Character.AI has rolled out new safety features and policies for building and interacting with the AI-powered virtual personalities it hosts. The new measures aim to make the platform safer for all users, but particularly younger people. The update includes more control over how minors engage with the AI chatbot, more content moderation, and better detection of the AI discussing topics like self-harm.

Though not cited in the blog post about the update, Character AI linked to the announcement in a post on X expressing condolences to the family of a 14-year-old who spent months interacting with one of Character.AI’s chatbots before taking his own life. His family has now filed a lawsuit against Character.AI for wrongful death, citing a lack of safeguards for the AI chatbots as a contributor to his suicide​.

AI chat guardrails

‘Sometimes the hardest journeys require the most patience’, posted Jared Isaacman on X as he and his crewmates await to embark on the ‘Polaris Dawn’ mission. This is a historical mission, with Polaris Dawn completing the first-ever private spacewalk and travelling farthest to the earth since Apollo. Over half a century has passed since humans have flown this far from Earth. Isaacman will fly as Polaris Dawn’s mission commander, and will be joined by the mission’s pilot, retired United States Air Force (USAF) Lieutenant Colonel Scott “Kidd” Poteet, and the first two SpaceX employees to launch to orbit, Sarah Gillis and Anna Menon, who both work as Lead Space Operations Engineers. Gillis and Menon will serve as mission specialists.

The team at TrialX is eagerly awaiting the launch of the mission they are integral to. In the last few months, they have been working to develop the EXPAND app, in partnership with the Translational Research Institute for Space Health (TRISH) at the Baylor College of Medicine for the Polaris Dawn crew.

As part of the mission, the four-member crew will conduct a wide range of experiments to collect data on various aspects of human health in space. These studies will examine environmental factors, health and hygiene, personality development, vital signs, cognitive function, vision, motion sickness, decompression sickness, Spaceflight Associated Neuro-Ocular Syndrome (SANS), space radiation, and more.

All health data will be collected using integrated wearables and devices, along with nearly 30 surveys embedded in the EXPAND App. This comprehensive approach is designed to provide valuable insights into how the unique conditions of space affect human health at both physiological and molecular levels.

Additionally, the crew will collect biological samples to contribute to a biobank, aimed at uncovering the molecular changes that occur during spaceflight. A paper detailing the creation and purpose of this biobank, with the most extensive collection of data ever assembled for aerospace medicine and space biology. has been published in Nature, underscoring its significance for future research.

TrialX’s remote data collection platform is the backbone of the EXPAND App, providing a robust and secure system for data collection, management and analysis. The platform integrates seamlessly with various health monitoring devices and sensors, ensuring real-time data capture, implementing top-tier security protocols and is designed to handle large volumes of data, including genomic data.

TrialX was chosen by TRISH in 2021 to build the innovative centralised data repository and analytics platform for the first all-civilian Inspiration4 mission. The database was built as part of TRISH’s EXPAND (Enhancing eXploration and Analog Definition) programme, a multi-year initiative developed to help the NASA Human Research Program to reduce the risk to human health in space. The database supports a variety of data types across a multitude of individual research studies and currently houses data from subjects across six space missions – Inspiration 4, MS-20, Axiom-1, Axiom-2, Axiom-3 and Polaris Dawn. It equips space researchers to reuse and integrate research data across different research studies and unlock innovative actionable insights.

TrialX has always been on the forefront of cutting edge clinical research solutions and space health informatics. From launching the first clinical trials app on Google Health in 2008, to powering online patient recruitment and research data collection for space missions, co-founders Sharib Khan and Chintan Patel are on a mission to facilitate One Billion Health Research Contributions.

“Our mission at TrialX has always been to bridge the gap between research and participants, whether on Earth or in space. Partnering with TRISH to support commercial space missions allows us to extend our commitment to advancing clinical research, using technology to empower both researchers and participants in groundbreaking ways,” said CEO and cofounder Sharib Khan.

Khan and Patel first crossed paths at Columbia University, where their strong rapport led to the founding of TrialX Inc. and a partnership that has spanned over 16 years. Headquartered in New York, with presence in India, the Philippines, and Romania, the company is committed to accelerating clinical research and bridging the gap between research and patients—both on Earth and beyond.

Khan brings a deep understanding of clinical research and digital health to the table. With a background in medical sciences and biomedical informatics, he has been a driving force behind TrialX’s mission to democratise clinical research by using technology to engage and empower patients. Patel, with his background in computer science and a passion for healthcare innovation, plays a crucial role in driving technology solutions that make clinical research more accessible and efficient. Their shared vision and commitment to innovation and patient-centricity continues to push the boundaries of what’s possible in clinical trials and space health.

