all come together and agreed on something.

The companies involved include OpenAI, Amazon, Anthropic, Anthropic, Google, Meta, Anthropic, Microsoft, Mistral AI, and Stability AI. This collective effort aims to proactively address child safety risks as these companies develop and build their AI models.

While the details are still being worked out, this commitment represents an important step in the AI industry taking responsibility for potential misuse of their technologies. By aligning on this critical issue, these leading AI firms are demonstrating a shared priority to protect the most vulnerable.

The International Olympic Committee has launched an Olympic AI agenda, aiming to leverage AI technology across various aspects of the Olympic Games. Some key areas where AI will be utilized include:

• Talent identification and personalized training: AI can help identify athletes and talent around the world, and provide more athletes with access to personalized training methods, superior sports equipment, and individualized programs.

• Judging and refereeing: AI can be used to improve fairness in sports by revolutionizing judging and refereeing.

• Safeguarding in sport: AI will be leveraged to improve safeguarding in sport.

• Event organization and broadcasting: AI will make organizing sporting events more efficient and transform sports broadcasting, making the spectator experience more individualized and immersive.

The Olympic Committee believes AI can play a transformative role in the future of sports, from talent development to event operations and fan engagement. This marks a significant step in the integration of AI technology across the Olympic ecosystem.

The people over at TED put out a video this week generated by Sora, showing what TED will look like in 40 years. While the audio is muted to avoid potential copyright issues, the video provides a glimpse into TED's vision of the future powered by AI.

The video starts with a futuristic TED stage and gradually transitions to showcase various AI-driven advancements. It eventually zooms in on the TED logo, which is the only element in the video that was not generated using AI.

If you're interested in watching the full 1.5-minute video, you can find the link in the description below.

The Rayban Meta smart glasses have received an upgrade this week. They now have the ability to see what you're looking at and connect to a large language model to talk about what you're looking at. There's also a new feature to broadcast what you're looking at over WhatsApp or Messenger.

In a demonstration video, we can see the user saying "Hey Meta, play some music", and the glasses start playing music. The person they're talking to can also see exactly what the user is looking at through the video calling feature, which only works over WhatsApp and Messenger, not directly through FaceTime or similar.

The user then tries out the feature by looking around and asking "Hey Meta, what am I looking at right now?" The glasses respond that they are looking at a desk with various electronic devices including monitors and speakers.

Overall, these new features allow for more interaction and sharing of the user's visual experience through the smart glasses, leveraging the connected language model to provide information about what the user is seeing.

During the Tesla earning calls this week, Elon Musk claimed that Tesla is not a car manufacturer, but rather an AI company. Musk stated:

"We should be thought of as an AI robotics Company. If you value Tesla as just an auto company, it's just the wrong framework. If you ask the wrong question, then the right answer is impossible. If somebody doesn't believe that Tesla is going to solve autonomy, I think they should not be an investor in the company."

An image has also been circulating that depicts the Tesla ecosystem, with humanoid robots and AI compute at the very top of the hierarchy, suggesting that Musk sees these as the core focus of the company, rather than just vehicle manufacturing.

Musk's comments and the visual representation indicate that he wants Tesla to be viewed as an AI-driven robotics company, rather than a traditional automaker. This shift in perspective aligns with Musk's long-standing interest in artificial intelligence and his belief that Tesla's future lies in autonomous driving and advanced AI capabilities.

This article came out as the creator was pressing the record button. It states that Musk's xAI is close to raising $6 billion from Sequoia and others, valuing the company at $18 billion.

The assumption is that this money will go towards more compute power to train larger and more powerful AI models even faster. However, the article does not explicitly state how the funds will be used.

The creation of xAI, a one-year-old AI company founded by Elon Musk, highlights Musk's focus on advancing AI technology. With this $6 billion raise, xAI is now valued at $18 billion, underscoring the significant investment and interest in Musk's AI endeavors.

In Los Angeles, the Metro is installing AI-powered cameras on buses to issue tickets to illegally parked vehicles. While the goal is to address the issue of vehicles parked in bus lanes, which can make it harder for people with disabilities to get on and off the buses, there are concerns that this move may be seen by the public as a massive overreach of government and a step towards a "Big Brother" surveillance state. The use of AI cameras to automatically issue parking tickets is a controversial decision that may face backlash from the general public.

