Gemini fails to understand tone as Avnet launches AI program

Key Takeaways

Avnet and HKSTP Launch AI Hardware Launchpad for Startups

Avnet and the Hong Kong Science and Technology Parks Corporation (HKSTP) have launched a new 12-month program called the DfMA Launchpad for AI. This initiative invites startups from around the world to apply for support in developing AI hardware. The program focuses on three key areas: Edge AI, Physical AI, and high-performance computing. Participants will receive up to HKD 100,000 in funding, access to advanced labs, and one-on-one mentorship from experts at Avnet and the University of Hong Kong. The goal is to help startups turn their prototype products into market-ready solutions using Hong Kong as a gateway to the Greater Bay Area.

Avnet and HKSTP Launch AI Hardware Launchpad for Startups

Avnet and HKSTP have opened an AI Hardware Launchpad designed to help global startups build better AI products. The program targets work in Edge AI, Physical AI, and high-performance computing. Selected startups can get funding of up to HKD 100,000 and access to the EMUS Lab and Hong Kong Science Park facilities. They will also receive technical mentorship from Avnet engineers and University of Hong Kong experts. This support covers AI model optimization, hardware-software design, and system validation. The initiative aims to shorten the time it takes to move from idea to a working product.

Avnet and HKSTP Invite Global Startups to New AI Hardware Launchpad

On May 12, 2026, Avnet, HKSTP, and the EMUS Lab announced a new 12-month program for AI startups. The DfMA Launchpad for AI MMP Programme is now accepting applications from teams worldwide. The program supports local and international startups in commercializing innovations in Edge AI, Physical AI, and high-performance computing. By using Hong Kong as a strategic location, the program connects startups with global supply chains and industry partners. Leaders from Avnet and HKSTP stated that this partnership will help turn prototypes into scalable, market-ready solutions across the region.

Avnet and HKSTP Invite Global Startups to New AI Hardware Launchpad

Avnet and the Hong Kong Science and Technology Parks Corporation have launched a new program to help startups build AI hardware. The 12-month DfMA Launchpad for AI MMP Programme is open to applications from startups anywhere in the world. The initiative focuses on Edge AI, Physical AI, and high-performance computing technologies. Participants can receive up to HKD 100,000 in funding and access to advanced research facilities. They will also get one-on-one technical mentorship from experts at Avnet and the University of Hong Kong. The goal is to help startups optimize their AI models and turn their ideas into real products.

Avnet and HKSTP Invite Global Startups to New AI Hardware Launchpad

Avnet, HKSTP, and the EMUS Lab have announced a new 12-month program for AI startups on May 12, 2026. The DfMA Launchpad for AI MMP Programme is now accepting applications from teams globally. This co-incubation program helps startups accelerate the commercialization of AI innovations in Edge AI, Physical AI, and high-performance computing. Startups will gain access to global market exposure and supply chain integration through Hong Kong. The program combines Avnet's supply chain skills with EMUS Lab's research infrastructure to help turn prototypes into market-ready solutions.

AI Security Architecture Is Key to Verifiable AI Systems

Experts say that AI security architecture is the most important factor in creating trustworthy AI systems. Without a unified data model, security AI tools often have to guess instead of providing clear answers. A unified ontology at the data layer allows for transparent investigations backed by raw evidence. The structure of the platform matters more than the specific AI model chosen. Platforms built on this unified architecture can provide faster and provable investigations that security teams can trust.

AI Coders Carry Open Laptops in Public to Keep Agents Running

Software developers using AI agents are carrying their laptops open in public places to keep their AI tools running. People have been seen walking through airports, offices, and even ice rinks with their devices cracked open. This behavior happens because the AI agents need a constant internet connection to function properly. One developer noted that people often think they are strange for doing this. The practice highlights how deeply integrated AI tools have become in the daily workflow of modern coders.

Gemini Fails to Understand Tone in TV Recommendations

A recent test of Gemini, an AI-powered TV recommendation engine, showed that it struggles to understand human emotions. The reviewer found that Gemini could not pick up on emotional clues when suggesting shows. Instead of recommending content based on mood, it suggested random titles that did not match the user's preferences. The AI seemed stuck in a loop, repeating the same suggestions even after feedback was given. This experience suggests that Gemini is not yet ready to be a reliable TV recommendation tool.

