Elliot Blanchard

Damn, I’m getting addicted to NotebookLM (Google’s ChatGPT moment). Here, I was feeding my last book “UX for XR” (btw. new edition coming in 2025 !) into the AI model and it generated a pretty decent podcast (note to publisher: It only works with the documents you upload and doesn't use your data to train its AI models. Google has my trust at the moment). There is a bit of fluff, a few witty lines, the conversation stays a bit on the surface, focus topics are kinda picked randomly (here it is the OODA loop). There is one mistake (I wouldn’t classify Adobe Aero as social), nevertheless I’m impressed by the natural sounding voice and basic understanding of the topics in the book. On the covered book contents: Obviously, Spark and the Hololens are on the way out (and Aero has been in beta for the last four years, which is not very assuring). Nevertheless there is still a lot of exciting stuff, new tech, tools and case studies, to pack into the next edition… https://lnkd.in/gwJjPVtJ

12

2 Comments

A very humble contribution of mine in this article from Forbes about Blockade Labs’ release of Blendbox. Here is my full unedited note on the potential of this new breed of tool: “I approach AI art in much the same way as I have approached photoshop over the past 20 years, utilizing techniques like kitbashing and photo-bashing. Blendbox, even in its alpha stage, is a promising step toward integrating AI into creative workflows that digital artists have used for years. The advantage of a tool like Blendbox, particularly as new features roll out, is in retaining control over composition and concept, rather than allowing AI to dictate the entire composition, colors, and other details. My hope and expectation for the future are to see features like image uploads, enabling me to incorporate my own assets and apply my own LoRas to maintain consistency across a body of work, aesthetic style, or even within IP-owned projects. With features such as these, combined with shortcuts and layer management similar to Photoshop, and potentially even 3D file imports, a GenAI-powered software could provide a level of control that Photoshop never achieved. Currently, I generate objects and layers directly on the platform, then import the results into Magnific for latent upscaling, and finally, into Minimax to produce a video. To address your question, Blendbox doesn’t quite compare with tools like Topaz or Magnific. Topaz is simply an upscaler, a pioneer but now largely outdated, while Magnific supports text-to-image, style references, and upscaling but lacks control over composition. In Magnific, control is only as precise as the prompt and the model’s ability to interpret it, which has limitations. For example, if I want a cow on the left, a cloud on the right, a character in the center with a bird on their shoulder, a top hat, a left red shoe, and a right blue shoe, some prompt information will likely be lost or elements misplaced. Blendbox’s real power lies in enabling control over an infinite array of elements, allowing you to position and size them precisely as desired, giving you mastery over the composition. This refined version improves clarity, flow, and readability while keeping the original meaning intact. Let me know if there’s anything else you’d like to adjust!” I invite you to read the full article here: https://lnkd.in/g_zXu6mw

16

1 Comment

This week on "On AI," I welcome back AI aficionado and philosopher, Lev Gorelov, for an insightful conversation. Link to episode: https://lnkd.in/ecYbq2uJ With a particular focus on the recent advancements with Claude III, Lev contrasts Claude III's approach to generations—geared towards creating engaging, personality-driven content—with the more traditional model (#openai 's #chatgpt) that prioritize business-oriented, "buttoned-up" outputs. He also explains the fundamentals of why, using the same training data (the entirety of the internet), the outputs vary. Throughout the episode, Lev shares insights into how these tools are not just about generating content or improving human productivity, but are enhancing the creative process. Specifically: * Giving us practical methods on how to leverage LLM's as collaborative partners that help bridge ideas and refine storytelling. * Emphasizing the importance of using AI tools creatively and responsibly, * Highlighting their role in brainstorming and editing rather than replacing human creativity. Join Beth and Lev as they explore the future of AI in creative professions and education, pondering the broader implications of these technologies on our understanding of originality and creativity (you might want a pen handy to take some notes). #AIPhilosopher #AINinja #Claude3 #BethOnAI #GenerativeAITalks #AIExperience #CreativeAISystems #AIContentCreation #PhilosophyMeetsAI #LevGorolevOnAI #OnAIpodcast #openai #anthropic #ClaudeIIIInsights #CreativeAI #AIInEducation #GenerativeContent #AIandCreativity #StorytellingWithAI #AIForCreatives #FutureOfAI #podcast #podcasting #aipodcasting

5

1 Comment

Procedural Ruins Tutorial Julian Bragagna | In this lesson, learn how to create these alien-looking ruins. Start by learning how to build a procedural tool that transforms a rough blockout into a detailed, high-poly mesh that can be used inside Unreal. These kinds of tools are designed to give you fast iteration times, making it easier to experiment and refine your environments efficiently. https://lnkd.in/ge8PpuGp #Houdini #ProceduralGeneration #GameDevelopment #3DModeling #DigitalArt #VFX #CGI #EnvironmentArt #LevelDesign #GameArt #ProceduralTools #3DAnimation #VisualEffects #GameDesign #Tutorial #SideFX #ProceduralRuins #UnrealArtists #HoudiniArtists

25

Have a look at the new LLaMa 8b #instruct by #Meta, running locally on my 4090 RTX (the video has not been speed-up). I think it's outstanding that it can match and sometimes surpass the previous LLaMa 2 70b, a model which was almost 10 times bigger. And, as you can see...it is FAST! p.s.: I used Ollama to run it, on Windows. https://lnkd.in/dhHyZUQJ #llama #llama3 #meta #nvidia #rtx #llm #generativeAI #genAI

