- AI KATANA
- Posts
- Google opens its Generative AI Platform to all
Google opens its Generative AI Platform to all
Also: Meta's open source AI MusicGen turns text and melody into new songs
Hi!
Today in AI, Google Cloud rolls out Generative AI support on Vertex AI, a one-stop-shop for AI application development, already in use by innovators like GitLab and Canva. Meanwhile, Meta presents MusicGen, an open-source AI capable of generating original music from text prompts, outperforming competing models like Riffusion and Mousai. In related news, OpenAI’s CEO, Sam Altman, encourages China to play a vital role in shaping AI safety guidelines, emphasizing the importance of global cooperation. In academia, a Harvard study unveils a promising technique to enhance the truthfulness and informativeness of AI responses, a vital progression in ensuring AI serves as a reliable tool in various applications. In terms of updates, ChatGPT now supports iOS and iPadOS with Siri and Shortcuts integration, and Microsoft incorporates AI voice chat in Bing's desktop search, powered by OpenAI's GPT-4 technology. Lastly, in the investment landscape, EliseAI raises $35M in Series C funding, fueling the automation revolution in real estate, and MatrixSpace secures $10M Series A funding to advance their AI-powered radar technology.
Let’s slice into it shall we?
🦾 Google opens its Generative AI Platform to all
🎸Meta's open source AI MusicGen turns text and melody into new songs
🇨🇳 OpenAI’s CEO calls on China to help shape AI safety guidelines
🎓Harvard Study - How to get Large Language Models (LLM) to tell the truth
Google Cloud has announced the general availability of Generative AI support on Vertex AI, providing a comprehensive platform for building custom generative AI applications. The platform gives developers access to Google's text model powered by PaLM 2, the Embeddings API for text, and a selection of other foundation models in Model Garden. It also includes the Generative AI Studio, which offers a range of user-friendly tools for model tuning and deployment. Additionally, the platform is backed by enterprise-grade data governance, security, and safety features. Companies like GA Telesis, GitLab, Canva, Typeface, and DataStax are already leveraging these features to innovate and build new AI capabilities.
Meta's open-source AI, MusicGen, can generate new pieces of music from text prompts and optionally align these to an existing melody, with the text setting the basic style. The technology is built on a Transformer model, similar to contemporary language models, that can predict the next section in a piece of music. It employs Meta's EnCodec audio tokenizer to decompose audio data into smaller components, thereby ensuring fast and efficient processing. The model was trained using 20,000 hours of licensed music, including an internal dataset of 10,000 high-quality music tracks and music data from Shutterstock and Pond5. The model and code have been released as open source on GitHub for research and commercial use, with a demo available on Huggingface.
OpenAI's CEO, Sam Altman, has urged China to play a crucial role in shaping the safety guidelines for AI. Altman underscored the high stakes of global cooperation due to the emergence of increasingly powerful AI systems. His company has been instrumental in promoting AI in China, particularly with the launch of ChatGPT last year. The CEO’s call comes at a time when both China and Silicon Valley are pouring talent and investments into AI, a strategic area that is defining the deepening tech rivalry between the two world's largest economies. The rapid advances in this technology have further emphasized the tensions over governmental regulations in the sector. China's leader Xi Jinping has expressed the need for greater state oversight to manage national security risks related to AI.
In a recent study, researchers put an AI model, known as LLaMA-7B, to the test to evaluate its ability to answer questions truthfully and informatively. They employed a technique called inference-time intervention (ITI), a method designed to enhance the quality of the model's responses. The results were quite promising. With the application of ITI, the AI model's responses were 54.5% truthful and 93.3% informative, a significant improvement over the baseline performance without ITI. The findings of this study are particularly relevant in today's digital age, where AI models are increasingly being used in various applications, from customer service to information retrieval. By enhancing the truthfulness and informativeness of AI responses, we can ensure that these models serve as reliable and trustworthy tools. This research is a significant step forward in the ongoing efforts to improve the reliability and usefulness of AI systems, making them more beneficial for users and businesses alike.
🛠️ AI tools updates
Gen-2 is capable of synthesizing videos from just text prompts. If you can say it, you can now see it, making it an extraordinary tool for visualizing your ideas. Gen-2 can also generate videos using a combination of a driving image and a text prompt, offering a unique blend of visual and textual inputs.
ChatGPT on iOS gets improved iPad support and Shortcuts integration OpenAI has released an update for its ChatGPT app for iOS and iPadOS, which includes iPad support for full-screen mode, drag and drop, Siri support, and Shortcuts integration.
Microsoft adds AI voice chat to Bing on desktop Microsoft has added voice support to its Bing search engine's chatbot on Edge for PCs, powered by OpenAI's GPT-4 technology, allowing users to ask Bing questions simply by speaking, with the chatbot now also supporting text-to-speech answers and can respond to questions with its own voice.
💵 AI Venture Capital updates
EliseAI, a conversational AI platform for the real estate industry, has raised $35 million in a Series C funding round led by Point72 Private Investments. The company will use the funding to continue product development and expand into new verticals such as healthcare. EliseAI’s products automate conversations between potential and current renters and buildings through SMS, email, phone, and webchat, and automate workflows through integrations with key software systems.
MatrixSpace has raised $10 million in VC funding, led by the Raptor Group, to accelerate technology advancements, customer adoption, and revenue growth. Intel Capital and a set of technology executives also participated in the funding round. MatrixSpace is revolutionizing radar technology with its compact, AI-powered system that digitizes the outdoors, dramatically extending the range of human senses over long distances to a degree previously unavailable. Customer applications include critical infrastructure security, general aviation and transportation, multiple defense applications, and robotics.
🫡 Meme of the day

⭐️ Midjourney prompt of the day

selfie group shot of spanish men 1700s
Have a productive week ahead! 💻 Before you go, have a look at Tracking Everything Everywhere All at Once. OmniMotion is a cool piece of tech, that allows for accurate, full-length motion estimation of every pixel in a video.