This Week in AI

Hello Niuralogists!

In the dynamic landscape of artificial intelligence, this week's edition is dedicated to bringing you the latest advancements. Our primary objective is to examine the implications of these updates on various aspects of our lives, encompassing workplaces, businesses, policies, and individual experiences. In this edition, we will explore notable developments, including the Runway ML and Getty: AI Video Partnership, as well as the considerations within Stability AI amidst executive turmoil.

For further insights, read more...

Runway ML and Getty: AI Video Partnership

Runway ML, an AI video startup, has partnered with Getty Images, a stock media giant, to develop a cutting-edge generative model named the Runway & Getty Images Model (RGM). This collaboration enables Hollywood studios and advertising agencies to refine the model using their data, facilitating the creation of personalized, high-quality video content. By combining Runway's advanced AI technology for realistic video generation with Getty's extensive image library and distribution network, the partnership aims to reshape workflows in major industries. However, this news comes amid a legal challenge for Runway, as artists allege that the models were trained on their content without proper consent. Despite ethical concerns surrounding AI usage, initiatives like this highlight the potential for widespread technology adoption in various sectors.

AI Speeds up the Process of Solving Complex Problems

For companies like FedEx grappling with the intricate challenge of efficiently routing holiday packages, the conventional approach involves employing specialized software known as a mixed-integer linear programming (MILP) solver. However, this method often entails lengthy processing times, lasting hours or even days, prompting companies to settle for suboptimal solutions due to time constraints. In response, researchers from MIT and ETH Zurich have harnessed the power of machine learning to expedite the process. Identifying a critical intermediate step in MILP solvers and implementing a filtering technique significantly simplified the complexity, allowing machine learning to pinpoint optimal solutions for specific problems. This data-driven approach accelerates MILP solvers by 30 to 70 percent and maintains accuracy, empowering companies to tailor general-purpose solvers to their unique challenges and obtain optimal or improved solutions in a more manageable timeframe.

GPT-4 Demonstrates Proficiency in Radiological Tasks

According to a recent study by Microsoft, GPT-4 showcases remarkable proficiency in radiology tasks such as summarization and disease classification, achieving state-of-the-art results. The study reveals that GPT-4's summarizations were either comparable to or preferred over those generated by experienced human radiologists. The model demonstrated exceptional capabilities in automatically organizing intricate reports into standardized formats. Researchers emphasized the potential of integrating GPT in enhancing disease interpretation, leveraging real-world data, and improving patient education. This study adds to the growing body of evidence supporting the transformative impact of AI in the medical field, highlighting GPT-4's contributions to enhancing both diagnostic processes and overall workflows.

Visual Electric Launches to Revolutionize AI Art Generation Beyond Chat Interfaces

If you've experimented with several text-to-image AI art generation services like Midjourney or various iterations of DALL-E by OpenAI, you've likely encountered a common theme—they often resemble a chat interface. Typically featuring space for user text prompts and responses delivering images within a messaging format, this interface has proven effective for many users. However, some argue that it's limiting, especially for established artists and designers seeking a more nuanced interaction with AI. San Francisco-based Visual Electric aims to provide a fresh perspective. Emerging from stealth after securing funding from Sequoia, BoxGroup, and Designer Fund last year, the startup introduces an alternative approach it believes better caters to the visual creativity needs of users, diverging from the conventional back-and-forth text interaction with AI models.

Stability AI Considers Sale Amidst Executive Turmoil

Stability AI, a renowned open-source AI company, is reportedly contemplating a sale in response to investor demands for CEO Emad Mostaque's resignation due to financial instability and management concerns. Notably, Coatue, Stability AI's major investor, has urged Mostaque to step down, a year after leading a $101 million funding round for the company. Although Stability AI has engaged in acquisition talks with Cohere and Jasper, it asserts that it is not actively seeking a sale. This development coincides with cultural issues within the company, leading to executive departures amid a rapid expansion of staff for new releases. Despite the company's name, signs of instability raise questions about the potential impact on its standing as a leading open-source player in the AI space.

Q&Ai

What lies ahead in the future for generative AI?

Speaking at the "Generative AI: Shaping the Future" symposium on November 28, the inaugural event of MIT’s Generative AI Week, keynote speaker and iRobot co-founder Rodney Brooks cautioned attendees against uncritically overestimating the capabilities of this burgeoning technology. The symposium, held at MIT’s Kresge Auditorium and attended by hundreds from academia and industry, conveyed messages of optimism regarding the potential of generative AI to contribute positively to the world, especially in the realms of art and creativity. However, interspersed with these hopeful prospects were warnings about the potential pitfalls if these AI tools are not developed responsibly. MIT President Sally Kornbluth, in her opening remarks, highlighted various projects undertaken by faculty and students using generative AI to make a positive impact on the world.

How to animate anything with AI?

Discover the latest viral AI model that introduces an exciting technique for animating a diverse range of elements, from characters to images. To experience a similar demo, follow these steps: First, navigate to the MagicAnimate demo on HuggingFace. Next, upload the subject you want to animate, with the consideration that the example motion sequences are optimized for full-body vertical images. Finally, click the "animate" button and be patient, as this demo might take a bit longer to process due to the influx of users currently exploring its capabilities.

What is the process for building a basic GPT using natural language?

Exclusive to ChatGPT Plus users, access GPT Builder by navigating to the Explore tab and selecting "Create a GPT." Utilize the "Configure" tab to engage ChatGPT conversationally, describing your desired output and allowing it to generate a draft. Once the basic drafts are complete, head back to the "Configure" tab to fine-tune and optimize your GPT. Experiment by uploading files, such as spreadsheets, to enrich the training data. Interact with your GPT through conversation to assess its adherence to instructions, and if adjustments are needed, engage in a dialogue to refine its understanding. Keep in mind that GPT Builder is in its early stages, and if predictions hold true, its future GPT Builder Shop could become a central hub akin to an App Store for AI applications, providing an advantageous headstart in the evolving landscape of GPT technology.

Tools

?? MindMap Generator visualizes ideas for learners & and professionals

?? Walles.ai gets instant insights & and answers from any site/PDF in the browser

?? OpenRep is an AI-powered organic digital marketing workflow optimization

?? Recall- Save summarizes any online content

?? Show HN: Dreamachines.ai discovers the potential of artificial intelligence through dream machines