. .Today's 7 Top AI stories: Big Co's & Big Innovation
Friends, here are the top stories in AI today. OpenAI nearing $1B, Google adding Llama 2 to its platform, a new model for language to quadrupedal locomotion: there are some interesting stories here!
Special thanks to readers who shared yesterday’s newsletter on LinkedIn. Futurist Ross Dawson started a whole conversation around the Comb-shaped professionals story. Product leader Aaron Fulkerson shared a “nerd alert” about the AI and D&D story. Thanks guys!
Now here’s today’s stories.
OpenAI Passes $1 Billion Revenue Pace
An exclusive from The Information: "OpenAI is currently on pace to generate more than $1 billion in revenue over the next 12 months from the sale of artificial intelligence software and the computing capacity that powers it. That’s far ahead of revenue projections the company previously shared with its shareholders, according to a person with direct knowledge of the situation." [OpenAI Passes $1 Billion Revenue Pace as Big Companies Boost AI Spending]
Google's Vertex AI adds 100 models, including Llama 2, to its platform
Google Cloud Tech has upgraded Vertex AI, now offering access to 100 large models like MetaAI's Llama 2. Look at that collaboration! (It may be OpenAI against the world.) Google says the platform is used by firms like GE Appliances and has seen a 15x growth in customer accounts last quarter. It has also enhanced foundation models like PaLM 2, Codey, and Imagen, which now includes a digital watermarking feature powered by Google DeepMind’s SynthID. New tools like Vertex AI Extensions for real-time data retrieval and Vertex AI data connectors for data ingestion have also been launched to maximize model value. [Vertex AI extends enterprise-ready generative AI development with new models, tooling]
SayTap: Language to quadrupedal locomotion
Google Research has developed SayTap, a system that lets quadruped robots respond to human language instructions. It translates commands into robot actions like walking, running, and jumping using foot contact patterns. The research will proceed at Google DeepMind, potentially exploring emotion-specific commands and multi-modal inputs. It's easy to imagine this work being combined with two of Google's other recent papers: the robotics model family that allows robots to develop emergent skills with previously unseen objects, and yesterday's paper on Google research helping robots predict as-yet-unseen objects. Imagine saying "robot, go to this place and do this thing; you'll see some things you've never seen before, but I want you to anticipate those things and learn how to use them once you find them." Whoa. [SayTap: Language to quadrupedal locomotion]
Duet AI, Google’s AI assistant suite, expands across Google Cloud
Google's Duet AI, its generative AI suite, will extend to more Google Cloud products, the company announced at its Cloud Next conference today. Duet aids in code refactoring, infrastructure configuration, and context-aware code suggestions. Duet AI can also develop APIs from natural language prompts and is linked with BigQuery and Looker. However, concerns exist about coding errors and copyright issues with code-generating systems. [TechCrunch]
Meta's CoTracker is a model for tracking any point (pixel) on a video.
CoTracker, created by Meta AI Research, GenAI, and Oxford University, can track any video point, using Optical Flow benefits like pixel, grid, or selected point tracking. From sports performance to surveillance, the technology has many potential use cases. Users can access a pretrained CoTracker model from PyTorch Hub or GitHub. [CoTracker is a model for tracking any point (pixel) on a video. [Github]
Google Cloud, NVIDIA Intensify AI Collaboration
Google Cloud and NVIDIA have expanded their partnership to enhance AI computing, software, and services. They aim to offer customers AI infrastructure and software for generative AI, speeding up data science tasks. Google's PaxML, an LLMs framework, is now optimized for NVIDIA's accelerated computing. Google Cloud will also be among the first to use the NVIDIA DGX GH200 AI supercomputer's potential for generative AI tasks. [Google Cloud and NVIDIA Expand Partnership to Advance AI Computing, Software and Services]
Perplexity Adds Claude-2 for Enhanced Research Capabilities
Perplexity (a chatbot platform I've been using a lot lately) has added AnthropicAI's Claude-2 to its Pro platform, enhancing research with support for longer context and larger files up to 25MB. (Which is less than the 5 files of up to 10MB each in Claude's web UI.) Perplexity says Claude-2 improves writing quality with more natural content and quick responses. [via @AravSrinivas]
That's it! More news tomorrow!