Generative AI Video Tools
Highlights
1. A recent survey found that most executives are dissatisfied with their progress in GenAI adoption.
2. Microsoft introduced Copilot Pages, a collaborative AI-powered workspace, and Copilot agents to automate processes.
3. Amazon announced Amelia, an AI assistant that helps third-party sellers.
4. Runway and Luma Labs’ new AI video APIs allow developers to integrate generative video tools into their applications.
5. Snap’s new augmented-reality Spectacles are highly impressive in terms of functionality.
6. Neuralink’s Blindsight implant aims to restore vision while Synchron’s new brain-computer interface allows people with paralysis to control Amazon Alexa.
Innovation Insights
1. The stairway to GenAI impact (BCG)
Two-thirds of executives are dissatisfied with their progress in GenAI adoption. To realize GenAI’s full potential, businesses must focus on new revenue, cost reduction, and productivity gains, potentially achieving up to 10-15 times ROI in three years. Achieving this requires clear EBIT goals, prioritizing high-value use cases, and embedding accountability at the leadership level. Scaling GenAI across an organization involves more than just technology—companies must address training, process redesign, and change management for success. Establishing a GenAI transformation office ensures structured implementation, allowing companies to achieve significant EBIT improvements over time.
2. Power, chips and constraints: The breakthrough AI needs (The Economist)
Generative AI faces rising costs and diminishing breakthroughs due to the high energy demands of large models, raising concerns about its long-term economic viability. Investors have heavily funded AI startups, as well as Nvidia, but competition is intensifying as specialized AI chips and more efficient models emerge. Some tech firms are innovating with smaller, task-specific models to overcome constraints. As a result, the landscape is shifting, making it uncertain which companies will dominate AI in the future. Governments must focus on fostering talent and innovation, rather than solely restricting rivals, to maintain leadership in AI advancements.
3. Scaling: The state of play in AI (One Useful Thing)
Large Language Models (LLMs) are becoming increasingly capable as their scale, in terms of parameters and computational power, grows, making them better at complex tasks. Training larger models requires exponentially more data and computing power, leading to significant increases in cost and effort. Recent advancements suggest that improvements in AI not only come from scaling during training but also from allowing models to “think” longer during inference, leading to more accurate results. This dual scaling—both in training and thinking—indicates that AI capabilities will continue to advance rapidly. As AI progresses, independent AI agents will handle more complex problems, transforming industries and society, but also bringing new challenges and opportunities.
4. The Narrative AI Advantage? A Field Experiment on Generative AI-Augmented Evaluations of Early-Stage Innovations (Harvard Business School)
This preprint describes a field experiment using GPT-4 to evaluate early-stage innovations for the MIT Solve Global Health Equity Challenge, comparing human-only evaluations with AI-augmented processes. It found that AI-assisted evaluators were 9% more likely to reject submissions. Deeper engagement with AI’s objective rejections led to more overrides of the AI while deeper engagement with AI’s subjective rejections led to more alignment with AI. Overall, it concluded that while AI enhances evaluation consistency, human oversight remains essential, especially for subjective judgments.
AI Innovations
1. Slack
Slack’s new agent-powered work operating system integrates AI, CRM data, and automation, with features like Salesforce channels, AI-enhanced search, and third-party agents (Slack).
2. ClickUp
ClickUp has launched an AI-powered chat integrated into its project management platform, aiming to streamline communication and task management by directly linking conversations to tasks and using AI to automate workflows and insights, positioning itself as a comprehensive “everything app” for work (VentureBeat).
3. Microsoft
Microsoft is launching the next wave of its 365 Copilot, introducing Copilot Pages, a collaborative AI-powered workspace, and Copilot agents to automate business processes. Improvements to Copilot in apps like Teams, Excel, PowerPoint, and Outlook are enhancing productivity, with advanced AI capabilities integrated across various tasks. These updates aim to streamline workflows, boost efficiency, and enable AI-driven collaboration in real-time (Microsoft).
4. Alibaba
Qwen2.5 introduces a range of open-source language models, including specialized coding and math models, with enhanced performance, multilingual support, and extensive capabilities designed to drive innovation in AI development (GitHub).
6. Mixtral
Mistral has introduced a free tier for its serverless platform, reduced pricing across all models, unveiled the upgraded Mistral Small v24.09 model, and launched free vision capabilities with Pixtral 12B for enhanced image understanding in its AI tools (Mixtral).
7. Amazon
Amazon has introduced Amelia, an AI assistant designed to help third-party sellers quickly resolve account issues, access sales and inventory data, and streamline their business operations (CNBC).
8. Google
Google Research has developed a new AI model that can identify vocalizations of eight whale species, including the mysterious “Biotwang” sound attributed to Bryde’s whales (Google).
Google has launched the Open Buildings 2.5D Temporal dataset to track building changes, estimate heights, and count buildings across the Global South from 2016 to 2023, using satellite imagery to support urban planning, disaster response, and research efforts (Google).
9. World model
1X Technologies’ new generative model, trained on real-world sensor data from robots, aims to bridge the “sim2real gap” by more accurately simulating object interactions and environment dynamics for robotics, but still faces challenges like occasional unrealistic predictions (VentureBeat).
10. Video
Gen-3 Alpha Video to Video allows users to change the style of input videos by using a text prompt (Runway).
Runway and Luma Labs’ new AI video APIs allow developers to integrate generative video tools into their applications, potentially revolutionizing how video content is created and expanding AI’s accessibility in everyday products (Tom’s Guide).
YouTube is introducing AI-powered tools and features, such as Dream Screen and automatic dubbing, to empower creators, enhance audience connections, and expand monetization opportunities (YouTube).
Runway and Lionsgate have partnered to develop an AI model customized on Lionsgate’s film catalog, aimed at augmenting filmmakers’ creative processes and exploring new opportunities in AI-driven content creation (Runway).
11. Wildfire detection
AI-powered technologies, including satellite constellations and camera systems, are being developed to detect wildfires faster, giving first responders a crucial head start in preventing large-scale damage (MIT Tech Review).
Other Innovations
1. AR
Snap’s new augmented-reality Spectacles are highly impressive in functionality but also look quite goofy when worn (MIT Tech Review).
2. Computing
New technologies, like “in-memory” and “neuromorphic” computing, inspired by the brain’s efficiency, aim to overcome “von Neumann bottleneck” by processing data faster and using less energy. Additionally, optical computing shows promise for AI tasks, using light instead of electricity to dramatically increase speed and energy efficiency (The Economist).
3. Apple Watch
Apple Watch has a new function: sleep apnea detection (Yahoo).
4. Virtual brain model
Scientists have created a virtual brain model based on a fruit fly’s visual system that can predict neuron behavior, offering a tool for testing ideas before live experiments and helping make AI systems more energy-efficient by mimicking brain strategies (NPR).
5. Brain-computer interface
Elon Musk’s Neuralink has received the FDA’s “breakthrough device” designation for its experimental Blindsight implant, which aims to restore vision even in individuals who have lost both eyes and their optic nerve (Reuters).
Synchron has developed a brain-computer interface that allows people with paralysis to control Amazon Alexa and other devices using their minds (Wired).
