Kevin Shatzkamer’s Post

📢 At Google I/O 2024, we once again shared a glimpse into the future of AI. Our DeepMind team has been hard at work on Project Astra, a groundbreaking initiative that aims to revolutionize how AI agents interact with us and our world. Project Astra aims to develop AI that goes beyond simple responses, striving to understand context, learn from each interaction, and even anticipate what you might need. I believe this is just the beginning of what's possible. Check out the latest announcements, including Project Astra, from Google I/O 2024: https://lnkd.in/dxuxF6hw Stay tuned for more updates as we continue to push the boundaries of AI. #GoogleIO #ProjectAstra #AI #Innovation

This is a big week in new #AI capabilities. Google is launching: ◼ Veo #generativeAI text-to-video. ◼ Project Astra AI Assistant: understands code, objects, and has a memory of what it's seen. Fairly natural real-time voice interaction. ◼ Circle to Search helps learn to solve physics problems, for instance (Only on Android). And much more!

So-so many things announced today at #GoogleIO made me proud of how at Google, we’re making AI helpful for everyone by improving knowledge, learning, creativity & productivity with our products! Veo (text-to-video models), Ask Photos and Project Astra (our universal AI agent) are probably my most favorite announcements, though it is really hard to choose across so many fantastic new releases when the hype is real! The combination of multimodality, long context and agents will transform how we interact with computing. AI-powered future, bring it on!

📢Thrilled to witness the unveiling of Project Astra at Google I/O today! This groundbreaking initiative represents Google's ambitious vision for the next generation of AI assistants, aiming to revolutionize how we interact with technology. Key highlights of Project Astra: 🚀Advanced multimodal understanding: Astra seamlessly combines visual and auditory perception, enabling a deeper comprehension of our world. 💡 Real-time responsiveness: Experience natural, fluid conversations with unprecedented speed and accuracy. 🤝 Proactive and adaptable: Astra anticipates your needs and learns from your interactions, evolving to become your personalized companion. 🤖 Grounded in real-world context: Astra leverages visual cues and environmental awareness to provide relevant and actionable assistance. #GoogleIO #ProjectAstra #ArtificialIntelligence #Innovation #FutureTech #Assistant

I/O was incredible, and it was particularly amazing to see Google DeepMind's core technologies integrated into so many rich ways across Google products. Some highlights of the things I'm really excited about: Astra - Gemini based models progressively becoming full real-time agents with human-like I/O (audio and visual) & memory. Mind blowing. Gemini - continuing to scale, particularly on context window, and size/inference speed. Context is a key building block of intelligence, as it enables systems to have a working memory. 2M tokens of input context is getting large. And it's only going to get larger :). Gemma - we just announced V2, a 27B model that will come out in June. Our team is hard at work there. We're picking model sizes that are great, and easy to deploy, fit in 1 TPU or equivalent, and will still have unbelievable perf. Search in visual content - back when I started MadBits and brought it to Twitter, we had developed one of the very early image-to-text neural nets, that was captioning visual content so it could be searched. Fast forward to today, these models got incredibly better, and can describe visual content to a degree that's hard to comprehend! The demo of finding your license plate from your collection of Google photos was awesome. Agents - there were early demos of agents doing tasks on your behalf, integrated in Google Workspace. These are just beginning, and going to get really useful for actual day to day tasks. Developers - in the dev conference in the afternoon, the number of integrations of both Gemini and Gemma into IDEs / toolkits and even Chrome! were amazing to see. I wish i had these tools when I started coding. And back to building... :).

In the future AI agents will buy media through APIs instead of humans through UIs and they'll do it better than humans ever could. This is what Daypart is building.

Proud to have been selected as one of the Top 15 companies for Telus' #StandWithOwners initiative. Telus has been a tremendous partner to our growth and we're excited to share their ad for MoveMate! Moving soon? Here's why you'll love MoveMate:   🎯 The price you see is the price you pay. No more estimates, hidden fees, or surprising invoices. 🙅♂️ No upfront charges. Book your move now and secure your preferred date and time at no cost. 🤳A stress-free move at your fingertips. No phone calls, emails and endless back-and-forths. Submit all of the information for your move and request additional services on the same platform. 💆♀️ Book with peace of mind. Moving day can be filled with unforeseen challenges and last-minute changes. Modify or add any detail to your move at any time with no penalty fees. 🌟 Movers you can trust. Forget cheap movers who show up late and break your most precious belongings. Our movers are vetted and trained to ensure you have a smooth moving experience. 🚚 Like Telus, MoveMate is Canada-wide 🇨🇦 Book your move today: www.movemate.ca

To be most flexible & highly responsible in the AI world of tomorrow you need a multimodality approach to combine the dots... Just have a look how Google - Gemini get benchmarked in the areas: text, code, audio, image and video.. https://lnkd.in/gAPDws8T

Google introduces Lumiere, a revolutionary video AI model using STUNET for fluid motion and 80-frame creation. Read more, https://lnkd.in/dtMG9Q4k #bluelupin #Google #LumiereAI #2024tech

2-minute video (thank you Dovid Schick for sharing): Google DeepMind posted this astonishing demo last month. The tester interacts with a prototype of AI agents supported by Google's Project Astra multimodal foundation model, Gemini. There are two continuous takes: one with the prototype running on a Google Pixel phone and another on a prototype glasses device. The agent takes in a constant stream of audio and video input. It can reason about its environment in real time and interact with the tester in a conversation about what it is seeing. The future is here and the implications far-reaching. #google #deepmind #innovation #ai