12

To all the AI 'experts', yes, the same people who were last year's web3 'experts' and wherever VC funding lands, next year's 'experts' start using the brain in your meat sack and ask what future you are advocating for. https://lnkd.in/gtxWFQx5

12

2 Comments

Mixed reality has a software and content problem. But Meta knows this. Today, Meta connect announced some great improvements in hopes of easing developer friction and increasing the amount of apps published. I wanted to share some of my favorites below. - Camera Passthrough API This is the big one that many have been asking for. Until recently, developers could not utilize the cameras on device to identify objects when developing for the Horizon platform. This will allow developers to identify objects which could be used to implement new types of anchors for 2d planes or integrate AI alongside object identification for a more cohesive experience in augmented reality. I'd love to know more about the privacy features associated with the new camera Passthrough API as it balances developer access for functionality and user privacy. Still, this access is crucial to making AR applications that can identify and exist within the users space. - 2D app increased support including React native This is going to lead to so much more content available on platform, with developers having the ability to port their existing 2D apps to the Horizon OS platform, I foresee an increase in utility for augmented reality use for productivity. - Progressive web app packaging The increased support for web-based applications and sites will allow for even more ways for developers to enter the ecosystem without friction. The ability to have your web app in store, discoverable while allowing for in-app payments is a great new way for existing sites and developers to create experiences for users. - Lastly DOCUMENTATION ORGANIZATION. This is CRUCIAL. The platform has developed quickly and often, which is great, but has led to deprecated SDKs and new SDKs coexisting within platforms like the Unity asset store. Having the documentation and resources updated regularly within the meta site will be a tremendous help in reducing friction for developers.

5

My cool finding of the day: Roblox is pioneering 4D generative AI, merging the realms of design, physics, and interaction! Essentially letting users generate with AI their 3D assets but also how they will behave in the game! This will unlock the creativity of many people and literally let them build whatever they would like around them in the Roblox "metaverse". Their creations will come to life with a simple command promising a future where anyone can bring ideas to "digital life". #GenerativeAI #Innovation #Gaming https://lnkd.in/gZW8msf6

28

2 Comments

Does anyone else feel like AR has been a buzzword forever, and yet when you try to see actual UX/UI images they're nonexistent? Alex Heath from The Verge finally got a chance to check out Meta's Orion AR glasses, but the description of the interface was sadly lacking. It's not just the physical artifact I want to celebrate- but the UX/UI work that Meta designers put into Orion’s interface. I can't imagine how proud I'd be of working on something like this... #AugmentedReality #ARDesign #UXUI #ProductDesign #DesignInnovation #MetaAR #ARInterface #EmergingTech #UXTrends #DigitalExperience #theverge https://lnkd.in/g_zEXYNZ

2

1 Comment

Check out argodesign's latest POV on platform engineering and the importance of designing for internal users to deliver great products for end-users. The POV features real-world examples, including DreamWorks, a multi-year program that reimagined the film production pipeline. If you're interested in #platformengineering, #uxgovernance, #designops, and #devops, this read is for you!

13

Hello fellow designers and creatives. I’ve decided to jump back into writing. Find me on Substack where my musings will focus on industrial design, design, strategy, sustainability, and systems thinking. Thanks for the sub. :) https://lnkd.in/e4Wpfbrp #industrialdesign #designstrategy #systemsthinking #sustainability

3

This seems to be a better breakdown of the GPT-4o features vs. the demo earlier: https://lnkd.in/eFZUxHK5 Lots of really interesting stuff in here.

2

Josh Clark's article raises some interesting points. We are in an era of rapid growth. Impossble tech becomes reality every day. How do we classify things in this Cambrian explosion of AI creatures? Based on the model built by Matt Webb, he suggests a triangular venn diagram of 3 things: - Grounded - does what it's supposed to - Interoperable - plays well with the other systems - Radically adaptive - changes as you use it as you need it to https://lnkd.in/dth_n_dZ

4

1 Comment

In this segment of Masters of AI, I drill down with Ed Rhine from Spillt to explore how AI transforms creative projects' ideation and pre-production stages. We dive into the power of using AI to generate ideas faster, refine concepts, and ultimately get to that crucial client "green light" quicker than ever before. Ed and I discuss how this approach speeds up the process and adds value by allowing teams to iterate quickly and bring clients on board sooner. It’s all about harnessing AI to streamline the creative journey from idea to approval, unlocking new efficiencies and creative possibilities. #ai #greenlight #ideation

3

What emerging trends in deep learning are you most excited about? As we delve deeper into the future of AI, innovations like transformers, attention mechanisms, and beyond are reshaping our understanding and capabilities. Transformers have revolutionized natural language processing with their ability to handle vast amounts of data and complex dependencies. Attention mechanisms have enhanced model performance by enabling them to focus on the most relevant parts of input sequences. These advancements, among others, are paving the way for unprecedented breakthroughs in AI research and applications. As we continue to push the boundaries of AI research, flashAttention 3 stands out as a groundbreaking advancement. This innovative mechanism enhances the efficiency and accuracy of attention models, significantly reducing computational overhead while maintaining high performance. flashAttention 3’s ability to streamline processing without sacrificing quality opens new avenues for complex AI applications, from natural language processing to real-time data analysis. By optimizing resource utilization, it paves the way for more scalable and accessible AI solutions. What trends do you find most promising? Share your thoughts and join the discussion! #DeepLearning #FutureTrends #AIResearch #Transformers #AttentionMechanisms #Innovation #AcademicResearch #flashAttention3

1

I used AI in combo with my drawings to create this reel. I dont consider this animation or film making but some do lol. Question...Can one call themselves an animator or a film maker using generative AI prompts?

4