This week, Meta announced Metaverse avatars powered by “concept-predictive” language models and AI. Meanwhile, industry leaders predict the emergence of autonomous AI agents by 2025. Deloitte predicts that by 2025, 25% of companies using generative AI will have autonomous AI agents, rising to 50% by 2027. Also this week, OpenAI finally released a video creation tool. “Sora” after teasing the home in the spring. In other news, there’s a shift away from simply scaling up models to algorithmic efficiency, humanoid robots spark interest in STEM in Zimbabwe, and AI models ‘long thinking’ to solve complex problems You can also see us working on this.
LCM operates based on a conceptual prediction paradigm rather than traditional token-level prediction.
Meta launches Meta Motivo, an AI model that powers Metaverse avatars and large-scale context models
Source: Internal announcement, meta
Meta announced Meta Motivo, a behavior-based model designed to control the movement of virtually embodied humanoid agents through unsupervised reinforcement learning. This could potentially allow for more realistic NPC and character animations in the Metaverse. The company also released Large Concept Models (LCMs), which take a new approach to language modeling by separating inference from linguistic representations and predicting concepts rather than tokens. Additionally, Meta launched Video Seal, an open-source framework for embedding invisible watermarks in videos, building on previous Audio Seal technology to demonstrate resilience to common editorial changes. .
Management predicts the emergence of autonomous AI agents in 2025, increasing profitability for the AI sector
Source: Reuters NEXT Conference
At the Reuters NEXT event, industry leaders predicted that incremental advances in reasoning will lead to autonomous agents capable of independent actions such as making purchases and scheduling meetings by 2025. OpenAI CFO Sarah Friar expects artificial general intelligence (AGI) to become a reality within the next few years. Venture capital sentiment has shifted from prioritizing revenue growth to focusing on profitability, indicating the maturation of the AI market. A specific example from a portfolio company shows that deploying an AI sales team reduced labor costs by 75%.
AWS Re:Invent 2024 Announces Nova Foundation Model and Healthcare AI Initiative
Source: Re:Invent analysis by Himanshu Jain
Himanshu Jain, Amgen’s technology strategy and innovation leader, outlines the Amazon Web Services platform evolution at Re:Invent 2024. The cloud services pioneer has announced the Nova foundation model family with four different text processing variants, alongside Canvas and Reel modules for images and images. Video generation respectively. Healthcare adoption is maturing, with Merck, Amgen, Geisinger, Gilead, and Natera pioneering the implementation of genAI in their businesses.
Broadcom reaches $1 trillion valuation thanks to AI chip partnership
Source: Reuters
Broadcom has surpassed the $1 trillion valuation threshold thanks to its expanding artificial intelligence semiconductor strategy. The company’s strategic positioning includes development partnerships with leading cloud and technology providers, particularly OpenAI’s inference chip development program targeted for deployment in 2026 through TSMC’s manufacturing capabilities. CEO Hock Tan predicts that the company’s AI chip division could generate nearly $90 billion in revenue by 2027.
(Wave)
Wayve’s end-to-end self-driving system learns directly from video
Source: Wave
London-based startup Wayve has introduced an AI-driven self-driving system that ditches traditional high-resolution mapping in favor of learning from raw, unlabeled driving footage. This approach allows the model to mimic human driving behavior, with plans to provide Level 3 driving assistance in the short term and eventually move to Level 4 and 5 full autonomy. Tesla’s self-driving modes, including Fully Self-Driving (FSD), also rely heavily on video data captured by the car’s cameras.
Biotech sector navigates transformational 2024 amid AI integration and funding changes
Source: Industry analysis by Dr. Andrii Buvailo via LinkedIn
The biotechnology sector experienced major structural changes in 2024 due to both strategic and technological consolidation. Notable developments include Xaira Therapeutics’ $1 billion financing, in contrast to the overall decline in M&A activity. Clinical advances will center around gene editing technologies, with Casgevy’s approval in 2023, and FDA approvals for multiple applications including oncology (Tecelra, Bizengri, KRAZATI), neurology (Vyalev), and rare diseases (Yorvipath). expanded into the therapeutic field of Integration of AI technology has accelerated through partnerships with NVIDIA and Google, and the ‘Techbio’ venture has demonstrated early clinical validation. But layoffs were another common theme that affected companies like Ginkgo Bioworks and 23andMe. Meanwhile, geopolitical tensions, particularly U.S.-China relations, prompted regulatory responses through the BIOSECURE Act. .
