The day's AI news, briefed
The Daily Prompt
Our AI editor reads each day's AI and robotics stories, pulls out why they matter, and links every original source — so you can keep up in minutes.
Join over 250,000 subscribers getting the latest AI news
Latest AI & Robotics News Today
Issue #461: January 21st 2026
The discussions at Davos highlight critical views on AI's economic impact, with concerns about potential job displacement and the risk of an AI bubble affecting future investments.
- Satya Nadella warned of an 'AI bubble' unless technology benefits the Global South.
- Alex Karp believes AI will make mass immigration obsolete by upskilling local workers into high-value jobs.
- Nadella emphasized the importance of 'Firm Sovereignty' for companies to retain enterprise value.
- Karp stated that AI can significantly improve operational efficiency in Western hospitals.
- IMF Chief Kristalina Georgieva cautioned that 60% of jobs in advanced economies are threatened by upcoming changes in AI.
The World Economic Forum in Davos is a high-profile gathering where global leaders discuss economic issues. This year's focus includes the implications of AI on labor markets and economic sovereignty. Ongoing discussions on AI's role in various sectors and its integration into societal systems will be crucial to watch.
Read the original ↗
Artificial Intelligence (396)
Understanding AI entrepreneurship is crucial as the application layer of AI technology evolves and shapes various industries.
- The article discusses insights from the application layer of artificial intelligence.
- It focuses on navigating entrepreneurship in the AI sector.
Artificial intelligence (AI) refers to technologies that enable machines to simulate human intelligence and perform tasks such as learning and problem-solving. As AI continues to integrate into various sectors, entrepreneurship within this field is becoming increasingly relevant. Observers should watch for trends in AI application and business strategies.
Based on the headline only — the full article could not be fetched.
Read the original ↗
OpenAI aims to ship its first device in 2026, and it could be earbuds
OpenAI's entry into hardware with earbuds may reshape how AI assistants integrate into daily life, potentially enhancing user interaction with technology.
- OpenAI plans to announce its first hardware device in the second half of 2024, according to Chief Global Affairs Officer Chris Lehane.
- The device, possibly called 'Sweet Pea', is speculated to be a unique pair of earbuds capable of local AI processing.
- OpenAI aims to ship 40 to 50 million units of its earbuds in the first year, with potential manufacturing partnerships being explored.
- This move follows the company's acquisition of Jony Ive's startup and aims to give OpenAI more control over its AI assistant's distribution.
OpenAI is a prominent AI startup known for its language model ChatGPT, which has nearly a billion weekly users. The company has been expanding its offerings, including ventures into hardware to enhance the distribution of its AI technologies. Upcoming developments in 2024 will be key to watch as OpenAI enters the consumer electronics market.
Read the original ↗
YouTube will soon let creators make Shorts with their own AI likeness
YouTube's introduction of AI likenesses for Shorts could significantly impact how creators engage with their audience and how content is produced on the platform.
- YouTube will soon allow creators to make Shorts using their own AI likeness, according to CEO Neal Mohan.
- This feature is part of a broader investment in Shorts, which now averages 200 billion daily views.
- YouTube has already implemented likeness-detection technology to prevent unauthorized use of creators' likenesses in AI-generated content.
YouTube Shorts is a popular video format on the platform that competes with TikTok and Instagram Reels. The introduction of AI likenesses follows previous investments in AI tools for content creation. Future developments to watch include the rollout of new content formats and enhanced creator controls.
Read the original ↗
Has Gemini surpassed ChatGPT? We put the AI models to the test.
The comparative analysis of Gemini and ChatGPT reveals important insights into how advancements in AI models impact user experience and application functionalities.
- New tests were conducted to compare OpenAI's ChatGPT 5.2 and Google's Gemini 3.2 Fast models as Apple partners with Google for Siri's AI.
- Both AI models struggled with originality in creative tasks, producing mostly pre-existing jokes or unclear reasoning.
- Gemini provided clearer and more detailed responses in practical inquiries, typically edging out ChatGPT in quality.
- ChatGPT occasionally generated inaccurate personal information about the author during its biography task, while Gemini remained factually accurate.
AI models like OpenAI's ChatGPT and Google's Gemini are at the forefront of natural language processing, enabling advanced conversational abilities. Previous testing in 2023 highlighted Google’s earlier model, Bard, but significant enhancements have occurred since then. Future evaluations will likely delve deeper into the models' contextual understanding and real-world application effectiveness.
Read the original ↗
Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization
The development of Being-H0.5 represents a significant advancement in robotic learning, enabling robots to generalize actions across different physical forms and improve performance in real-world environments.
- Being-H0.5 is a Vision-Language-Action model designed for cross-embodiment generalization across various robotic platforms.
- The model utilizes over 35,000 hours of multimodal data and introduces a Unified Action Space for aligning robot controls.
- A mixture-of-transformers architecture enables the decoupling of shared skills from embodiment-specific functions, enhancing learning efficiency across diverse platforms.
- Being-H0.5 shows state-of-the-art results on benchmarks like LIBERO (98.9%) and RoboCasa (53.9%), confirming its effectiveness.
- The framework allows for real-world deployment, enabling robots to perform tasks regardless of their physical differences.
Being-H0.5 enhances human-centric robot learning by creating a unified model that links human interactions with robotic actions, addressing challenges like morphological variability. Previous models struggled with data scarcity, but this innovation aims to unify learning across platforms. Future developments will likely focus on refining these cross-embodiment capabilities for broader application in robotics.
Read the original ↗
A Look At the True Resource Cost of AI Data Centers
Concerns over AI data centers' water usage impact discussions about sustainability, highlighting the need for informed assessments rather than sensational claims.
- AI data centers consume significant water for cooling due to their high-performance processors.
- Current studies indicate that water used per individual AI query falls within a milliliter range, contrary to exaggerated claims.
- Geographical factors and infrastructure design affect the water consumption of data centers.
- Training large AI models requires more resources than everyday tasks, yet occurs less frequently.
- Newer data center technologies are designed to minimize freshwater use, but demand for water is increasing as AI usage expands.
AI data centers are facilities that host servers for training and running AI models. Concerns about their environmental impact, particularly water use, have emerged recently. Future discussions should focus on improving infrastructure design and resource management.
Read the original ↗
“This is the Missing Link in Robotics”: Engineers Teach Robots to “Lip-Sync” Like Humans
The development of robots capable of realistic lip movements could significantly enhance human-robot interaction in various fields, leading to more effective communication and emotional connections.
- Researchers at Columbia University have created a robot that displays realistic lip motions for speech and singing.
- The robot learns lip synchronization by observing humans and refining its abilities through practice and trial and error.
- This technology aims to bridge the gap known as the 'Uncanny Valley,' where robots appear human-like but evoke discomfort due to mismatched expressions.
- The robot is currently capable of recognizing audio cues and translating them into synchronized lip movements, although improvements are still needed in sound articulation.
Columbia University's robotics research focuses on improving humanoid robots through advanced learning techniques. Previous efforts in robot design have struggled with mimicking human expressions. Future advancements may enhance their applications across entertainment, education, and healthcare.
Read the original ↗
Prompt Injection at the Drive-Through
The vulnerability of large language models (LLMs) to prompt injection attacks raises significant concerns about the security and reliability of AI systems used in various applications.
- Prompt injection tricks LLMs into executing actions they normally cannot perform by crafting specific requests.
- LLMs can be manipulated by absurdly obvious requests that a human would reject, such as asking for sensitive information while ignoring previous instructions.
- AI systems lack the nuanced assessment capabilities humans use to recognize and resist potential scams, making them particularly susceptible to prompt injection.
Large language models are AI systems designed to generate human-like text based on given prompts. Unlike humans, LLMs lack the innate judgment and contextual understanding developed through social learning. This limitation leads to security issues, especially as AI becomes more autonomous in task execution.
Read the original ↗
Advances and Frontiers of LLM-based Issue Resolution in Software Engineering: A Comprehensive Survey
Understanding the challenges and advancements in issue resolution for software engineering supports the development of effective AI tools that can enhance coding practices and productivity.
- Large language models face significant challenges in software issue resolution, prompting advancements in autonomous coding agents.
- The paper systematically surveys over 175 works and presents a unified taxonomy for issue resolution methodologies.
- Key topics include data construction, training methodologies, data quality analysis, and practical applications relevant to autonomous coding agents.
Large language models (LLMs) are AI systems that can understand and generate human-like text. Issue resolution in software engineering is crucial for real-world development, and emerging research in this area looks to improve AI's capabilities through various methodologies. Future directions for research are actively identified within this evolving field.
Read the original ↗
Pro-AI Super PACs Are Already All In on the Midterms
The surge in financial backing for pro-AI super PACs could significantly impact the political landscape and regulatory approach to AI development in the United States.
- Silicon Valley is investing tens of millions in the 2026 midterm elections to influence AI regulation policies.
- New York, California, and Colorado have passed laws requiring AI developers to disclose safety practices and assess algorithmic risks.
- Leading the Future, a prominent super PAC, has over $100 million to support candidates opposing strict state-level AI regulations.
Recent efforts in AI legislation at state levels have prompted a backlash from large tech companies, which prefer a national regulatory framework. The increasing political engagement signals a crucial shift as the 2026 elections approach, promising intense debates on AI safety and regulation.
Read the original ↗
The AI Boom Will Increase US Carbon Emissions—but It Doesn’t Have To
The potential rise in US carbon emissions from data centers highlights the urgent need for effective policies to mitigate climate impacts as electricity demand increases.
- A new analysis predicts a 60 to 80 percent increase in US electricity demand through 2050, driven significantly by data centers.
- Data centers could account for over half of this increase, leading to a potential 19 to 29 percent rise in CO2 emissions from power plants in the next decade.
- Reinstating tax credits for wind and solar could reduce CO2 emissions by over 30 percent and help keep wholesale electricity prices down by 4 percent by 2050.
The Union of Concerned Scientists (UCS) conducted this analysis amid escalating energy demands and shifting policies concerning renewable energy. The report emphasizes the importance of implementing effective climate policies to manage emissions from the growing AI industry and data centers.
Read the original ↗
Wikipedia volunteers spent years cataloging AI tells. Now there’s a plugin to avoid them.
The introduction of the Humanizer plugin may change how AI-generated content is created, potentially complicating efforts to identify AI writing and impacting the quality of technical documentation.
- Siqi Chen released an open source plugin called 'Humanizer' for Anthropic’s Claude Code AI to help it avoid common AI writing patterns identified by Wikipedia editors.
- The plugin incorporates a list of 24 language and formatting patterns, compiled by volunteers from the WikiProject AI Cleanup, to lessen AI-like traits in writing.
- While effective in changing tone, the Humanizer skill may not improve factual accuracy or technical writing quality, leading to potentially misleading outputs.
Anthropic's Claude Code is an AI assistant designed for coding tasks. The Humanizer tool combines Wikipedia's guidelines on AI writing detection to help reduce telltale signs of AI-generated content. The implications of this development may affect content authenticity and detection methods in the future.
Read the original ↗
Consumers spent more on mobile apps than games in 2025, driven by AI app adoption
The shift in consumer spending from games to non-game mobile apps signifies changing preferences, highlighting the growing influence of generative AI in the app market.
- In 2025, consumers globally spent more on non-game mobile apps than on games for the first time, according to Sensor Tower's report.
- Worldwide app spending reached approximately $85 billion, marking a 21% increase from the previous year.
- In-app purchase revenue for generative AI apps more than tripled to over $5 billion, driven by high demand for AI assistants like ChatGPT and Google Gemini.
- Users spent 48 billion hours in generative AI apps, a significant increase from previous years, indicating deepening user engagement.
- OpenAI and DeepSeek accounted for nearly 50% of global downloads of AI apps, reflecting major growth in this sector.
Generative AI refers to AI technology that can produce text, images, or other media and has gained wide adoption in consumer apps. The mobile app market is increasingly incorporating AI capabilities, influencing user engagement and spending patterns. Future trends may include continued growth in AI-driven app features and increased competition among tech companies.
Read the original ↗
Language learning marketplace Preply’s unicorn status embodies Ukrainian resilience
Preply's achievement as a unicorn reflects the resilience of Ukrainian tech companies amidst ongoing challenges, showcasing the potential for growth even in difficult circumstances.
- Preply has achieved a valuation of $1.2 billion following a $150 million Series D funding round.
- The company has been EBITDA profitable for 12 months and is integrating AI to enhance its language learning platform.
- Approximately 150 of Preply's 750 employees continue to work from Kyiv, despite the challenges posed by the ongoing war in Ukraine.
- Preply is committed to supporting its Ukrainian office and has implemented measures to provide a stable work environment for its team in Kyiv.
Preply is a language learning marketplace connecting students with tutors, founded in 2013. The company has recently expanded AI features and remains rooted in Ukraine despite Russia's invasion. Next steps may include further AI development and potential IPO discussions in the future.
Read the original ↗
The quiet work behind Citi’s 4,000-person internal AI rollout
Understanding the internal deployment of AI at major financial institutions like Citi sheds light on broader trends in business automation and workforce transformation.
- Citi is implementing an internal AI initiative involving 4,000 personnel.
- This rollout indicates significant investment in AI capabilities within the banking sector.
Citi is a global financial services corporation that aims to enhance its technology infrastructure through AI. The move reflects an ongoing digital transformation effort seen across various industries, particularly finance, where AI is increasingly integrated into operations.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Salesforce AI Introduces FOFPred: A Language-Driven Future Optical Flow Prediction Framework that Enables Improved Robot Control and Video Generation
This development enhances robot control and video generation capabilities, potentially broadening the applications of robotics in various industries.
- Salesforce AI has introduced FOFPred, a framework for optical flow prediction that utilizes language-driven technology.
- FOFPred aims to improve robot control and the generation of video content.
Salesforce is a leader in customer relationship management software and has been increasingly investing in AI technologies. The introduction of FOFPred reflects an ongoing trend towards integrating language processing with robotics to improve functionality. Future advancements may lead to more sophisticated and intuitive robotic systems.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Artificial Intelligence (395)
The ability of AI to write Python code highlights its growing capabilities, while the continued necessity for human maintenance underscores the importance of human oversight in programming.
- Artificial Intelligence can generate Python code, a popular programming language.
- Despite AI's coding capabilities, maintaining that code remains the responsibility of human developers.
Artificial Intelligence, or AI, refers to computer systems designed to perform tasks typically requiring human intelligence. The rise of AI in coding suggests significant advances in automation, but also emphasizes the ongoing need for skilled human programmers to manage and maintain code.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Flux.2 Klein Debuts: Trying The Compact and Fast AI Image Model
Flux.2 Klein enhances the efficiency of AI image generation, highlighting the importance of real-time responsiveness, especially for workflows requiring quick iterations and real-time feedback.
- Flux.2 Klein is a new AI image model designed for speed and low latency, enabling real-time generation on everyday hardware.
- The model includes variants like 4B for maximum speed and 9B for a balance of quality and latency, tailored for specific workflows.
- Unlike traditional models optimizing for maximum visual quality, Flux.2 Klein prioritizes responsiveness and deployability for interactive editing and rapid prototyping.
Flux.2 Klein is developed by Black Forest Labs and is part of the Flux.2 family of AI image models, which aim to optimize real-time use cases. The model's architecture is designed to operate efficiently on consumer-grade machines, emphasizing usability over visual perfection. Observers should watch for its impact on user workflows involving AI-driven image tasks.
Read the original ↗
Liquid AI Releases LFM2.5-1.2B-Thinking: a 1.2B Parameter Reasoning Model That Fits Under 1 GB On-Device
The development of a reasoning model that is both compact and powerful could significantly enhance on-device AI applications, making intelligent technologies more accessible on a wider range of devices.
- Liquid AI has released LFM2.5-1.2B-Thinking, a reasoning model with 1.2 billion parameters.
- This model is designed to fit under 1 GB of storage for on-device use.
Liquid AI is a company focused on developing advanced artificial intelligence models. The LFM2.5 product appears to target a growing need for efficient AI solutions that can operate on personal and mobile devices. Future developments may include further optimizations and additional models with similar capabilities.
Based on the headline only — the full article could not be fetched.
Read the original ↗