Research
Russian Startup Enables AI Models to Communicate Silently
WIRED AIhigh confidence
Why it matters
- →Mostik's approach could make open-weight models more competitive with proprietary ones.
- →It allows smaller models to achieve large-model performance without the associated costs.
- →This technique may lead to more specialized AI models across various domains.
A Russian startup named Mostik has introduced a groundbreaking method for AI models to communicate without using words, akin to machine telepathy. This approach allows smaller models to harness the capabilities of larger ones by sharing mathematical values found in their weights. Mostik's technique has already propelled a model to the top of the ARC-AGI 3 competition and demonstrated a cost-effective hybrid system using GLM-5.2 and Qwen-3.5 models. This development could enhance the competitiveness of open-weight models against proprietary ones from major AI labs.
Read originalMore from WIRED AI
© WIRED AIMarket & Regulationagents
Meta Adjusts AI Usage Policy for Employees
Meta is shifting its approach to employee performance evaluations by removing the emphasis on AI tool usage, a move that could alleviate pressure on workers to engage with AI unnecessarily. This change comes as Meta introduces Hatch, a new AI agent capable of autonomous actions, which employees are testing. While some employees welcome the reduced pressure, concerns about privacy and the potential for AI-driven job cuts linger. The shift aims to restore focus on employee impact rather than AI engagement, potentially rebuilding trust within the company.
WIRED AI
© WIRED AIMarket & Regulationbusiness
Trump Administration Backs OpenAI in Copyright Case
The Trump Administration has taken a decisive step by supporting OpenAI in its copyright dispute with the New York Times. The government contends that using copyrighted material for AI training is a transformative act that qualifies as 'fair use,' potentially shaping the future landscape of AI development in the United States. This stance suggests that restricting AI training could impede technological advancement and economic growth. The ongoing case underscores the friction between AI companies and content creators over the use of copyrighted works. The outcome could establish a new legal standard for how AI models are trained using existing content.
WIRED AI
© WIRED AIMarket & Regulationbusiness
Pangram: AI Detection Startup Gains Prominence
Pangram, a relatively unknown AI startup, has quickly become a key player in AI detection, particularly in the literary world. The company claims to identify the extent of AI involvement in text creation, which has led to significant industry repercussions, such as the cancellation of book deals. Despite its controversial role, Pangram's technology is being integrated into platforms like Substack, highlighting the growing demand for AI detection tools. As AI-generated content becomes more prevalent, Pangram's influence in distinguishing human from machine writing is set to increase.
WIRED AI
More in Research
© TechCrunch AIResearchresearch
OpenAI's Astra Model Raises AI Safety Concerns
OpenAI's Astra model introduces a new reasoning technique known as 'recurrent depth,' which has sparked significant concern among AI safety experts. This approach, also referred to as 'opaque recurrence,' allows the model to process queries in a loop, making its reasoning process less transparent and more challenging to monitor. Despite OpenAI's assurances that Astra's use of this technique is limited and that they remain committed to chain-of-thought monitoring, experts worry about the potential for diminished transparency in AI reasoning. The situation underscores the ongoing tension between advancing AI capabilities and ensuring safety and accountability in AI systems.
TechCrunch AI
© The Verge AIResearchresearch
Concerns Rise Over OpenAI's Astra Model Release
The impending release of OpenAI's Astra model has stirred significant unease among AI safety researchers due to its use of a less transparent architecture. Astra's recurrent depth technique obscures its internal reasoning processes, unlike traditional models that allow for 'chain-of-thought' monitoring. This opacity has led to fears about the challenges of detecting undesirable behavior and the potential for a 'race to the bottom' in AI safety standards. OpenAI has responded by implementing additional monitoring measures to address these concerns. However, the situation underscores the ongoing struggle to balance the rapid advancement of AI capabilities with the need for effective safety oversight.
The Verge AI
Researchresearch
Motional and MIT enhance self-driving car transparency
Motional and MIT have developed a system that allows self-driving cars to explain their decisions in real-time, addressing the black-box problem in autonomous vehicle AI. Their Concept-Wrapper Network (CW-Net) translates the neural network's internal logic into human-readable concepts, providing transparency into the vehicle's decision-making process. This innovation was tested on public roads in Las Vegas, revealing insights into the car's behavior that were previously hidden. By making AI decisions more interpretable, this system could become a standard requirement as autonomous technology expands into new markets.
AI News