| |
| Here's a look at today's AI briefing: - Google's most advanced AI model is now powering Bard.
- AMD launches sales for its new AI chip.
- Meta debuts image generator on the web.
- McDonald's to use Google AI in its operations.
- ChatGPT was the most viewed article on English Wikipedia this year.
Beth
p/beth-duckett | |
| 1 | | Google's Bard chatbot is now being powered by Gemini, the company's most advanced AI model. According to Google, it's the biggest improvement to Bard since its launch in March. CEO Sundar Pichai said the model will eventually come to Google's search engine, ad products, Chrome browser, and beyond. More: - The upgrade is boosting Bard's abilities in tasks like understanding, summarizing, reasoning, brainstorming, writing, and planning.
- Initially, Bard is being powered by Gemini Pro, a tailored version of the large language model.
- According to Google, Gemini Pro outperformed OpenAI's GPT-3.5 in six out of eight industry-standard benchmarks for AI models.
- Gemini is currently only in English, with other languages expected soon.
Related: - Starting Dec. 13, developers and business customers can access Gemini Pro through Google Generative AI Studio or Vertex AI in Google Cloud.
- Meanwhile, Google has developed Gemini Nano, a reduced version of the model, which is designed to run locally and offline on the Pixel 8 Pro.
- As of today, Pixel 8 Pro users will see Gemini Nano power Smart Reply on the Gboard keyboard and the auto-summarization feature in the Recorder app.
- Google Search is also already testing Gemini.
|     | |
| 2 | | Advanced Micro Devices (AMD) is now selling its MI300X chip, an AI accelerator that competes against Nvidia's H100 chip for AI training and inference. Meta, Microsoft, Oracle, and OpenAI have already signed up to use the MI300X when it starts shipping next year, taking on market leader Nvidia. More: - During a presentation today, AMD CEO Lisa Su called the MI300X "the most advanced AI accelerator in the industry."
- According to Su, MI300X rivals Nvidia's H100 AI GPU in training LLMs but outperforms it by 1.4x when running Meta's Llama 2 for inference tasks.
- Pricing details of the MI300X weren't shared publicly. Su said AMD's chip must offer a more competitive cost compared to Nvidia's ~$40,000 H100 to win over customers.
- AMD now projects a $45B market for data center AI processors this year, growing to $400B by 2027, reflecting the surging demand for AI chips.
Other products: - AMD also announced the MI300A APU, a version of the MI300X. It is now in production and being built in data centers; it will also power the El Capitan supercomputer.
- The company unveiled the Ryzen 8040 mobile processors, which incorporate neural processing units and offer added AI processing performance compared to predecessors.
Customers: - Meta and Microsoft, who are both big buyers of Nvidia H100 chips, plan to use the MI300X.
- Meta will use the chip for AI inference workloads and Microsoft through its Azure cloud web service.
- OpenAI said it will utilize AMD GPUs in Triton, a non-LLM software product used in AI research.
| | | |
| |
|
| 3 | | Meta is now offering a web-based image generator for U.S. users to try for free. The launch of "Imagine" on the web comes as Meta tests over 20 generative AI features across its platforms, including Facebook and Instagram. More: - Imagine first launched in Meta's messaging app, and it is now available as a standalone experience on the web, though it requires a Meta account to log in.
- Powered by Meta's Emu image generation model, Imagine churns out four high-resolution images for each text prompt.
- All the created images contain a visible watermark in the lower left corner that labels the content as AI-generated.
- Meta says it will soon add an invisible watermarking system.
Related: - Meta outlined other AI updates in a blog post today. They include a tool that uses Imagine to remix AI-generated images shared by friends in Messenger and Instagram chats.
- Meta is also upgrading its AI assistant to provide better mobile responses and search result summaries.
- Meta's 28 AI characters, based on celebrities like Paris Hilton and Tom Brady, are now fully accessible across the U.S. on WhatsApp, Messenger, and Instagram.
| | | |
| 4 | | Starting next year, McDonald's will use Google's generative AI offerings to power upgrades and optimize operations at its stores. This includes a chatbot called "Ask Pickles" to help restaurant workers with quick answers to questions based on a range of training data, including restaurant manuals and equipment data. More: - McDonald's plans to apply generative AI to data to optimize operations at "thousands" of restaurants.
