News TodAI

AI models can acquire backdoors from surprisingly few malicious documents

Anthropic study suggests "poison" training attacks don't scale with model size.

Bank of England warns AI stock bubble rivals 2000 dotcom peak

Central bank says market concentration hasn't been this extreme in 50 years.

Vandals deface ads for AI necklaces that listen to all your conversations

Critics attacked subway ads to defend human friends and broadly criticize AI.

Insurers balk at paying out huge settlements for claims against AI firms

OpenAI, Anthropic consider using investor funds to settle potential lawsuits.

Ars Live: Is the AI bubble about to pop? Ed Zitron is on with Ars at 3:30pm EDT today

Our discussion happens today, October 7, at 3:30pm US Eastern time.

Dead celebrities are apparently fair game for Sora 2 video manipulation

OpenAI's likeness protections for living "public figures" don't apply to "historical figures."

OpenAI wants to make ChatGPT into a universal app frontend

Spotify, Canva, Zillow among today's launch partners, more coming later this year.

Deloitte will refund Australian government for AI hallucination-filled report

Consulting firm quietly admitted to GPT-4o use after fake citations were found in August.

AMD wins massive AI chip deal from OpenAI with stock sweetener

ChatGPT maker will be allowed to buy 10% of AMD for a penny per share.

OpenAI, Jony Ive struggle with technical details on secretive new AI gadget

ChatGPT maker working with former Apple design boss to launch personal assistant.

A biological 0-day? Threat-screening tools may miss AI-designed proteins.

Ordering DNA for AI-designed toxins doesn't always raise red flags.

Why iRobots founder wont go within 10 feet of todays walking robots

Rodney Brooks says humanoid robots pose hidden safety challenges and won't learn dexterity from video alone.

Meta wont allow users to opt out of targeted ads based on AI chats

US users stuck with AI ad targeting as EU users win more control over their feeds.

OpenAI mocks Musks math in suit over iPhone/ChatGPT integration

xAI's claim that Apple gave ChatGPT a monopoly on prompts is baseless, OpenAI says.

OpenAIs Sora 2 lets users insert themselves into AI videos with sound

Sora social app launches with deepfake-style "cameos" and feed controls.

Can todays AI video models accurately model how the real world works?

New research shows highly inconsistent performance on a variety of physical reasoning tasks.

Googles Gemini-powered smart home revamp is here with a new app and cameras

Google promises a better smart home experience thanks to Gemini.

The AI slop drops right from the top, as Trump posts vulgar deepfake of opponents

A sombrero and a fake mustache were also involved.

Alexas survival hinges on you buying more expensive Amazon devices

Echo speakers and displays for Alexa+ require more expensive components.

Critics slam OpenAIs parental controls, while users rage, Treat us like adults

OpenAI still isnt doing enough to protect teens, suicide prevention experts say.

DeepSeek tests sparse attention to slash AI processing costs

Chinese lab's v3.2 release explores a technique that could make running AI far less costly.

With new agent mode for Excel and Word, Microsoft touts vibe working

Agent Mode in Word, Excel works like vibe coding tools but for knowledge work.

Californias newly signed AI law just gave Big Tech exactly what it wanted

After the failure of S.B. 1047, new AI disclosure law drops kill switch for disclosure mandate.

Burnout and Elon Musks politics spark exodus from senior xAI, Tesla staff

Disillusionment with Musk's activism, strategic pivots, and mass layoffs cause churn.

Anthropic says its new AI model maintained focus for 30 hours on multistep tasks

New Claude Sonnet 4.5 beats OpenAI and Google on coding tests.

Big AI firms pump money into world models as LLM advances slow

World models aim to navigate the physical world by learning from videos, robotic data.

Why LA Comic Con thought making an AI-powered Stan Lee hologram was a good idea

I suppose if we do it and thousands of fans... don't like it, we'll stop doing it.

Can AI detect hedgehogs from space? Maybe if you find brambles first.

Cambridge researchers use satellite-based bramble detection as a proxy for mapping hedgehog habitats.