Elon Musk Compares AI Future to The Terminator in Court

During a trial between Elon Musk and Sam Altman, Musk discussed the dangers of artificial intelligence with the judge. His lawyer argued that AI risks are the most important topic in the case, comparing a bad future to the movie The Terminator. The judge, Yvonne Gonzalez Rogers, tried to steer the trial away from sci-fi scenarios and focus on the actual legal issues. She noted that the trial is not about whether AI will destroy humanity. Despite the judge's efforts, the topic of AI safety remained a constant presence in the courtroom.

Cursor VP Explains How Humans and AI Agents Work Together

Tido Carriero, VP of Engineering at Cursor, discussed how to build teams that use AI agents in software development. He explained that while AI agents can handle code generation and security checks, human oversight is still essential. Humans must review product plans and make key architectural decisions that AI cannot fully understand. The company uses AI to automate tasks like triaging issues and running experiments, but humans finalize the results. This partnership allows teams to be more efficient while maintaining quality and security.

Delta CEO Threw Away AI Speech for Graduates

Delta CEO Ed Bastian asked AI to write his graduation speech for Emory University students, but he decided to discard the draft. He felt the AI-generated text lacked soul and warmth, sounding like an algorithm rather than a personal message. Instead, he wrote a new speech by hand using pencil and paper to ensure it reflected his true voice. Bastian told the graduates that authenticity and character are the most important assets they can have. He emphasized that shortcuts do not lead to lasting success and that people skills matter more than technical abilities.

Parameter Golf Shows How AI Agents Help AI Research

An open machine learning challenge called Parameter Golf involved over 1,000 participants and more than 2,000 submissions. The goal was to minimize loss on a fixed dataset while keeping the model size under 16 MB. Participants used AI coding agents to lower the cost of experimentation and speed up the competition. The challenge revealed many creative techniques, including careful tuning of optimizers and new methods for model compression. Organizers found that AI agents changed the pace of research and helped discover new talent in the machine learning community.

AI Sales Tools Expose Weak Systems Instead of Fixing Them

Many companies believe AI will fix their sales problems, but experts say it often exposes existing weaknesses instead. Leaders often rush to adopt AI tools without fixing the underlying issues in their sales processes. When data is poor or systems are disorganized, AI creates more noise and activity without generating real revenue. High-performing organizations use AI to improve decisions, not just to automate tasks. The key is to build a strong foundation first, then use AI to scale what is already working well.

AI Car Salespeople Are More Trusted Than Humans

AI is changing how people buy cars by acting as friendly virtual salespeople on dealership websites. These AI assistants, often named Sheila or Bob, are available 24/7 to answer questions and set up appointments. They are programmed to be polite and non-threatening, which makes customers feel more comfortable than talking to a human salesperson. AI also speeds up car design and manufacturing by running complex simulations in seconds instead of hours. Dealerships use these tools to improve quality control and ensure parts are ready before a customer arrives.

Sources

• Avnet and HKSTP Invite Global Startups to New AI Hardware Launchpad
• Avnet and HKSTP Launch AI Hardware Launchpad for Startups
• Avnet and HKSTP Invite Global Startups to New AI Hardware Launchpad
• Avnet and HKSTP Invite Global Startups to New AI Hardware Launchpad
• Avnet and HKSTP Invite Global Startups to New AI Hardware Launchpad
• AI Security Architecture: The Key to Verifiable AI
• AI coders walk around in public with laptops open to keep agents going
• Gemini Is Tone Deaf (And Other Adventures In Using AI For Television)
• Elon Musk Fears That AI’s Future Could Look Like ‘The Terminator’
• Cursor's VP of Engineering on Building AI Agent Teams
• Delta’s CEO let AI write a speech for Gen Z college grads—but he threw it away and started over with pencil and paper for one key reason
• What Parameter Golf taught us about AI-assisted research
• The AI Sales Problem No One Wants to Admit
• Meet the AI Car Salesperson Customers Trust More Than Humans