More recently, TRISH, in collaboration with consortium partners Massachusetts Institute of Technology and California Institute of Technology, has partnered with TrialX to develop a software platform for a portable device that acts as a “to-go” version of the EXPAND master repository. This portable device is designed to extend the repository’s capabilities, allowing spaceflight participants to access their previously collected health information and store new data offline during their mission.

Upon returning to Earth, the device will automatically synchronise all collected data with the EXPAND Master Repository. If a participant continues on another segment of their space journey, the portable device will seamlessly transfer their existing health data from one mission segment to the next. By tackling the challenges of data transfer and interoperability in space, this portable solution aims to enhance the efficiency and quality of healthcare for astronauts.

The Polaris Dawn mission was initially slated to launch on August 27, but was delayed to August 28 due to technical issues and then postponed again, due to splashdown weather concerns for the end of the mission. 

Disney+ is easily the best streaming service for watching kids movies. Family audiences have all they could ask for between all-time classics and modern hits. Subscribers have countless hours’ worth of child-friendly content at their fingertips, and this guide is updated monthly to spotlight some of the best kids movies on Disney+.

If you haven’t signed up already, the Disney Bundle is a terrific package for getting the most value from your streaming budget. This gives audiences access to Disney+, Hulu’s ad-based plan, and ESPN+’s sports coverage in a streamlined bundle. With this streaming package, fans will have more than enough Star Wars, Marvel Studios, and sports content.

We’ve also rounded up the best kids movies on Hulu, the best kids movies on Netflix, and the best kids movies on Amazon Prime Video if you don’t find what you’re looking for on Disney+. For older audiences, check out the best new movies to stream this week and the best movies on Disney+.

T-Mobile is going to help self-driving cars run on a loop with its 5G network. The 5G network will be private, and ensure consistent and reliable connection for moving passenger vehicles.

T-Mobile to offer reliable and private 5G network for self-driving cars

5G networks were supposed to bring a revolution in wireless internet. Carriers promised ultra-high-speed connections with low latency. Although 5G has yet to live up to the lofty expectations, it is going to help autonomous passenger vehicles.

T-Mobile is trying to prove how 5G networks can be helpful in multiple scenarios. The company recently indicated it is offering a new 5G-based service to help emergency workers and citizens stay connected. Now, the carrier announced in a press release that it’s working with Miller Electric Company.

Specifically speaking, T-Mobile is going to support autonomous shuttles with 5G network connectivity. According to the telecom service provider, “T-Mobile will deliver highly secure, reliable two-way communications between the shuttles and the command center.

T-Mobile will also share real-time telemetry data from onboard sensors as well as audio and video feeds to connect employees with passengers. Simply put, T-Mobile is offering a dedicated, private 5G network to a fleet of self-driving cars.

How will a private 5G network work with passenger vehicles?

Miller Electric Company is reportedly bringing fully autonomous shuttles to a three-mile strip in Jacksonville. Self-driving shuttles will connect the EverBank Stadium in Jacksonville to the new waterfront project. The fleet should start operations next year.

The bus-shaped shuttles will rely on T-Mobile’s Advanced Network Solutions (ANS) and connect to a private 5G network. One of the parties involved in the project has stated that existing standards, such as Wi-Fi, aren’t as reliable or scalable as 5G.

A dedicated and private 5G network should remain invisible and inaccessible to the general population. In other words, T-Mobile could be reserving a small slice of the 5G spectrum for self-driving cars. This would ensure congestion-free connectivity for high-bandwidth applications.

In its press release, T-Mobile has urged cities and government departments to explore the possibilities 5G networks can unlock. The company has implied that 5G isn’t limited to smartphones and home broadband. This strongly suggests T-Mobile is trying to expand in other markets using its 5G networks.

The United Nations’ Environmental Program has released a new with yet more dire news about our odds of avoiding climate disaster caused by greenhouse gas emissions. According to this assessment, the current trajectory of international commitments will see the planet’s temperature increasing 2.6 degrees Celsius or more over the course of this century. That amount of temperature change would lead to more catastrophic and life-threatening weather events.

