• AI KATANA
  • Posts
  • Google Just Launched Gemini, Its Long-Awaited Answer to ChatGPT

Google Just Launched Gemini, Its Long-Awaited Answer to ChatGPT

Also: BlackRock to roll out first generative AI tools to clients next month

Author

AI KATANA
December 06, 2023

Hi!

Today's email newsletter covers several groundbreaking developments in the AI and technology sector. Google has unveiled Gemini, a versatile AI model integrated into its Bard chatbot, marking a significant step in AI's multimodal capabilities. A new study raises concerns about AI chatbots' potential for misuse, indicating their ability to "jailbreak" other chatbots and bypass safety measures. In the competitive AI processor market, AMD has launched the MI300X chip, directly challenging Nvidia's dominance. BlackRock is set to release its first generative AI tools, integrating them into its risk management systems to enhance productivity and efficiency. Nvidia plans to build a new supercomputer in Singapore, reinforcing its commitment to AI technology leadership. Lastly, Alibaba's "Animate Anyone" tool, capable of animating still images, has sparked discussions around digital privacy and misinformation, while VAST Data's surging valuation and IPO plans highlight the booming interest in AI-driven data infrastructure solutions.

Sliced:

  • 🆕 Google Just Launched Gemini, Its Long-Awaited Answer to ChatGPT

  • 🤖 Jailbroken AI Chatbots Can Jailbreak Other Chatbots

  • ⚙️ AMD rolls out new rival to Nvidia’s AI chip

  • 🏦 BlackRock to roll out first generative AI tools to clients next month

  • 🌏 Nvidia CEO hints at new supercomputer in Singapore

Google has launched Gemini, a highly advanced AI model, integrated within its Bard chatbot. This launch is seen as Google's response to the success of OpenAI's ChatGPT. Gemini is unique in its ability to process not just text, but also images, video, and audio, positioning it as a potentially groundbreaking algorithm in Google's history. It's available in over 170 countries and will soon be accessible to developers through Google Cloud’s API. Gemini has three versions: Ultra, Nano, and Pro, each offering different capabilities. Gemini Pro powers Bard, enabling more advanced reasoning and planning, while Gemini Ultra, the most powerful version, is set for release in 2024 after extensive safety checks. Gemini's multimodal capabilities represent a significant step in AI development, demonstrating Google's commitment to maintaining its position as a leader in AI technology.

A new preprint study has revealed that AI chatbots can be manipulated to "jailbreak" other chatbots, leading them to provide users with dangerous information. The study observed targeted AIs breaking built-in restrictions to offer advice on illegal activities like synthesizing methamphetamine, building bombs, and laundering money. Researchers successfully used a chatbot acting as a research assistant to develop prompts that could jailbreak other chatbots, bypassing their programmed safety measures. This was notably effective against several large language models (LLMs), including GPT-4, Anthropic's Claude 2, and the open-source Vicuna chatbot. The study highlights the inherent risks in AI-powered chatbots, suggesting vulnerabilities beyond the control of individual AI developers. The findings call for greater attention to the potential misuse of AI technologies and the difficulty in completely eliminating harmful capabilities in AI models.

AMD has launched the MI300X, an advanced AI chip designed to rival Nvidia’s dominance in the AI processor market, which is expected to be worth $400 billion by 2027. This chip is touted by AMD's CEO, Lisa Su, as the most advanced AI accelerator, surpassing Nvidia's current offerings. The MI300X is part of a strategy to capture a significant share of the rapidly growing AI chip market, with expectations to reach $1 billion in sales by mid-2024. AMD's MI300X will also include a supercomputing variant, the MI300A. The launch signifies AMD's ambition to compete with Nvidia, whose AI chips have been essential for companies like Amazon, Meta, Microsoft, and Google in developing generative AI tools. This new chip has also garnered interest from major tech players, including OpenAI, which plans to use the MI300X in its latest version of Triton AI software. AMD's efforts extend to software as well, with the introduction of the ROCm 6 software platform, competing with Nvidia’s Cuda platform.

BlackRock, the world's largest money manager with $9.1 trillion in assets, is set to roll out its first generative AI tools to clients in January. This move is part of BlackRock's broader initiative to harness AI technology for enhancing productivity. The company has developed a "co-pilot" for its Aladdin and eFront risk management systems, using large language model technology. This AI integration will assist clients in extracting information more efficiently. BlackRock's AI tools will initially produce "first drafts" that require subsequent quality control, ensuring data remains within BlackRock's secure environment. Additionally, the company plans to deploy Microsoft’s AI add-on to Office 365 across its operations. This strategic implementation of AI aims to democratize coding, reduce fixed costs, and boost margins, indicating BlackRock's commitment to leveraging emerging technologies to stay ahead in the financial services industry.

Nvidia's CEO, Jensen Huang, has hinted at the construction of a new supercomputer in Singapore during his visit to discuss potential large investments in AI with Prime Minister Lee Hsien Loong and the Singapore Economic Development Board (EDB). This initiative is part of Nvidia's broader strategy to maintain its leadership in AI technology. Huang mentioned that while Nvidia already has a supercomputer in Singapore, they plan to build an even larger one. The company is also in talks with EDB about investing in a significant, iconic site for AI in Singapore. Additionally, Nvidia is collaborating with the Infocomm Media Development Authority on an AI project named Sea-Lion, a large language model tailored to Southeast Asia, covering 11 languages in the region. This move by Nvidia aligns with Singapore's refreshed national AI strategy, Nais 2.0, and demonstrates the company's commitment to advancing AI technology globally, particularly in Southeast Asia.

🛠️ AI tools updates

The AI tool "Animate Anyone," a project by the Alibaba Group's Institute for Intelligent Computing, represents a significant advancement in AI-generated imagery. Unlike previous AI tools, it can animate still images of humans, anime, cartoons, and humanoid figures, creating lifelike movements with minimal disruptions or noise. This is achieved by combining models like ReferenceNet and a temporal overlay for smoother transitions. Currently available on GitHub as an open-source tool, it's primarily used by developers. While offering potential benefits for industries like video game development and filmmaking, it also raises concerns about digital privacy and the spread of misinformation. The technology is not yet perfect, with ongoing efforts to refine the algorithms, especially for complex backgrounds.

💵 Venture Capital updates

VAST Data, a data infrastructure company, has recently achieved a valuation of $9.1 billion, more than doubling its worth from two years ago, following a $118 million funding round led by Fidelity Management & Research, with participation from several venture capital firms. The New York-based firm, specializing in data-intensive computing solutions for enterprises, has seen its revenue triple year-over-year, surpassing $200 million. Backed by Nvidia and partnering with AI cloud computing providers, VAST Data is exploring an IPO in 2024, subject to market conditions. With a global workforce exceeding 700 employees, the company is expanding into new international markets. Notably, the company maintains a strong financial position, marked by a consistent positive cash flow and a high gross margin of nearly 90%.

🫡 Meme of the day

⭐️ Generative AI image of the day