Adobe encourages candidates to use AI during the application process.Stacy Martinet, the company's chief communications officer, toldFortunethat work samples or designs created with AI tend to stand out. According to Martinet, the company is searching for people who combine ...

Mark Zuckerberg is still on his AI talent shopping tour.Andrew Tulloch, co-founder of AI startup Thinking Machines Lab, is leaving the company to join Meta, theWall Street Journalreports. Tulloch informed colleagues of his decision on Friday. A spokesperson for Thinking Machines ...

OpenAI could soon face a billion-dollar fine.Authors and publishers suing the company for copyright infringement have uncovered internal messages and emails about the deletion of a dataset containing pirated books. The plaintiffs now want access to communications between OpenAI a...

Chinese researchers let LLMs share meaning through internal memory instead of text

Open ASR Leaderboard tests more than 60 speech recognition models for accuracy and speed

Google Deepmind's "Vibe Checker" aims to rate AI code by human standards

A new information-theory framework reveals when multi-agent AI systems truly work as a team

OpenAI accused of pressuring AI regulation advocates with subpoenas

Google boasts 1.3 quadrillion tokens each month, but the figure is mostly window dressing

Anthropic finds 250 poisoned documents are enough to backdoor large language models

Claude Levels Up, Qwen3 Proliferates, Big AI Diversifies Product Lines, LoRA Adapters on Tap

Claude Levels Up, Qwen3 Proliferates, Big AI Diversifies Product Lines, LoRA Adapters on Tap

Reinforcement Learning Heats Up, White House Orders Muscular AI Policy, Computer Use Gains Momentum, Fine Control of Fine-Tuning

Reinforcement Learning Heats Up, White House Orders Muscular AI Policy, Computer Use Gains Momentum, Fine Control of Fine-Tuning

DeepSeek Sharpens Its Reasoning: DeepSeek-R1, an affordable rival to OpenAI''s o1

DeepSeek Sharpens Its Reasoning: DeepSeek-R1, an affordable rival to OpenAI''s o1

Tumbling Training Costs, Desktop AI Supercomputer, Tighter AI Export Restrictions, Improved Contrastive Loss

Tumbling Training Costs, Desktop AI Supercomputer, Tighter AI Export Restrictions, Improved Contrastive Loss

OpenAI''s Trillion-Dollar Bet, Generating Viruses, Modeling Planet Earth, Paying for Training Data

OpenAI''s Trillion-Dollar Bet, Generating Viruses, Modeling Planet Earth, Paying for Training Data

AI Agents Spend Money, Online Betting Automates, ChatGPT Users Shift, Reinforcement Learning Accelerates

AI Agents Spend Money, Online Betting Automates, ChatGPT Users Shift, Reinforcement Learning Accelerates

Drone Swarms Go To War, States Ban AI Mental-Health Treatments, Qwen3-Next Accelerates, Transformers Get Energized

Drone Swarms Go To War, States Ban AI Mental-Health Treatments, Qwen3-Next Accelerates, Transformers Get Energized

Stronger Chatbot Guardrails, Weaker Google Monopoly, AI-Assisted Education, 10 Million Tokens of Context

Stronger Chatbot Guardrails, Weaker Google Monopoly, AI-Assisted Education, 10 Million Tokens of Context

Gemini''s Environmental Impact, China''s Emerging AI Hub, Chatbot Job Interviews, Security For Agents

Gemini''s Environmental Impact, China''s Emerging AI Hub, Chatbot Job Interviews, Security For Agents

AI-Powered Phones Get Proactive, Robot Antelope Joins Herd, LLM Environmental Impacts Get Measured

AI-Powered Phones Get Proactive, Robot Antelope Joins Herd, LLM Environmental Impacts Get Measured

China Questions Nvidia, When Models Memorize, Mixture of Video Experts, OpenAI & Oracle Join Forces

China Questions Nvidia, When Models Memorize, Mixture of Video Experts, OpenAI & Oracle Join Forces

GPT-5''s Rough Takeoff, AI Video Blockbusters, India''s Homegrown LLMs, Synthetic Data Generation

