This week's latest AI industry updates - December 17, 2024

Stay informed with SymphonyAI's weekly Eureka newsletter, covering all the important AI industry developments and technology advancements you need to know.

SymphonyAI news 

• Enterprise IT—Analyst report: Total Economic Impact study reveals 204% ROI for SymphonyAI enterprise IT service management customers 

• Retail—Analyst report: AI-based innovations propel retail profitability and sustainability, research finds 

• Industrial—Award: SymphonyAI’s IRIS Foundry named AI impact of the year champion in Digital Engineering Awards  

•  Retail—Case study: Southern Co-op implements AI-based assortment and space 

• Financial Services—Article: Four ways generative AI accelerates financial investigations 

Big tech announcements 

• Google introduces Gemini 2: Google released Gemini 2.0, featuring enhanced multimodal capabilities and agentic AI features, designed to enable more intelligent and interactive AI assistants across various applications. 

• Google debuts new AI video generator Veo 2: Google released Veo 2, an improved AI video generation model competing with OpenAI's Sora, offering more realistic videos and updated Imagen 3 for enhanced image generation, both available on a limited basis through Google Labs. 

• Google unveils AI coding assistant Jules: Google launched Jules, an AI coding assistant built on Gemini 2.0, capable of autonomously fixing bugs and preparing code changes, integrating with GitHub's workflow system to enhance developer productivity. 

• ChatGPT gets screensharing and real-time video analysis: OpenAI has added video and screen sharing capabilities to its advanced voice mode in ChatGPT, rivaling Gemini 2.  

• OpenAI launches ChatGPT Projects: The new feature allows users to organize files and conversations, similar to Google's NotebookLM. 

• Microsoft launches Phi-4, a smaller, highly efficient AI model: Microsoft’s new AI model outperforms larger competitors—including Google’s Gemini Pro 1.5 and Gemini Ultra and OpenAI’s GPT-4—in mathematical reasoning. Phi-4 uses 14 billion parameters, much smaller than the hundreds of billions or possibly trillions of parameters used by Google’s and OpenAI’s models.  

• Meta updates its smart glasses with real-time AI video: Ray-Ban Meta smart glasses receive new AI-powered features, including live AI with video capabilities, real-time language translation, and Shazam integration for music recognition. 

• Meta FAIR shares new research, models, and datasets: Meta Fundamental AI Research (FAIR) releases new AI research artifacts, including Meta Motivo for embodied agents, Meta Video Seal for video watermarking, and other advancements in AI capabilities and safety. 

AI impact 

• Nearly one in four US workers use generative AI on a weekly basis: A quarter of US workers report using generative AI weekly, but companies are struggling to establish policies and controls to manage this rapid adoption, leading many employees to use AI tools secretly due to unclear guidelines and fear of negative perceptions. 

• By 2028, 30% of Fortune 500 companies could use AI-only service channels: Gartner predicts that by 2028, 30% of Fortune 500 companies will adopt single, AI-enabled channels for customer service, transforming interactions through text, image, and sound. 

• Klarna’s CEO says it stopped hiring because of AI: Klarna's CEO claims the company stopped hiring due to AI efficiencies, but job listings and company statements indicate they are still selectively hiring for essential roles.  

• Systematic use of Gen AI cognitively overwhelms many people, Microsoft Research finds: A Microsoft Research study reveals that effective use of generative AI systems requires strong metacognitive abilities, with many people struggling to meet the cognitive demands of professional AI use, leading researchers to propose specific improvements in planning, self-evaluation, and interface design. 

• Thoughtworks partners with AWS to accelerate gen AI Adoption in Asia Pacific: Thoughtworks and AWS have signed a strategic collaboration agreement to help enterprises across Asia Pacific develop and implement generative AI solutions, aiming to improve operational efficiencies, customer engagement, and revenue growth across various industries.

AI agents: The next wave of AI 

• DeepMind VP sees a fundamental shift towards autonomous AI agents: DeepMind's Oriol Vinyals discusses AI's evolution from specialized systems to autonomous agents, highlighting challenges in scaling, data scarcity, and the development of more complex capabilities like planning and logical thinking. 