UN members are due to submit their latest Nationally Determined Contributions ahead of the COP30 conference in Brazil next year. The NDCs lay out each country’s plan for reduced greenhouse gas emissions. One part of the NDCs are to reach the goal set by the Paris Agreement to limit global temperature increases to 1.5 degrees Celsius, and one part targets keeping global temperature increases to within a less ideal 2 degrees Celsius. While the report says it is technically possible to reach the Paris Agreement goal, much larger actions will be required to cut emissions by the necessary amount.

“Increased deployment of solar photovoltaic technologies and wind energy could deliver 27 percent of the total emission reduction potential in 2030 and 38 percent in 2035,” the report gives as an example of what’s still needed. “Action on forests could deliver around 20 percent of the potential in both years.”

“Every fraction of a degree avoided counts in terms of lives saved, economies protected, damages avoided, biodiversity conserved and the ability to rapidly bring down any temperature overshoot,” UN Environment Program Executive Director Inger Andersen wrote in the report’s forward.

International collaboration, government commitments and financial contributions will also be essential for getting back on track to either the 2-degree or 1.5-degree goals. “G20 nations, particularly the largest-emitting members, would need to do the heavy lifting,” the report reads.

If all of this sounds familiar, that’s probably because the UN has issued the same stark warnings in each of its annual reports on emissions for now. And other reports have echoed their calls, such as damning earlier this year that just 57 companies are responsible for 80 percent of carbon dioxide emissions worldwide.

Cohere today released two new open-weight models in its Aya project to close the language gap in foundation models. 

Aya Expanse 8B and 35B, now available on Hugging Face, expands performance advancements in 23 languages. Cohere said in a blog post the 8B parameter model “makes breakthroughs more accessible to researchers worldwide,” while the 32B parameter model provides state-of-the-art multilingual capabilities. 

The Aya project seeks to expand access to foundation models in more global languages than English. Cohere for AI, the company’s research arm, launched the Aya initiative last year. In February, it released the Aya 101 large language model (LLM), a 13-billion-parameter model covering 101 languages. Cohere for AI also released the Aya dataset to help expand access to other languages for model training. 

Aya Expanse uses much of the same recipe used to build Aya 101. 

“The improvements in Aya Expanse are the result of a sustained focus on expanding how AI serves languages around the world by rethinking the core building blocks of machine learning breakthroughs,” Cohere said. “Our research agenda for the last few years has included a dedicated focus on bridging the language gap, with several breakthroughs that were critical to the current recipe: data arbitrage, preference training for general performance and safety, and finally model merging.”

Aya performs well

Cohere said the two Aya Expanse models consistently outperformed similar-sized AI models from Google, Mistral and Meta. 

Aya Expanse 32B did better in benchmark multilingual tests than Gemma 2 27B, Mistral 8x22B and even the much larger Llama 3.1 70B. The smaller 8B also performed better than Gemma 2 9B, Llama 3.1 8B and Ministral 8B. 

Cohere developed the Aya models using a data sampling method called data arbitrage as a means to avoid the generation of gibberish that happens when models rely on synthetic data. Many models use synthetic data created from a “teacher” model for training purposes. However, due to the difficulty in finding good teacher models for other languages, especially for low-resource languages. 

It also focused on guiding the models toward “global preferences” and accounting for different cultural and linguistic perspectives. Cohere said it figured out a way to improve performance and safety even while guiding the models’ preferences. 

“We think of it as the ‘final sparkle’ in training an AI model,” the company said. “However, preference training and safety measures often overfit to harms prevalent in Western-centric datasets. Problematically, these safety protocols frequently fail to extend to multilingual settings.  Our work is one of the first that extends preference training to a massively multilingual setting, accounting for different cultural and linguistic perspectives.”

Models in different languages

The Aya initiative focuses on ensuring research around LLMs that perform well in languages other than English. 

Many LLMs eventually become available in other languages, especially for widely spoken languages, but there is difficulty in finding data to train models with the different languages. English, after all, tends to be the official language of governments, finance, internet conversations and business, so it’s far easier to find data in English. 

It can also be difficult to accurately benchmark the performance of models in different languages because of the quality of translations. 

Other developers have released their own language datasets to further research into non-English LLMs. OpenAI, for example, made its Multilingual Massive Multitask Language Understanding Dataset on Hugging Face last month. The dataset aims to help better test LLM performance across 14 languages, including Arabic, German, Swahili and Bengali. 

Cohere has been busy these last few weeks. This week, the company added image search capabilities to Embed 3, its enterprise embedding product used in retrieval augmented generation (RAG) systems. It also enhanced fine-tuning for its Command R 08-2024 model this month.