#3 The time where you can have a conversation with your dog has come
In this week’s edition of our bi-weekly innovation newsletter: a British talking robot dog, new ways to create NeRF’s (neural radiance fields) with gaussian splat, the reason why Google bought Anthropic and a long read about how we use Unity when creating seamless 3D branded games in your web-browser.
Spot: the British talking robot dog
We’ll kick this bi-weekly blog off with a talking dog, yes this is for real and not CGI. To be specific, it’s a robot dog from a company called Boston Dynamics. This company already has a pretty crazy track record when it comes to uploading viral videos with their very own developed robots. They have created dancing robots and robots that can complete obstacle courses as a fully skilled parkour-master. And now they have combined the best of generative AI language models with their own robot algorithms, creating a posh English sounding robot-dog named Spot. Check out the video below to see what all the fuss is about.
The way they’ve entered the voice ai into the robot is pretty cool. First they use a speech-to-text model like Whisper to decode anything the handler is saying to the robot. After that it is loaded as a prompt into a language model like GPT4. Using the prompt response as input for a text to speech model likely from Eleven Labs.
Gaussian splatting
The past half year showed us a big uprise in developments in the world of neural radiance fields. Which are basically a new way to interpret videos and/or photos into a 3D model. Instead of having to lidar-scan the object you want to have in 3D, you can just upload 10 photos of the object into a NeRF AI model and it will create a 3D representation of it. The latest advancement in this field is the use of Gaussian splats. Traditional NeRF techniques often struggled with fine details and edges in the 3D representations, but Gaussian splats have significantly improved the situation. Essentially, they allow for a more accurate placement and blending of the light points (radiance fields), leading to sharper images and a more realistic 3D model. This is particularly useful in industries like architecture and game design, where precision is key.
Gaussian splatting is being developed and adopted by AI software companies like Luma AI and Polycam. Check out the video below to learn more about how Polycam provides gaussian splatting.
Google's Acquisition of Anthropic
In a significant development, Google has invested a whopping 2 billion dollars in Anthropic, an AI safety and research company. This move isn't just about expanding its AI capabilities; it's a strategic play to keep up with the intensifying competition in the AI space, especially against formidable rivals like OpenAI, which has received substantial backing from Microsoft. Google's investment in Anthropic is indicative of the high stakes in the race to lead in AI technology.
Why Anthropic? Well, it's not just any AI company. Anthropic is known for its focus on AI safety and alignment research, areas that are becoming increasingly crucial as AI systems grow more powerful and pervasive. By ensuring that AI systems align with human values and operate safely, Google is positioning itself not only as a leader in AI capabilities but also in responsible AI development.
Exploring the 3D world of Unity with branded games
Digital experiences are becoming increasingly beautiful, better, and more spectacular. While 2D campaigns can be very appealing in themselves, the experience truly becomes immersive when we add an extra dimension. An essential tool we use at LiveWall to create 3D applications is Unity. One of our gems? Without a doubt, the advent project for Rituals.
Read the full blog on our website via this link: https://www.livewall.co/blogs/exploring-the-3d-world-of-unity-with-branded-games
Stay tuned for our next edition, where we'll continue to explore the cutting-edge developments reshaping our world. Until then, keep innovating!