Microsoft has launched a new small language model called F3. This 3.8 billion parameter model is designed to run directly on mobile devices, providing faster response times compared to larger models that require cloud processing.

The key benefits of F3 include:

• Cheaper to run than larger models
• Performs better on personal devices like phones and laptops
• Suitable for quickly looking up basic information

F3 is available on Microsoft Azure, Hugging Face, and Olama, allowing developers to experiment and integrate the model into their applications. The release of smaller language models is a trend as companies aim to bring more language model inference directly to end-user devices, reducing latency and cloud dependency.

In some random AI news, the rapper Drake created an AI diss track. At least, that's what everyone is saying, though Drake himself has not confirmed it. The interesting thing about this story is that Drake has previously talked about not being a fan of AI-generated music, yet the rumors suggest he is now making his own AI versions of songs to create these diss tracks.

The estate of Tupac is now threatening to sue Drake if he doesn't take down the song and remove all traces of it. The song reportedly uses AI voices of Tupac and Snoop Dog. This raises questions of hypocrisy, as Drake has criticized others for making AI music, yet he may be doing the same thing himself.

The use of AI to generate music, especially for the purpose of diss tracks, is a controversial and complex issue. It highlights the ethical concerns around the use of AI technology in the music industry and the potential for abuse or misuse. As AI capabilities continue to advance, these types of situations are likely to become more common, and the industry will need to grapple with the implications.

Synthesia, a company known for its text-to-AI avatar technology, has recently showcased their latest model called Express One. What sets this new model apart is its ability to display more emotion and natural body language based on the context of the spoken text.

In a 10-second clip, the AI avatar demonstrates a range of expressions, from happiness to frustration, as it speaks the corresponding text. The facial expressions and body language dynamically change to match the emotional tone of the words being said, creating a more lifelike and engaging presentation.

Synthesia offers a selection of avatars that can be utilized with this new expressive capability. Additionally, users have the option to create their own custom avatars, although this process requires a green screen setup and recording at certified studios to achieve the best results.

While the basic usage of the Synthesia avatars is free, the custom avatar feature is available only through their paid plans.

Sanctuary AI, a company I discussed in a previous video about AI robots, has just released a new and improved humanoid robot. This new robot features several key upgrades:

• Improved commissioning and build speed
• Reduced material costs
• Expanded range of motion
• Enhanced hand durability
• Upgraded tactile sensors

These improvements across various aspects of the robot's design and capabilities demonstrate Sanctuary AI's commitment to advancing humanoid robotics technology. The company seems to be making steady progress in creating more capable and cost-effective robotic platforms.

While I didn't cover this specific announcement in my previous AI robots video, I'm glad to see Sanctuary AI continuing to innovate in this space. Robotics, along with AI, are areas I'm particularly passionate about, so I always enjoy seeing the latest advancements from companies working in this field.

Meta (formerly Facebook) has taken an interesting step this week by opening up its mixed reality ecosystem. The company has made its Horizon OS available for other companies to use in their virtual reality (VR) headsets, beyond just the Meta Quest devices.

This move allows the likes of Asus Republic of Gamers, Lenovo, and Xbox to adopt the Horizon OS for their own VR headsets. This is a strategic play by Meta, as it aims to expand the reach of its Horizon Store - the VR app marketplace similar to the Apple App Store.

By making Horizon OS available across multiple VR headsets, Meta is positioning its store to become the go-to destination for VR apps, much like how the App Store dominates the mobile app ecosystem. This could help Meta generate more revenue from the sale of VR apps and experiences, regardless of the hardware being used.

The move also highlights Meta's ambition to establish its Horizon platform as the de facto standard for mixed reality computing, similar to how Android and iOS have become the dominant mobile operating systems. By opening up the ecosystem, Meta is hoping to attract more developers and users to its Horizon platform.

Overall, this decision by Meta demonstrates its desire to exert more control over the emerging mixed reality landscape, leveraging its existing Horizon ecosystem to gain a competitive advantage over other VR platforms and hardware manufacturers.

This robot dog with a flamethrower attachment is a concerning development. Equipping robots with potentially dangerous tools like flamethrowers raises serious safety and ethical concerns. While the creators claim it could be used for wildfire prevention or snow removal, the risks of such a device seem to outweigh any potential benefits. Putting powerful tools of destruction on autonomous robots is a troubling step towards a future many would find undesirable. Careful consideration of the implications and stringent safety protocols are essential before deploying such technology.