(Datacenter stock image from Adobe Stock)
NVIDIA CEO highlights ‘long-term thinking’ AI models for improved inference
Source: Jensen Huang in industry conference presentation, WSJ
NVIDIA CEO Jensen Huang’s emphasis on “long-term thinking” AI models signals a pivotal shift in computational methodologies with systems designed to extend inference cycles. An analysis of NVIDIA’s quarterly earnings reveals that this evolution has the potential to dramatically expand the demand for inference computing, especially for complex scientific and mathematical challenges. These systems are capable of 100 days of computational processing and represent a departure from the rapid response framework, as OpenAI’s o-series models have already demonstrated value in genetic research at Harvard Medical School. This deliberative AI processing approach, reflecting Kahneman’s System 2 cognitive framework, potentially reduces error rates while informing fundamental changes in infrastructure requirements. This development coincides with NVIDIA’s record quarterly revenue of $35.1 billion, reflecting strong market validation of the demand for advanced AI architectures.
Exxon enters AI data center power battle with natural gas CCS facility
Source: TechCrunch
TechCrunch reports that ExxonMobil has announced plans to build a more than 1.5 gigawatt natural gas power plant specifically targeted for AI data center operations, marking the company’s first off-site power generation project. The facility will incorporate carbon capture and storage (CCS) technology and aim to reduce emissions by 90%. The plant operates independently from the grid infrastructure and addresses the critical power supply constraints expected to affect nearly 50% of new AI data centers by 2027.
Alphabet stock price as of December 13th (Google Finance)
Alphabet’s stock price soars and quantum computing chip announcement
Source: Alphabet Investor Update, Google
After announcing the Gemini 2.0 model (see below for details), which offers better performance than previous versions, Alphabet’s stock price hit an all-time high. Upcoming projects like Project Mariner and the Jules coding agent, as well as the Willow quantum computing chip, which outperforms today’s supercomputers, demonstrate a robust roadmap for Alphabet. Market excitement reflects confidence in Alphabet’s leadership in advanced AI and quantum technologies.
Google DeepMind debuts Gemini 2.0, OpenAI announces Sora
Source: Philipp Schmid (LinkedIn), Google, Sora official website
Google DeepMind has released Gemini 2.0, which runs twice as fast as Gemini 1.5 Pro and has highly accurate scores. This marks a major advance in the search giant’s pursuit of agent AI, moving beyond simply organizing information to creating AI that can actively assist users. Currently available for testing is Gemini 2.0 flash, an experimental model designed for low latency and enhanced performance. As well as outperforming its predecessor in terms of speed, it introduces several new features including the ability to handle multimodal input such as images, video, audio, natively created images and manipulated text-to-speech audio. You can now also generate multimodal outputs such as . Additionally, Gemini 2.0 Flash natively supports the use of tools and integrates with Google search, code execution, and user-defined functions.
Meanwhile, OpenAI announced Sora, a model that supports high-precision text-to-video generation. The video below shows a sample of Sora’s features. The website created a 5-second video in about a minute with simple prompts related to AI agents.
Converge Bio and Remilk partner on LLM-based protein production
Source: Dub Gertz (LinkedIn)
Converge Bio announced a partnership with Remilk to employ its LLM-based GenAI platform to develop protein alternatives to traditional dairy products. This collaboration highlights the role of AI-driven biotechnology in sustainable food production, enabling more efficient identification and synthesis of novel protein structures.
Thibault Joouy
Podcast series “Tech and Drugs” featuring Kiin AI
Source: Thibault GEOUI (LinkedIn)
The first episode of the “Tech and Drugs” podcast will launch on December 17th at 11am CET and will feature Kiin AI guests such as Filippo Abbondanza and Mark Davies. Hosted by Thibault Geoui, this series focuses on the intersection of technology and medicine, providing industry insights on AI-driven drug development. Geoui is at the heart of scientific analysis at Charles Riber Laboratories in Frankfurt.
Advances in AI-driven drug discovery from Insilico Medicine: Special feature on ISM5411
Source: Dr. Andriy Buvaylo (LinkedIn)
Insilico Medicine has achieved a milestone with ISM5411, an AI-designed PHD inhibitor targeting inflammatory bowel disease (IBD), now documented in Nature Biotechnology. The compound was developed using Insilico’s Pharma.AI and Chemistry42 production chemistry platforms, reflecting the accelerating impact of AI in drug development. We progressed from target selection to preclinical candidate in 12 months.
Acknowledgment
Dr. Andrii Buvailo, science and technology communicator and founder of BioPharmaTrend.com, and Frédéric Célerse, AI research scientist for chemistry at EPFL, for their efforts in identifying quality content related to AI in biopharmaceuticals. Special thanks to Drs. And beyond that.