GPT-5''s Rough Takeoff, AI Video Blockbusters, India''s Homegrown LLMs, Synthetic Data Generation

Open Agentic LLMs Proliferate, Robot Removes Gallbladders, Reasoning Models Boost Emissions, OpenAI Re-Opens

Open Agentic LLMs Proliferate, Robot Removes Gallbladders, Reasoning Models Boost Emissions, OpenAI Re-Opens

Trump Resets AI Policy, Qwen3''s Agentic Advance, U.S. Chips for China, The Trouble With AI Friends

Trump Resets AI Policy, Qwen3''s Agentic Advance, U.S. Chips for China, The Trouble With AI Friends

Power Moves in AI Coding, Moonshot''s Agentic LLM, How to Comply with EU AI Regs, AI Agents Evolve

Power Moves in AI Coding, Moonshot''s Agentic LLM, How to Comply with EU AI Regs, AI Agents Evolve

Grok Raises Questions, Meta Poaches Talent, California Reframes AI Regulations, Multi-Agent Systems Get Stronger

Grok Raises Questions, Meta Poaches Talent, California Reframes AI Regulations, Multi-Agent Systems Get Stronger

How to Make LLMs Commit Blackmail, Robotic Beehive, Walmart''s AI App Factory, Training Web Agents

How to Make LLMs Commit Blackmail, Robotic Beehive, Walmart''s AI App Factory, Training Web Agents

Amazon''s $100 Billion Bet, Meta''s Sensor-Packed Glasses, Anthropic''s Reason-Free Reasoning, Google''s Extreme Weather Prediction

Amazon''s $100 Billion Bet, Meta''s Sensor-Packed Glasses, Anthropic''s Reason-Free Reasoning, Google''s Extreme Weather Prediction

Judge Rules Training AI on Copyrighted Works Is Fair Use, Agentic Biology Evolves, Meta Befriends Alexandr Wang

Judge Rules Training AI on Copyrighted Works Is Fair Use, Agentic Biology Evolves, Meta Befriends Alexandr Wang

Nvidias AI empire: A look at its top startup investments

Thinking Machines Lab co-founder Andrew Tulloch heads to Meta

Its not too late for Apple to get AI right

Prezent raises $30 million to acquire AI services firms - starting with founders other company

Former UK Prime Minister Rishi Sunak to advise Microsoft and Anthropic

The world is just not quite ready for humanoids yet

Intel unveils new processor powered by its 18A semiconductor tech

Startup Battlefield company SpotitEarly trained dogs and AI to sniff out common cancers and will show off its tech at Disrupt

Googles virtual try-on shopping tool expands to more countries, now lets you try on shoes

Zendesk says its new AI agent can solve 80% of support issues

Zendesk says its new AI agent can solve 80% of support issues

DeepSeek-V3 New Paper is coming! Unveiling the Secrets of Low-Cost Large Model Training through Hardware-Aware Co-design

DeepSeek Unveils DeepSeek-Prover-V2: Advancing Neural Theorem Proving with Recursive Proof Search and a New Benchmark

Can GRPO be 10x Efficient? Kwai AIs SRPO Suggests Yes with SRPO

DeepSeek Signals Next-Gen R2 Model, Unveils Novel Approach to Scaling Inference with SPCT

Beyond Next-Token Prediction? Metas Novel Architectures Spark Debate on the Future of Large Language Models

Sunday Special: Scientists reverse Alzheimer's in mice

Robotics Special: Figure drops 03 humanoid robot

Google launches new AI platform for businesses

ALSO: How to create professional LinkedIn headshots with NanoBanana

Microsoft eyes licensing deal with Harvard Medical School

ALSO: How to build a real app in 10 minutes

Elon Musk's xAI rolls out AI video generation model

ALSO: How to design creatives with GenSpark AI Designer

OpenAI rolls out apps and agent platform

ALSO: How to turn ChatGPT into your personal tutor

OpenAI reportedly launching a no-code agent builder

ALSO: How to create a consultant quality strategy (in minutes)

