leonardogiusti

At first glance, it looks like activity recognition. But look closer: the model isn't just detecting "typing" but spontaneously distinguishing right-hand from left-hand keystrokes. We didn't train Newton, our foundation model, to do this. No explicit labels. No supervision. @PhysicalAI #PhysicalAI

Interesting result from the team: Newton's embeddings are spontaneously organizing gestures such as “holding”, “shaking” and “smooth movement”. The model finds natural groupings in the high-dimensional embedding space. No explicit labels, no supervision.

Newton's embeddings spontaneously organizing more physical activities: notice how "walking", "running" and "jumping" cluster in different spots in the embedding space (bottom left). #PhysicalAI

It was a great conversation with @sfelc_ about what it takes to think big.

How will AI revolutionize our interactions with devices? Mark Wilson @ctrlzee argues that we've been too focused on gadget assistants "at the expense of imagining the experience AI can create." He cites our work as one of these potential experiences: "Archetype believes it can build something that feels less like a friend and more like real-time intuition [...] perhaps foreseeing an accident on a job site as easily as Google currently auto-completes a sentence." It's a great read, as always! fastcompany.com/91231902/the-g…

1/ Our large behavior model, Newton, creates a detailed narrative of a real-world scenario using only a microphone and radar — no camera required! By fusing real-time sensor data with contextual information, it can figure out if someone may need help and offer recommendations. 🧵

1/ Imagine understanding real-world events just by fusing data from two simple sensors - no cameras needed. Here's how our AI model Newton interprets complex situations through sound and radar to keep homes safer.

1/ We're hiring for 3 roles: 🔹 Principal Software Engineer 🔹 Senior Applied AI Researcher 🔹 Staff Technical Program Manager Join us in revolutionizing how AI understands and interacts with the physical world.

“Working with sensor data that’s already available, you can pull off some amazing feats, translating those signals to track people moving through a room, noting whether they are standing or sitting. Archetype believes it can build something that feels less like a friend and more like real-time intuition, possibly even capable of predicting waveforms that come next." @ctrlzee

Have you ever wondered if the universe has a secret language? What if we could crack that code with AI? Today at @TEDAI2024, we’re introducing Newton — an AI model that can learn how the world works on its own, just by observing it.

Top stories in AI today: -Newton AI learns physics from scratch -NotebookLM launches business pilot -Automate web data extraction with Zapier -Worldcoin unveils next-gen eye scanner -5 new AI tools & 4 new AI jobs Read more: therundown.ai/p/physics-new-…

Archetype AI published a new paper demonstrating how an AI model can encode and predict physical behaviours. Newton, the startup's ‘Large Behavior Model’, can learn complex physics principles directly from raw sensor data, without any human guidance.

"AI will be like water or electricity, ubiquitous. We won't think of a company as an AI company, just like we don't think of a company as an Internet company." @Bill_Gross Thank you to the @Infineon team—OktoberTech™ Silicon Valley 2024 was a blast!

1/ We are excited to share a milestone in our journey toward developing a physical AI foundation model. In a recent paper by the Archetype AI team, "A Phenomenological AI Foundation Model for Physical Signals," we demonstrate how an AI foundation model can effectively encode and predict physical behaviors it has never encountered before, without being explicitly taught underlying physical principles.

Archetype AI’s Newton model learns physics from raw data—without any help from humans venturebeat.com/ai/archetype-a…

We're excited to join @TEDAI2024 to share our vision for the future of Physical AI. Don't miss out on a panel discussion featuring @DrJimFan, @songyeeyoon, and Archetype AI CEO @ipoupyrev! RSVP here: tedai-sanfrancisco.ted.com

We were selected for the new AWS Generative AI Accelerator in a highly competitive process. Archetype AI was one of 80 companies, which is less than 2%, that were selected out of several thousand applicants. So, I am going to Seattle for the kickoff next week. 👇

1/ "With AI, we're headed towards the emergence of something that we are all struggling to describe, and yet we cannot control what we don't understand," says @mustafasuleyman. Here is why metaphors for AI interaction matter:🧵

1/ How can we create seamless interactions between humans and AI technology? In his recent talk for @DomusAcademy, Leonardo Giusti, Archetype AI Co-Founder and Head of Design, shared his vision for the future of Physical AI.

1/ Research spotlight: When AI becomes part of our environment, how do we interact with it? In 2022, @PhysicalAI team members @leonardogiusti, @lbedal2, @ipoupyrev co-authored the paper "Fields: Towards Socially Intelligent Spatial Computing." Here are their top insights: