AI news on Wednesday, June 17, 2026

SIGNIFICANCE

★ Top story · Models6h ago

Google’s Gemini-Powered AI Home Speaker Goes on Sale June 25

Google will release its new $100 smart speaker on June 25, marking the first time the company has integrated Gemini generative AI directly into a home device. This launch represents a strategic shift for Google, as it attempts to move beyond the rigid command-and-control functionality of the Google Assistant era. By prioritizing conversational interaction over hardware audio improvements, the company is betting that generative AI can revitalize interest in the smart home category.

BBloomberg Technology TTechCrunch AI WWired AI Read original ↗

Wednesday — June 17, 2026

Google’s Gemini-Powered AI Home Speaker Goes on Sale June 25

GLM-5.2: Built for Long-Horizon Tasks

New research shows how AMIE, our medical AI, could help manage health conditions.

A near-autonomous AI chemist improves a challenging reaction in medicinal chemistry

Nvidia research shows robots that train themselves through AI coding agents

OpenAI researchers want to predict how often AI models will fail before launch

Anthropic CEO Amodei on Compute Spend, Surpassing OpenAI

US Order on Anthropic Models Signals New Era for AI Controls

Hyperscalers may soon be unable to fund their AI buildout from cash flow alone

MolmoMotion: Language-guided 3D motion forecasting

From the Hugging Face Hub to robot hardware with Strands Agents and LeRobot

Europe frets about U.S. AI as tech world flocks to France for G7, VivaTech

Curiosity-Critic: Cumulative Prediction Error Improvement as a Tractable Intrinsic Reward for World Model Training

Mordal: Automated Pretrained Model Selection for Vision Language Models

HeteRo-Select: Informativeness as the Participation Driver in Heterogeneous Federated Learning

Branch-and-Browse: Efficient and Controllable Web Exploration with Tree-Structured Reasoning and Action Memory

ThinkJEPA: Empowering Latent World Models with Large Vision-Language Reasoning Model

A Benchmark for Omni-Modal Reasoning in Long Videos

When LLMs Analyze Scars: From Images to Clinically-Meaningful Features

CausalT5k: Diagnosing Refusal and Failure Modes in Trustworthy Causal Reasoning Across Causal Rungs

How Musicians Can Get Paid for Training AI

DPRM: A Plug-in Doob h transform-induced Token-Ordering Module for Diffusion Language Models

RAIGen: Rare Attribute Identification in Text-to-Image Generative Models

DVD: Discrete Voxel Diffusion for 3D Generation and Editing

Learn from Your Mistakes: Self-Correcting Masked Diffusion Models

Beyond MACs: Hardware Efficient Architecture Design for Vision Backbones

LVLMs and Humans Ground Differently in Referential Communication

Brep2Shape: Boundary and Shape Representation Alignment via Self-Supervised Transformers

Learn-To-Learn on Arbitrary Textual Conditioning: A Hypernetwork-Driven Meta-Gated LLM

A Recipe for Long-Context Reasoning in Large Language Models via On-Policy Optimization and Distillation

ASML CEO on AI Demand, Data Centers in Space and Musk's Terafab

Olmo Hybrid: From Theory to Practice and Back

States Are Now Writing the Workplace AI Rules - Employment Law This Week®

Disentangling Perception and Reasoning in Multimodal LLMs via Reward Design

Atlas: Orchestrating Heterogeneous Models and Tools for Multi-Domain Complex Reasoning

NeuroClaw Technical Report

Gradual Fine-Tuning for Flow Matching Models

Noise-Driven Exploration and Transient Freezing Select Flat Minima in Stochastic Gradient Descent

AnalogFed: Privacy-Preserving Discovery of Analog Circuits at Scale with Federated Generative AI

Detail++: Training-Free Detail Enhancer for Text-to-Image Diffusion Models

Regression Language Models for Code

Only 16 percent of Americans think AI will have a positive impact on society, a new study shows

BofA Public Finance Head Sees AI Boosting Underwriting Dominance

Rethinking Cross-Layer Information Routing in Diffusion Transformers

AIMER: Calibration-Free Task-Agnostic MoE Expert Pruning

PACE-RAG: Patient-Aware Contextual and Evidence-Constrained RAG for Clinical Drug Recommendation

TextMesh4D: Zero-shot Text-to-4D Mesh Generation

PLATE: Plasticity-Tunable Efficient Adapters for Geometry-Aware Continual Learning

MoSE: Mixture of Slimmable Experts for Efficient and Adaptive Language Models

Learning to Decide with AI Assistance under Human-Alignment

Adversarial Attacks Leverage Interference Between Features in Superposition

Multiple JetBrains IDE plugins caught stealing AI keys

TRACE: Learning to Compute on Circuit Graphs

When Tables Go Crazy: Evaluating Multimodal Models on French Financial Documents

Principled RL for Flow Matching Emerges from the Chunk-level Policy Optimization

m2sv: A Scalable Benchmark for Map-to-Street-View Spatial Reasoning

ALAS: An Automatic Latent Alignment Score for Audio Language Models

Riemann-Bench: A Benchmark for Moonshot Mathematics

MedicalAgentsBench for Complex Medical Reasoning: Comparing Internalized Reasoning Models versus Externalized Agent-based Frameworks

DiffAttn: Diffusion-Based Drivers' Visual Attention Prediction with LLM-Enhanced Semantic Reasoning

ECB’s Lagarde Warns AI Is Huge Risk for Financial Stability

Canadian pension giant joins race to fund India’s AI-fueled data center boom

Bridging Modality Disconnect in Self-Reflection via Closed-Loop Visually Grounded Verification

Advances in 4D Representation: Geometry, Motion, and Interaction

Fluently Lying: Adversarial Robustness Can Be Substrate-Dependent

AI Health Startup Wants to Assist Half of Latin American Doctors

Conformal Path Reasoning: Trustworthy Knowledge Graph Question Answering via Path-Level Calibration

Partial Ring Scan: Revisiting Scan Order in Vision State Space Models

The White House Wants Anthropic to Block All Jailbreaks. That May Not Be Possible

Vercel Releases Eve: An Open-Source AI Agent Framework Where Each Agent is a Directory of Files Mapped to Capabilities

SceneCompleter: Dense 3D Scene Completion for Generative Novel View Synthesis

DICE: Diffusion Large Language Models Excel at Generating CUDA Kernels

Clarify Before You Draw: Proactive Agents for Robust Text-to-CAD Generation

Provably Efficient Regularized Online RLHF with Generalized Bilinear Preferences

Critique of World Model: A Generative Latent Prediction Architecture for World Modeling

Pramaana Labs raises $27M seed round from Khosla Ventures to bring formal verification to AI

Flash-GRPO: Efficient Alignment for Video Diffusion via One-Step Policy Optimization

EngTrace: A Symbolic Benchmark for Verifiable Process Supervision of Engineering Reasoning

Adaptive Activation Steering for Efficient LLM Reasoning via Closed-Loop PID Control