Sunday Special: Eggs, skulls, and bone-healing guns

Robotics Special: DoorDash unveils new delivery robot

Perplexity makes Comet browser free, adds new features

ALSO: How to use Claude AI in your Slack workspace

OpenAI hits $500B valuation, seals Samsung chip deal

ALSO: How to make scroll-stopping video ads in minutes

OpenAI unveils Sora 2 with TikTok-style feed

ALSO: How to create an auto-updating slide deck with Manus

ZTE hosts Global Summit & User Congress 2025 in Milan under the theme "Expanding Intelligence, Creating Possibilities"

McKinsey wonders how to sell AI apps with no measurable benefits

Starlink is burning up one or two satellites a day in Earths atmosphere

Techie found an error message so rude the CEO of IBM apologized for it

No account? No Windows 11, Microsoft says as another loophole snaps shut

Leak suggests US government is fibbing over FEMA security failings

Employees regularly paste company secrets into ChatGPT

AI gets more 'meh' as you get to know it better, researchers discover

Intel's open source future in question as exec says he's done carrying the competition

UK Home Office opens wallet for GBP60M automated number plate project

Level-10 vuln lurking in Redis source code for 13 years could allow remote code execution

OpenAI will stop saving most ChatGPT users deleted chats

Court ends controversial order forcing OpenAI to save deleted ChatGPT logs.

Like putting on glasses for the first time-how AI improves earthquake detection

AI is comically good at detecting small earthquakes-heres why that matters.

OpenAI GPT-5: great taste, less filling, now with 30% less bias

Managers are throwing entry-level workers under the bus in race to adopt AI

Microsoft lets bosses spot teams that are dodging Copilot

Climate goals go up in smoke as US datacenters turn to coal

Microsoft hypes PCs with NPUs, still can't offer a good reason to buy one

It's trivially easy to poison LLMs into spitting out gibberish, says Anthropic

Amazon's Quick Suite is like agentic AI training wheels for enterprises

Google rearranges Agentspace into Gemini Enterprise

This is your brain on bots: AI interaction may hurt students more than it helps

Gartner warns agentic AI startups: Prepare to be consolidated

Humans flunk the Turing test for voices as bots get chattier

SoftBank snaps up ABB's robotics biz for $5.4B to fuel 'physical AI' dreams

Hobble your AI agents to prevent them from hurting you too badly

CoreWeave bets on serverless agent builder to woo penny-pinching enterprises

Ciscos new router unites disparate datacenters into AI training behemoths

Bank of England smells hint of dotcom bubble 2.0 in AI froth

IBM's big iron to get Spyre AI accelerator upgrade this month

Which Agent Causes Task Failures and When?Researchers from PSU and Duke explores automated failure attribution of LLM Multi-Agent Systems

In recent years, LLM Multi-Agent systems have garnered widespread attention for their collaborative approach to solving complex problems. However, its a common scenario for these systems to fail at a task despite a flurry of activity.

ByteDance Introduces Astra: A Dual-Model Architecture for Autonomous Robot Navigation

ByteDance introduces Astra, an innovative dual-model architecture revolutionizing robot navigation in complex indoor environments.

MIT Researchers Unveil SEAL: A New Step Towards Self-Improving AI

MIT introduces SEAL, a framework enabling large language models to self-edit and update their weights via reinforcement learning.

Researchers from PSU and Duke introduce Multi-Agent Systems Automated Failure Attribution

Automated failure attribution is a crucial component in the development lifecycle of Multi-Agent systems. It has the potential to transform the challenge of identifying what went wrong and who is to blame from a perplexing mystery into a quantifiable and analyzable problem

Adobe Research Unlocking Long-Term Memory in Video World Models with State-Space Models

By combining State-Space Models (SSMs) for efficient long-range dependency modeling with dense local attention for coherence, and using training strategies like diffusion forcing and frame local attention, researchers from Adobe Research successfully overcome the long-standing ch...

Zhipu.AIs Open-Source Power Play: Blazing-Fast GLM Models & Global Expansion Ahead of Potential IPO

Zhipu.AI open-sources faster GLM models (8x speedup), launches Z.ai, aiming for global expansion, potentially ahead of IPO.

