Google & Microsoft's Latest AI Projects

Author

AI Breakfast
November 17, 2023

Good morning. It’s Friday, November 17th.

Did you know: ChatGPT Plus subscriptions have appeared on eBay after CEO Sam Altman announced the temporary halt of new users?

In today’s email:

  • Microsoft Ignite 2023 AI Announcements

  • Google AI Initiatives

  • AI in Creative and Business Solutions

  • 10 New AI Tools

  • Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

Interested in reaching 44,251 smart readers like you? To become an AI Breakfast sponsor, apply here. 

Today’s trending AI news stories

Microsoft Ignite 2023 AI Announcements

> At Microsoft Ignite 2023, major AI announcements included the rebranding of Bing Chat to Copilot in Bing, expanding its availability across Windows and various Microsoft platforms.

> New custom AI chips, Azure Maia 100 AI Accelerator and Azure Cobalt 100 CPU, were introduced to enhance Azure's capabilities.

> A suite of new AI tools, including Copilot for Azure and Copilot Studio, was unveiled, offering diverse functionalities.

> Microsoft Teams received AI enhancements like advanced noise reduction and a new background decoration feature.

> A unified project management tool merging Microsoft To Do, Planner, and Project was announced for Microsoft Teams.

See the full Microsoft Ignite 2023 keynote here (Watch)

Google AI Projects

> Google Photos introduces Photo Stacks, improved document sorting, and calendar-linked reminders, enhancing image organization and event recall on Android and iOS.

> Google DeepMind's Mirasol AI uses 'artificial brainstorming' for innovative chess problem-solving, combining diverse AI strategies.

> DeepMind's Lyria, in partnership with YouTube, advances music creation, enabling creators to generate unique soundtracks and assisting in songwriting and production.

> The "Student of Games" AI, a collaboration involving DeepMind and others, demonstrates proficiency in various games, indicating progress towards artificial general intelligence.

> Google incorporates SynthID, an undetectable watermarking technology, into its DeepMind AI-generated music to ensure AI creation recognition.

> Google postpones the release of its Gemini AI, a rival to OpenAI's ChatGPT, due to challenges and slower cloud sales growth compared to Microsoft's AI-enhanced cloud operations.

AI in Creative and Business Fields

> Adobe's new AI venture, Project Sound Lift is an advanced tool that expertly filters out and eliminates specific background noises from a recording, such as traffic or applause, allowing a clearer focus on primary sounds like speech or musical instruments.

> Azure AI Speech has launched a pioneering Text to Speech Avatar in public preview, enabling the creation of lifelike speaking avatars and interactive bots from textual inputs. This advancement harnesses deep neural networks, transforming video production and digital engagement and signals a major lead on the field of AI.

> Dell and Hugging Face partner to simplify LLM deployment In a strategic move for AI integration in business. The alliance creates a bespoke Dell portal on Hugging Face, easing the implementation of AI models on Dell infrastructure, marking a pivotal step towards making advanced AI technology more accessible for companies.

> Tencent Stockpiled Nvidia AI Chips for ‘a Couple of Generations’ The company has allayed fears about U.S. export controls impacting its AI development by amassing a significant reserve of Nvidia Corp.’s H800 AI accelerators. According to President Martin Lau, these resources are sufficient to advance Tencent's Hunyuan AI model for several more iterations, as stated in an earnings call.

In partnership with LEMONAIDE

Lemonaide is the premier generative AI tool for musicians. Ethically-trained AI generates an endless number of MIDI tracks to your studio. 100% royalty-free, with dozens of customizable parameters.

Bring generative AI to your beats with a free 3-day trial:

Thank you for supporting our sponsors!

New AI-powered tools from around the web

> Pillar.io, an AI-powered all-in-one creator store, offers easy selling of digital products and services with features like Apple Pay checkout, personalized templates, and automated bookings.

> Sprig AI Product Insights Feed offers a centralized, real-time feed of user sentiment and product opportunities, streamlining the process of understanding user feedback. This AI-driven tool helps prioritize insights, identifying trends, opportunities, strengths, and correlations, enhancing product strategy.

> Flot.AI integrates ChatGPT into various apps and websites, enhancing productivity and creativity. It offers features like writing assistance, language translation, and task simplification, all powered by GPT-4.

> Taskade AI Agents allows users to build, train, and run custom AI agents, streamlining teamwork and task automation. These adaptable agents, powered by GPT-4 Turbo, can perform diverse roles in marketing, sales, and customer support.

> Zeliq is an AI-powered all-in-one sales solution offering a 1.8 billion prospect database and unlimited email enrichment. It enables automated multi-channel outreach, including calls, emails, and SMS, and features AI-driven performance enhancement tools to revolutionize sales processes.

> VideoGPT by VEED enables the creation of high-quality videos through ChatGPT using text descriptions. It features quick generation with stock footage, music, and text-to-voice.

> Meshy-1 is an AI-driven 3D modeling tool that transforms text or images into detailed 3D assets quickly. It is designed for 3D artists, hobbyists, XR developers, and game makers, offering user-friendly modes.

> DVC Extension for VS Code is an open-source tool that transforms VS Code into a comprehensive platform for managing ML experiments. It offers functionalities to run, track, compare experiments, and manage reproducible pipelines within the IDE.

> Polywork AI allows users to create personalized websites in seconds by simply pasting their LinkedIn URL. It offers various templates and customization options, including layout, color, and type, with domain connection and purchase options.

> JotUrl AI-driven Call-To-Actions tool enhances digital marketing and content curation by generating previews for non-embeddable pages. It is designed to optimize conversions, particularly beneficial for content creators and influencers using deep linking tools.

arXiv is a free online library where researchers share pre-publication papers.

The paper presents a novel approach for creating realistic, real-time 3D avatars using Gaussian splats and volumetric cage deformations, a departure from linear blend skinning. The method, leveraging 3D Gaussian Splatting, renders high-quality human representations from dense multi-view videos, surpassing current photorealistic avatars in speed and quality.

This paper by Google Research presents an innovative approach to evaluating LLMs by integrating insights from diverse evaluators like human-based, model-based, automatic-metrics-based systems. This method uses LLMs not just for direct evaluation but to effectively combine various evaluators’ scores, enhancing the correlation with human judgment.

The study presents a method for creating a 360-degree, high resolution 3D view of a person from a single image. It uses high-capacity 2D diffusion models for image synthesis, focusing on clothed humans. The method synthesizes multiple views by inpainting missing areas, guided by shape cues.

Emu Edit by GenAI, Meta, is an advanced multi-task image editing model setting new benchmarks in instruction-based image editing. It uniquely combines editing with computer vision tasks, covering a diverse range of operations like object alteration and texture changes. Emu Edit stands out for its precise execution of user instructions, showcasing exceptional performance in both compliance and image fidelity.

The paper explores enhancing language model reasoning by incorporating both valid and invalid reasoning demonstrations, inspired by how humans learn from positive and negative examples. This approach, called contrastive chain of thought, guides models to reason step-by-step while reducing mistakes.

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in 44,251 smart readers like you? To become an AI Breakfast sponsor, apply here.