• Could Google's Gemini 2.0 be the beginning of truly autonomous AI? Google’s release of Gemini 2.0 introduces AI agents capable of completing complex tasks autonomously, potentially marking a significant step towards truly autonomous agentic AI. 

• How AI agents are revolutionizing automation: AI agents are evolving from assistants to autonomous collaborators, with vertical AI agents and multi-agent systems poised to revolutionize enterprise automation by reimagining workflows, eliminating operational overhead, and creating new possibilities across various industries.

Noteworthy AI applications and use cases 

• Pika Labs releases AI video generator 2.0 with new features: Pika Labs releases version 2.0 of its AI video generator, introducing a "Scene Ingredients" feature that allows users to incorporate their own images into AI-generated videos, offering wider availability compared to OpenAI's Sora. 

Platform and model developments 

• Large language models aren't so large anymore, AI analysts estimate: Recent AI language models like GPT-4o and Claude 3.5 Sonnet achieve similar performance to larger models with significantly fewer parameters, challenging the notion that bigger models are always better. 

• Anthropic’s fastest model, Claude 3.5 Haiku, is now generally available: Anthropic makes Claude 3.5 Haiku, a smaller, faster AI model, available to all users through its Claude chatbot, offering competitive performance and features despite some limitations. 

• Cohere’s smallest, fastest R-series model excels at RAG, reasoning in 23 languages: Cohere released Command R7B, a small, fast AI model designed for enterprise use cases, outperforming competitors in math, coding, and retrieval-augmented generation tasks. 

• Lambda launches ‘inference-as-a-service’ API: Lambda launches a low-cost Inference API service for deploying AI models, using its extensive GPU resources to offer competitive pricing and scalability for enterprises and developers. 

• Elon Musk's Grok AI on X is now faster and cites legacy media as sources: xAI has upgraded its Grok AI chatbot on X, improving speed and performance while adding the ability to cite legacy media sources in its responses.

AI regulations, public policy, and ethical AI 

• Major cloud providers could get key role in AI chip access outside the U.S.: The U.S. plans to empower major cloud providers like Google and Microsoft as global gatekeepers for AI chip access outside the U.S., aiming to prevent Chinese access while streamlining exports to allies. 

• Meta asks California attorney general to block OpenAI’s conversion to for-profit: Meta supports Elon Musk's legal action to block OpenAI's transition from nonprofit to for-profit status, arguing it could have significant implications for Silicon Valley and questioning the validity of OpenAI's new business model. 

• Amid lawsuits and criticism, Character AI unveils new safety tools for teens: Character AI announces new teen safety tools amid lawsuits and criticism, including a separate model for teens, content blocks, usage notifications, and clearer disclaimers, while repositioning itself as an entertainment platform rather than an AI companion service.

Research  

• New LLM optimization technique cuts memory costs up to 75%: Sakana AI researchers developed "universal transformer memory," a technique using neural networks to optimize language models' memory usage, potentially reducing costs and improving performance for enterprise AI applications.

Funding 

• Anybotics raises $60M to bring autonomous industrial robots to the U.S.: Swiss robotics company Anybotics raises an additional $60 million, closing its Series B round at $110 million, to expand its autonomous inspection robot business globally, particularly in the U.S. market. 

• 200 VCs wanted to get into Lumen Orbit’s $11M seed round: Lumen Orbit, a startup building data centers in space, raised $11 million in seed funding at a $40 million valuation amid high investor interest, aiming to launch its first demonstrator satellite in May. 

• Legion Technologies secures $50M from Silicon Valley Bank: Legion Technologies, an AI-powered workforce management platform for hourly employees, secures $50 million in financing from Silicon Valley Bank, bringing its total funding to $195 million. 

• Twelve Labs raises $30 million: Twelve Labs, a startup developing AI models for video analysis and search, has raised $30 million to expand its capabilities and grow into new markets, including potential security and defense applications. 

• Refute raises £2.3 million to counter misinformation: The London-based startup detects and responds to disinformation for commercial entities has raised £2.3 million in pre-seed funding to combat the growing threat of AI-fueled misinformation campaigns targeting companies.