AI Video Generation Race Shifts from Capability to Profitability, Challenging Soras Dominance

The AI video generation landscape is shifting from capability to profitability, challenging OpenAI Soras dominance. Competitors are surpassing Sora in quality and efficiency, with users preferring alternatives. The focus is now on improvements like precise control and style custo...

Nvidia Intensifies Robot Push with New Humanoid Platform as Industry Giants Eye Lucrative Future

Nvidia will launch Jetson Thor for humanoid robots in H1 2025, entering a growing market where Google is also active. The robotics sector is projected for substantial growth. Nvidia offers integrated hardware and software solutions. Simultaneously, Chinas rapidly developing domes...

Automating Artificial Life Discovery: The Power of Foundation Models

A research team introduces Automated Search for Artificial Life (ASAL). This novel framework leverages vision-language FMs to automate and enhance the discovery process in ALife research.

Llama 3 Meets MoE: Pioneering Low-Cost High-Performance AI

Researchers from the University of Texas at Austin and NVIDIA proposes upcycling approach, an innovative training recipe enables the development of an 8-Expert Top-2 MoE model using Llama 3-8B with less than 1% of the compute typically required for pre-training.

DeepMinds JetFormer: Unified Multimodal Models Without Modelling Constraints

A DeepMind research team introduces JetFormer, a Transformer designed to directly model raw data. This model maximizes the likelihood of raw data without depending on any pre-trained components, and is capable of both understanding and generating text and images seamlessly.

NVIDIAs nGPT: Revolutionizing Transformers with Hypersphere Representation

An NVIDIA research team proposes the normalized Transformer, which consolidates key findings in Transformer research under a unified framework, offering faster learning and reduced training steps-by factors ranging from 4 to 20 depending on sequence length.

From Token to Conceptual: Meta introduces Large Concept Models in Multilingual AI

A research team at Meta introduces the Large Concept Model (LCM), a novel architecture that processes input at a higher semantic level. This shift allows the LCM to achieve remarkable zero-shot generalization across languages, outperforming existing LLMs of comparable size.

NVIDIAs Hybrid: Combining Attention and State Space Models for Breakthrough Performance of Small Language Models

An NVIDIA research team proposes Hymba, a family of small language models that blend transformer attention with state space models, which outperforms the Llama-3.2-3B model with a 1.32% higher average accuracy, while reducing cache size by 11.67x and increasing throughput by 3.49...

From Response to Query: The Power of Reverse Thinking in Language Models

In a new paper Time-Reversal Provides Unsupervised Feedback to LLMs, a research team from Google DeepMind and Indian Institute of Science proposes Time Reversed Language Models (TRLMs), a framework that allows LLMs to reason in reverse-scoring and generating content in a manner o...

Yann LeCun Teams New Research: Revolutionizing Visual Navigation with Navigation World Models

In a new paper Navigation World Models, a research team from Meta, New York University and Berkeley AI Research proposes a Navigation World Model (NWM), a controllable video generation model that enables agents to simulate potential navigation plans and assess their feasibility b...

The Future of Vision AI: How Apples AIMV2 Leverages Images and Text to Lead the Pack

An Apple research team introduces AIMV2, a family of vision encoders that is designed to predict both image patches and text tokens within a unified sequence. This combined objective enables the model to excel in a range of tasks, such as image recognition, visual grounding, and ...

Redefining Music AI: The Power of Sonys SoniDo as a Versatile Foundation Model

In a new paper Music Foundation Model as Generic Booster for Music Downstream Tasks, a Sony research team presents SoniDo, a groundbreaking music foundation model that offers robust framework for improving the effectiveness and accessibility of music processing.

DeepMinds Socratic Learning with Language Games: The Path to Self-Improving Superintelligence

Researchers from Google DeepMind introduce the concept of Socratic learning. This refers to a form of recursive self-improvement in artificial intelligence that significantly enhances performance beyond the initial data or knowledge available to the system, as well as a practical...