Tuya Smart unveiled an upgraded Hey Tuya voice assistant at its 2026 Global Developer Summit in Shenzhen on April 24. The new version integrates with Google Mail, Calendar, and Docs, letting users send emails, schedule meetings, and process documents using voice commands. It also supports Matter and Home Assistant for broader hardware compatibility. Tuya introduced new AI tools like Vibe Coding, which lets developers build apps through natural language, and new AI models including Physical AI Foundation V2.8, Wukong AI 3.0, and OmniMem V2.0 to power its AI Home, AI Robot, and AI Energy ecosystems.
In other AI news, Together AI introduced a new framework called distribution-aware speculative decoding (DAS) that speeds up reinforcement learning training for large language models. The rollout phase, where models generate responses for training, can consume up to 70 percent of total training time. DAS achieves up to a 50 percent speedup in RL rollouts by using an adaptive suffix tree drafter and length-aware scheduling, all without degrading reward quality. This could significantly cut compute costs for AI training.
Security concerns around AI tools also made headlines this week. A restricted AI security tool ended up in the hands of Discord users, and three exploits were discovered that could turn Windows Defender into an attacker tool. AI vendors are promoting their tools for finding vulnerabilities but rejecting responsibility for fixing them. Separately, a journalist tracked a Dutch Navy frigate using a 5 dollar Bluetooth tracker hidden in a postcard, leading the Dutch government to ban greeting cards with batteries.
On the education front, the University of Arkansas at Little Rock is hosting its second annual AI Hackathon and HealthTech Startup Week for students ages 16 to 26. Teams will solve healthcare challenges using AI tools, with no coding experience required. Participants compete for a share of 20,000 dollars in prizes and receive mentorship from healthcare professionals and AI engineers. Organizers expanded the program to 150 students this year, including dorm rooms for those traveling from far away.
In a military experiment, officers from the 3rd Brigade Combat Team, 101st Airborne Division tested AI tools during a rotation at the Joint Readiness Training Center at Fort Polk, Louisiana. The AI helped draft brigade operations orders in about 23 hours, faster than most staffs achieve under pressure. It also assisted with extracting tasks from orders and analyzing mission factors, but human officers made all final decisions. The experiment showed that AI works best as support for staff, not as a replacement for command judgment.
Other notable developments include the world's first museum dedicated to AI art, Dataland, opening in Los Angeles. Created by digital artist Refik Anadol, its inaugural exhibition Machine Dreams: Rainforest uses a Large Nature Model trained on millions of nature images to create digital sculptures and soundscapes. The museum uses only open-source data and images from national parks to address ethical concerns. A 17-year-old student in New Jersey was charged for creating and sharing AI-generated nude images of classmates, highlighting the need for responsible technology use. The Take it Down Act, signed into law last year, requires media platforms to remove such images. Finally, a White House AI official hired by the Commerce Department was replaced after just four days on the job, though reasons were not disclosed.
Key Takeaways
- Tuya Smart upgraded its Hey Tuya voice assistant with Google Mail, Calendar, and Docs integration, plus support for Matter and Home Assistant.
- Together AI's DAS framework speeds up RL training rollouts by up to 50%, addressing the phase that consumes up to 70% of training time.
- UA Little Rock's AI Hackathon offers $20,000 in prizes for students ages 16 to 26, with no coding experience required.
- Army officers used AI to draft brigade operations orders in about 23 hours, but humans made all final decisions.
- A New Jersey teen was charged for creating and sharing AI-generated nude images of classmates; the Take it Down Act requires platforms to remove such content.
- Dataland, the world's first AI art museum, opens in Los Angeles with an exhibition using a Large Nature Model trained on millions of images.
- Security incidents include a restricted AI tool leaked to Discord users and exploits that could turn Windows Defender into an attacker tool.
- A developer created a free, local AI pipeline that automatically summarizes Kindle highlights from the My Clippings.txt file.
- A White House AI official was replaced after only four days on the job, for undisclosed reasons.
- Tuya introduced new AI models including Physical AI Foundation V2.8, Wukong AI 3.0, and OmniMem V2.0 for its AI ecosystems.
Tuya Smart upgrades Hey Tuya voice assistant for AI home push
Tuya Smart unveiled an upgraded Hey Tuya voice assistant at the 2026 Global Developer Summit on April 24, 2026. The new version can send emails, schedule meetings, and process documents using voice commands. It also supports Matter and Home Assistant for broader hardware compatibility. Tuya introduced new AI tools like Vibe Coding for app creation and expanded into AI Home, AI Robot, and AI Energy ecosystems.
