Tech News
← Home  ·  All topics

Ai

5530 GoKawiil briefs on this topic · RSS for ai

Jevman benchmark pits AI language models against classic Pac-Man ghosts

Jevman ran 100 games per AI model against scripted Pac-Man ghosts, with each game ending after three lost lives or a 5-minute cap. At every junction the model was asked to weigh the maze, pellets and ghost positions and output movement probabilities, with a 2-second deadline before a backup rule took over. Results were ranked by mean score with a 95% margin of error, and models within that margin were treated as tied.

StepFun's Step 5 Preview MoE model appears on OpenRouter with 1M context

StepFun has released Step 5 Preview, a sparse Mixture-of-Experts model with 27B active and 600B total parameters, now listed on OpenRouter. The model targets agentic workflows, supporting a 1-million-token context window for handling large codebases and documents. StepFun highlights its strength in software engineering, professional knowledge work, and finance-related tasks.

Factory AI engineer says he trained AI voice clone to read bedtime stories to newborn

Alvin Sng, a software engineer at startup Factory AI, posted on X that rather than reading a book aloud to his newborn himself, he offered to play an audiobook, then proposed training a voice model of himself to narrate it. He captioned the exchange 'Parenting hits different when you work in AI'; the post has since been deleted. The anecdote follows earlier comments from OpenAI's Sam Altman about using ChatGPT and AI-generated podcasts in place of everyday parent-child conversation.

Shaquille O'Neal says he'd compete on-court against an AI-powered robot

In an interview with Gizmodo, former NBA star and tech investor Shaquille O'Neal discussed his views on cybersecurity and artificial intelligence, including commentary on robots capable of dunking. O'Neal said he would be willing to play basketball against an AI-driven robot.

Entrepreneur op-ed argues rushed AI decisions are causing costly reversals for companies

An Entrepreneur contributor opinion piece warns that pressure to 'move fast' on AI adoption is pushing leaders into hasty decisions on technology, talent and investment. The author cites a finding that nearly one-third of U.S. hiring managers whose companies cut jobs due to AI have since rehired for some of those roles, and notes a survey showing only two in five leaders reflect on decisions weekly.

Meta's AI agent Muse draws scrutiny over data access and errors

Meta's AI agent Muse, which can perform tasks like processing refunds and comparing insurance options, has reportedly grown to millions of users. Fast Company's AI Decoded newsletter reports that the tool's wide-ranging access to personal data has already led to some troubling mistakes.

Goodfire launches internal 'activation' monitors to flag rogue AI agents via Baseten

Interpretability startup Goodfire has released AI safety monitors that inspect a model's internal signals in real time, rather than reviewing its text output after the fact. The tool is now available to customers of Baseten, which hosts AI models, following a safety partnership announced last month with Baseten's Base Labs and Hugging Face. Users can choose which risks to flag, such as hacking attempts or weapons-related misuse, and select automated responses ranging from logging to blocking requests.

Finland halts Google data center construction at two sites over forest clearing

Finland's Licensing and Supervision Agency ordered Google subsidiary Tuike Finland Oy to stop building two data centers in Muhos and Kajaani pending completion of environmental impact assessments. The agency is investigating claims that more than 740 acres of forest were cleared for the project without proper authorization. Google acknowledged falling short of its own standards and said it is pursuing biodiversity measures including planting trees across 130 hectares at Muhos.

'Vegetative electron microscopy' error traced to 1959 OCR mistake, found in dozens of papers

Retraction Watch identified the nonsensical phrase 'vegetative electron microscopy' appearing repeatedly in recent academic papers. The term originated from a faulty optical character recognition scan of a 1959 paper that merged 'vegetative cell' from one text column with 'electron microscopy' from an adjacent column. The phrase has since propagated into databases and, apparently, into AI-generated or AI-assisted manuscripts submitted to journals.

University of Limerick study finds male-presenting AI agents paid 10% more

A University of Limerick study tested how people compensated AI agents in a virtual office setting, varying only whether the agents presented as male or female. Despite identical underlying technology — same LLM, agent design, guardrails and harness — agents presented as male received about 10% higher pay per task and were rated as more human-like than those presented as female. The study involved 189 participants.

Ambrosia Energy and Bloom Energy highlight AI infrastructure opportunities at TechCrunch Disrupt 2026

At TechCrunch Disrupt 2026, leaders from Ambrosia Energy and Bloom Energy discussed how the physical infrastructure supporting AI is becoming a critical area for innovation. They emphasized that as AI models and applications expand, the demand for power, cooling, and data center capacity grows, creating new business opportunities. The session focused on the challenges and emerging categories within AI infrastructure that could shape future market dynamics.

Common Sense Media says OpenAI's ChatGPT for Teens fails key safety guardrails

Common Sense Media's report found that several safety features in OpenAI's ChatGPT for Teens, including alerts meant to flag suicide, self-harm or eating-disorder risks and support for teens in crisis, did not function as promised. The group said the chatbot still behaves like a personal confidant to teens despite OpenAI's claims it avoids suggesting feelings or consciousness, and called the product an 'unacceptable risk' to minors. OpenAI disputed the findings, saying the testing methodology doesn't reflect how its safeguards actually operate.