AI research

Frustrated authors withdraw papers after realizing their reviewers are just lazy LLMs

AI research

Gemini 3 Pro tops new AI reliability benchmark, but hallucination rates remain high

AI research

Researchers push "Context Engineering 2.0" as the road to lifelong AI memory

AI research AI and society AI in practice
Ad
AI in practice
Short

OpenAI is bringing ChatGPT Voice directly into the main text chat, making it easier to switch between speaking and typing without jumping into a separate mode. Users can talk naturally, see responses as text, revisit earlier messages, and view visual content like images or maps without losing context. The update is rolling out on mobile and the web. Anyone who prefers the old setup can still turn the standalone voice mode back on in the settings under "Voice Mode."

Video: OpenAI

ChatGPT's Advanced Voice Mode has been available since September 2024. In June 2025, OpenAI boosted the voice system's expressiveness and added real-time translation.

AI in practice
Short

Claude Opus 4.5 scores higher than its rivals in prompt-injection security, but the results show how limited these defenses still are. A benchmark by the security firm Gray Swan found that a single "very strong" prompt injection attack breaks through Opus 4.5's safeguards 4.7 percent of the time. Give an attacker ten attempts and the success rate jumps to 33.6 percent. At 100 attempts, it reaches 63 percent. Even with those gaps, Opus 4.5 still performs better than models like Google's Gemini 3 Pro and GPT-5.1, which show attack rates as high as 92 percent.

Anthropic

Prompt injection works by slipping hidden instructions into a prompt to bypass safety filters, a long-standing weakness in large language models. The issue becomes even more serious in agent-style systems, which expose more potential entry points and make these attacks easier to exploit.

Ad
Ad
AI and society
Short

Google is in talks with Meta and several other companies about letting them run Google's TPU chips inside their own data centers, according to a report from The Information. One person familiar with the discussions said Meta is considering spending billions of dollars on Google TPUs that would start running in Meta facilities in 2027. Until now, Google has only offered its TPUs through Google Cloud.

The new TPU@Premises program is Google's attempt to make its chips a more appealing alternative to Nvidia's AI hardware. According to someone with knowledge of internal comments, Google Cloud executives have said the effort could help the company reach ten percent of Nvidia's annual revenue. Google has also built new software designed to make TPUs easier to use.

AI research
Short

US President Donald Trump signed an order on Monday to launch a shared AI platform for federal research data. Called the Genesis Mission, the effort aims to make large datasets from federal agencies usable for new AI models, according to White House adviser Michael Kratsios.

The Department of Energy will link its supercomputers, research datasets, and automated lab systems through the new platform. Kratsios said the goal is to have AI plan experiments, speed up simulations, and generate predictions on topics like protein structures and plasma behavior.

Energy Secretary Chris Wright pointed to the surge in private AI investment but argued that more of that momentum needs to shift toward scientific and technical research. He said the data held by federal labs is essential for that work. The order also highlights priority areas including biotechnology, space, energy, and semiconductor research.

Ad
Ad
AI in practice
Short

Amazon has announced a major investment in its AI footprint for federal work, saying it will spend up to $50 billion to expand AI and supercomputing infrastructure for U.S. government agencies. The project supports the U.S. government’s AI Action Plan and is expected to help agencies accelerate discovery, decision-making, and mission workflows, including through faster analysis and automation.

Amazon’s investment underscores the strategic importance of AI and supercomputing in maintaining technological superiority, safeguarding critical infrastructure, and driving industrial innovation.

Starting in 2026, AWS plans to add nearly 1.3 gigawatts of new compute capacity across its Top Secret, Secret, and GovCloud (US) regions. Once live, agencies will be able to use services such as SageMaker, Bedrock, Trainium chips, and Anthropic models to build their own AI applications, speeding up data analysis and improving workflows in areas like cybersecurity, healthcare research, and autonomous systems.

Ad
Ad
AI research
Short

Google has added a slide generator to NotebookLM, giving users a quick way to turn their sources into simple slide decks. The tool can help structure notes or produce early drafts, and Google says it can also enhance existing slides visually.

Right now, NotebookLM delivers slides only as PDFs. Export options for Google Slides and PowerPoint are in development, Google says. The feature is available immediately, with daily usage limits based on the user's account.

The slide tool, along with a new infographic feature, runs on Google's Nano Bana Pro model (Gemini 3 Pro Image Generation). It is the first model capable of turning highly detailed prompts into precise, text-heavy images.

Google News

Bank details

IBAN: DE88 2507 0070 0053 0014 00
BIC: DEUTDE2HXXX
Account holder: Deep Content GmbH
Purpose: Support THE DECODER
AI research

Frustrated authors withdraw papers after realizing their reviewers are just lazy LLMs

AI research

Gemini 3 Pro tops new AI reliability benchmark, but hallucination rates remain high

AI research

Researchers push "Context Engineering 2.0" as the road to lifelong AI memory

Google News