- The rollout includes both in-store hardware and software upgrades, along with services delivered via Google Cloud.
- While details weren't specifically provided, McDonald's says it will result in "hotter, fresher food" for customers and help managers "quickly spot and enact solutions to reduce business disruptions."
- Google CEO Sundar Pichai said the company is "excited to see how McDonald's will use our generative AI, cloud, and edge computing tools."
Zoom out: - McDonald's plans to open 10,000 new stores worldwide by 2027 for the fastest-ever growth phase in its 60-year history.
| | | |
| 5 | | ChatGPT was the most viewed article on English Wikipedia in 2023, according to the Wikimedia Foundation. The article on OpenAI's chatbot dominated the list with nearly 49.5 million page views this year. 
What the numbers show: ChatGPT's article on Wikipedia maintained steady daily page views of 100,000 to 400,000 in the first half of the year, according to the Wikimedia Foundation. ChatGPT articles across all supported languages received over 78 million views. The bigger picture: ChatGPT received 153,000 visits when it was released on Nov. 30, 2022, surging to 15.5 million by the end of the first week. In week two, it reached 58 million, according to Similarweb. By October 2023, ChatGPT claimed 1.7 billion visits for the month, slightly down from its peak of 1.8 billion monthly visits in May. | | | |
| 6 | | 🍋 FRESH FUNDING 🍋 - Generative AI startup Together AI closed its Series A funding round at $102.5M. Kleiner Perkins led the round with additional support from Nvidia, Emergence Capital, NEA, Prosperity 7, Greycroft, and 137 Ventures.
- Speech AI firm Assembly AI bagged $50M in Series C funding led by Accel. Insight Partners, Smith Point Capital, Daniel Gross, Nat Friedman, and Y Combinator participated in the round.
- AutogenAI, a British AI startup that uses ChatGPT to automate paperwork filing for public and private contracts, raised $39.5M in fresh funding from Salesforce.
- Per The Information, AI software and storage startup Vast Data has raised fresh funding at a valuation of $9B, more than double its previous valuation of $3.7B in 2021. Fidelity Investments is reportedly leading the round with additional support from NEA and Bond.
For the latest on AI startup funding and more in the VC community, subscribe to our Inside Venture Capital newsletter. | | | |
| 7 | | QUICK HITS - The Kubernetes Cookbook is your comprehensive companion, equipping you with the knowledge and techniques to harness the full potential of Kubernetes.*
- Google unveiled its most powerful TPU, Cloud TPU v5p, and an AI Hypercomputer as part of its cloud offerings.
- Elon Musk clarified that his AI company, xAI, is "not raising money right now." This comes after a recent SEC filing showing X.ai planned to raise up to $1B in an equity offering. The filing showed the company had secured nearly $135M from four investors, with $865M left to be sold.
- OpenAI will pursue its tender offer in January, led by Josh Kushner's Thrive Capital. The AI startup will not issue any new shares, allowing existing employees to cash in on their holdings. OpenAI is targeting a valuation of $86B, more than triple its previous valuation of $28B.
- Apple released MLX, a machine learning framework optimized for Apple Silicon. Developers can use MLX to create AI models that run on MacBooks and other Apple devices.
- Microsoft's Seeing AI app, now available on Android, uses generative AI to provide detailed image descriptions for blind and low-vision users. It launched for iOS devices back in 2017.
- Toast to 2024: Key affiliate marketing trends to help your campaigns start off on the right foot in the new year. Get the free report.*
*This is a sponsored post. | | | |
 | AI and technology writer | | Beth is a contributing editor and writer of Inside's AI and Tech newsletters. She has written for notable publications like USA Today and the Arizona Business Gazette. During her time as a public policy reporter at The Arizona Republic, she received recognition with a Pulitzer Prize nomination and a First Amendment Award for her collaborative reporting on pension cost increases. Beth also authored a book on the solar photovoltaic industry in 2016. You can reach her at Beth.Duckett@yahoo.com. | | This newsletter was edited by Aaron Crutchfield | |
| |
| |
|
| |
Comments
Post a Comment