A supercomputer at Beth Israel Deaconess Medical Center in Boston uses machine learning on data from over 250,000 patients to predict medical problems including death. Project lead Dr. Steve Horng says the system can forecast a patient's probability of dying within 30 days with 96 percent confidence, while emphasizing it is meant to augment, not replace, clinicians.
Meta aims to populate Facebook with AI-generated characters that have profiles and post content, according to the Financial Times. A product executive said the company sees these AIs existing like accounts, while a former Meta creator-team lead warned of safety and trust risks.
OpenAI has acquired Chat.com, a domain previously used for adult video chat, and now redirects visitors to ChatGPT. Former owner Dharmesh Shah, who bought it for $15.5 million in 2023, said he sold it for more, though the exact price is undisclosed.
Tesla's Autopilot 8.0 update will let radar initiate braking without camera confirmation and process six times as many radar objects, a change CEO Elon Musk says could have prevented the fatal 2016 Model S crash that killed Joshua Brown.
Property managers are deploying AI chatbots like EliseAI to handle tenant inquiries and complaints, according to The New York Times. The trend raises questions about accountability when chatbots give wrong information, as illustrated by a case where Air Canada was ordered to pay damages after its AI misstated a bereavement policy.
Stability AI released Stable Video Diffusion, an image-to-video generator that has produced bizarre animated versions of memes like 'distracted boyfriend.' The launch follows executive departures and a former VP's public resignation over fair-use disagreements.
Murray Campbell, an IBM researcher who helped build Deep Blue, discussed the legacy of the 1997 chess match in which the machine defeated world No. 1 Garry Kasparov. Campbell traces the project back to Deep Thought, the first program to beat a grandmaster in a professional tournament, and explains why chess carried more weight than earlier computer wins in backgammon and checkers.
Security researchers at Wiz discovered that DeepSeek left a database completely unencrypted and publicly accessible, exposing chat histories, API secrets, and operational details. DeepSeek locked the database within an hour of being alerted, but Wiz said contacting the company was difficult and that the exposure could have allowed attackers to access internal data with minimal effort.
Emily Kennedy, a Carnegie Mellon graduate, developed Traffic Jam, an AI-powered tool that scans online sex advertisements to identify human trafficking victims. Since founding Marinus Analytics in 2014, her software has led to the rescue of hundreds of victims in the U.S. and Canada and is expanding internationally.
An Axios survey found that none of the 2020 presidential campaigns contacted had a plan to counter deepfakes, computer-generated media depicting real people saying or doing things that never occurred. The findings came after an altered video of Nancy Pelosi spread on Facebook and as governments treated the technology as a national security threat.
UK artificial intelligence expert Catriona Campbell argues in her new book that virtual children in the metaverse could offer a cheaper, environmentally friendly answer to overpopulation. She predicts such AI offspring will be indistinguishable from real children within 50 years and available for a small monthly fee.
Despite OpenAI's usage policies prohibiting sexually explicit content, users are crafting detailed prompts to bypass ChatGPT's guardrails. The loophole highlights broader challenges in moderating AI-generated content, as rival platforms like DeepAI and Poe also face similar issues.
Robot ethics professors Alan Winfield and Marina Jirotka propose fitting robots with an 'ethical black box' to record decisions and sensory data, modeled on aviation flight recorders. The call comes amid broader concerns about AI oversight and industry efforts to establish ethical standards.
President Donald Trump said Monday that DeepSeek's release should be a 'wake-up call' for US industries, calling its low cost a positive development. The Chinese startup's open-source R1 model, built for under $6 million, has rattled US tech stocks and raised questions about spending on AI infrastructure.
Deep learning pioneer Jürgen Schmidhuber predicted at WIRED2016 that by 2050 self-replicating robot factories will operate in the asteroid belt, with AI eventually colonizing the galaxy. He cited current deep learning capabilities in speech, image recognition, and medical diagnosis as evidence.
Google has acquired Eyefluence, an eye-tracking interface startup, according to an announcement on Eyefluence's website. The deal comes as Google prepares to launch its Daydream VR platform and Daydream View headset in November, potentially enabling VR control through eye movements.
Astronomer CEO Andy Byron resigned after a Coldplay concert Jumbotron showed him with Chief People Officer Kristin Cabot. The board accepted his resignation, named cofounder Pete DeJoy interim CEO, and Byron negotiated an exit package, according to Axios.
Nikkei Asia found 17 English-language papers on arXiv from 14 institutions in eight countries containing hidden prompts designed to trick AI review tools into giving positive assessments. Authors' responses ranged from remorse to defense, highlighting unresolved questions about AI's role in peer review.
The University of Chicago Law School will ban phones, tablets, and laptops in first-year classes as part of an 'AI-resilient' curriculum. Exams will be in-class without internet access, and students will discuss research papers orally with professors, though the school still encourages some AI use for studying.