Tuya Smart launches upgraded Hey Tuya with expanded AI tools
Tuya Smart announced the upgraded Hey Tuya voice assistant at the 2026 TUYA Global Developer Summit in Shenzhen, China on April 24, 2026. The new version integrates with Google Mail, Calendar, and Docs, allowing users to send emails and schedule meetings by voice. It supports Matter and Home Assistant for a wider hardware ecosystem. Developers can use Vibe Coding to build AI and IoT applications through natural language. Tuya also introduced new AI models including Physical AI Foundation V2.8, Wukong AI 3.0, and OmniMem V2.0 to support AI Home, AI Robot, and AI Energy ecosystems.
UA Little Rock hosts AI Hackathon for students ages 16 to 26
The University of Arkansas at Little Rock is hosting its second annual AI Hackathon and HealthTech Startup Week. Students ages 16 to 26 from Arkansas and surrounding states can form teams to solve healthcare challenges using AI tools. No coding experience is required, and participants compete for a share of 20,000 dollars in prizes. The event offers mentorship from healthcare professionals and AI engineers, plus industry-recognized AI certifications. Organizers expanded the program to 150 students this year, including dorm rooms for those traveling from far away.
Army experiment uses AI to speed up planning and save cognitive time
Officers from the 3rd Brigade Combat Team, 101st Airborne Division tested AI tools during a rotation at the Joint Readiness Training Center at Fort Polk, Louisiana. They used AI to draft brigade operations orders in about 23 hours, which is faster than most staffs achieve under pressure. The AI helped with tasks like extracting tasks from orders and analyzing mission factors, but human officers still made all final decisions. The experiment showed that AI works best as support for staff, not as a replacement for command judgment.
New Jersey teen charged for creating AI nude images of classmates
A 17-year-old student at Montgomery High School in New Jersey was arrested in March for creating and sharing AI-generated nude images of fellow classmates. State police charged the teen with harassment and possession and distribution of child sexual exploitation material. The school district said the images were made outside school hours and off school grounds. Experts urge parents to teach children about responsible technology use. The Take it Down Act, signed into law last year, requires media platforms to remove such images.
World first museum of AI art to open in Los Angeles
Dataland, the world first museum dedicated to AI art, will open in Los Angeles. The museum was created by digital artist Refik Anadol and features an inaugural exhibition called Machine Dreams: Rainforest. The exhibition uses a Large Nature Model trained on millions of nature images to create digital sculptures and soundscapes. The museum addresses concerns about AI art ethics by using only open-source data and images from national parks. Critics question whether AI art can be considered true art, but the museum aims to show AI art is more than just images.
AI security tools face accountability issues and new exploits emerge
This week security news highlights several incidents. A restricted AI security tool ended up in the hands of Discord users. A journalist tracked a Dutch Navy frigate using a 5 dollar Bluetooth tracker hidden in a postcard, leading the Dutch government to ban greeting cards with batteries. Three exploits were discovered that could turn Windows Defender into an attacker tool. AI vendors are promoting their tools for finding vulnerabilities but rejecting responsibility for fixing them.
Developer builds free AI pipeline to summarize Kindle highlights
A developer created a local, zero-cost AI pipeline that automatically cleans, structures, and summarizes Kindle highlights. The project extracts highlights from the My Clippings.txt file on any Kindle device. It then uses AI to generate a summary of the book based on those highlights. The developer shared the method so others can use it to better retain information from their reading without spending hours manually summarizing.
Together AI new technique cuts RL training time by half
Together AI introduced a new framework called distribution-aware speculative decoding, or DAS, that speeds up reinforcement learning training for large language models. The rollout phase, where models generate responses for training, can consume up to 70 percent of total training time. DAS achieves up to a 50 percent speedup in RL rollouts by using an adaptive suffix tree drafter and length-aware scheduling. The technique works without degrading reward quality, making it a valuable tool for cutting compute costs in AI training.
White House replaces new AI official after only four days
An artificial intelligence researcher hired by the Commerce Department to run a key federal technology center was replaced after just four days on the job. The Washington Post reported the situation based on four people familiar with the matter. The reasons for the quick replacement were not disclosed in the article.
Sources
- Tuya's assistant now sends emails by voice as it widens AI home push
- Tuya Smart Unveils Upgraded Hey Tuya and Expanded AI Capabilities for Hardware Innovation, Advancing AI Home, AI Robot, and AI Energy Ecosystems
- Students Invited to Build, Compete, and Innovate at UA Little Rock AI Hackathon
- Artificial Staff, Human Command: An AI Integration Experiment
- NJ high school teen accused of creating AI-generated nude photos of classmates
- The World's First Museum of A.I. Art Will Open in Los Angeles as the Art World Ponders Questions of Ethics and Sustainability
- Infosec This Week: AI Accountability Slips, Windows Defender Turns Attacker, and... Someone Tracked a Warship
- I Built an AI Pipeline for Kindle Highlights
- Together AI Slashes RL Training Time
- White House pushed out new AI official after just four days on the job
Comments
Please log in to post a comment.