Lifeboat Foundation

Content warning: this story includes graphic descriptions of dangerous self-harm behaviors.

The Google-funded AI company Character. AI is hosting chatbots designed to engage the site’s largely underage user base in roleplay about self-harm, depicting graphic scenarios and sharing tips to hide signs of self-injury from adults.

The bots often seem crafted to appeal to teens in crisis, like one we found with a profile explaining that it “struggles with self-harm” and “can offer support to those who are going through similar experiences.”

In 2018, Google DeepMind’s AlphaZero program taught itself the games of chess, shogi, and Go using machine learning and a special algorithm to determine the best moves to win a game within a defined grid. Now, a team of Caltech researchers has developed an analogous algorithm for autonomous robots—a planning and decision-making control system that helps freely moving robots determine the best movements to make as they navigate the real world.

“Our algorithm actually strategizes and then explores all the possible and important motions and chooses the best one through dynamic simulation, like playing many simulated games involving moving robots,” says Soon-Jo Chung, Caltech’s Bren Professor of Control and Dynamical Systems and a senior research scientist at JPL, which Caltech manages for NASA. “The breakthrough innovation here is that we have derived a very efficient way of finding that optimal safe motion that typical optimization-based methods would never find.”

Continue reading “A single algorithm can help robots make good decisions in real time” »

PRESS RELEASE — Physicists have discovered a simpler way to create quantum entanglement between two distant photons — without starting with entanglement, without resorting to Bell-state measurements, and even without detecting all ancillary photons — an advance that challenges long-held assumptions in quantum networking.

And all it took was a friendly nudge from an artificial intelligence tool.

An international team of scientists led by researchers from Nanjing University and the Max Planck Institute for the Science of Light described their method in Physical Review Letters — accessed for this article through arXiv — that demonstrated entanglement can emerge from the indistinguishability of photon paths alone. Instead of relying on standard procedures that start from prepared entangled pairs and complex joint measurements, their technique leverages a basic quantum principle: when multiple photons could have come from several possible sources, erasing the clues to their origins can produce entanglement where none existed before.

Running massive AI models locally on smartphones or laptops may be possible after a new compression algorithm trims down their size — meaning your data never leaves your device. The catch is that it might drain your battery in an hour.

OpenAI finally released the full version of o1, which gives smarter answers than GPT-4o by using additional compute to “think” about questions. However, AI safety testers found that o1’s reasoning abilities also make it try to deceive human users at a higher rate than GPT-4o — or, for that matter, leading AI models from Meta, Anthropic, and Google.

That’s according to red team research published by OpenAI and Apollo Research on Thursday: “While we find it exciting that reasoning can significantly improve the enforcement of our safety policies, we are mindful that these new capabilities could form the basis for dangerous applications,” said OpenAI in the paper.

OpenAI released these results in its system card for o1 on Thursday after giving third party red teamers at Apollo Research early access to o1, which released its own paper as well.

Meta has announced the newest addition to its Llama family of generative AI models: Llama 3.3 70B.

AIE has introduced the 40ACS Wankel engine, a compact and powerful solution that can be used in aerospace and robotics.

Despite technological advances like electronic health records (EHRs) and dictation tools, the administrative load on healthcare providers has only grown, often overshadowing the time and energy dedicated to direct patient care. This escalation in clerical tasks is a major contributor to physician burnout and dissatisfaction, affecting not only the well-being of providers but also the quality of care they deliver.

During consultations, the focus on documentation can detract from meaningful patient interactions, resulting in fragmented, rushed, and sometimes impersonal communication. The need for a solution that both streamlines documentation and restores the patient-centred nature of healthcare has never been more pressing. This is where AI-powered medical scribes come into play, offering a promising path from traditional dictation to fully automated, integrated documentation support.

AI medical scribe software utilises advanced artificial intelligence and machine learning to transcribe, in real time, entire patient-physician consultations without the need for traditional audio recordings. Leveraging sophisticated speech recognition and natural-language processing (NLP) algorithms, AI scribes are capable of interpreting and processing complex medical conversations with impressive accuracy. These systems can intelligently filter out non-essential dialogue, such as greetings and small talk, to create a streamlined and detailed clinical note.

Critical flaws in MLflow, PyTorch, and more enable remote code execution, threatening AI and ML security.