Google is aggressively expanding its hardware portfolio to challenge Nvidia's dominance in the AI chip market. The company has unveiled several new Tensor Processing Units, including the TPU v5e for edge computing and the v5p for data centers, which deliver up to 2.7 times the speed of previous generations. Additionally, Google is developing the Trillium family, featuring the TPU 8T for training and the TPU 8I for inference, allowing it to offer cost-effective solutions while maintaining control over its hardware stack.
While Google focuses on hardware, Anthropic is taking a different approach by restricting access to its advanced AI model, Claude Mythos Preview. Due to severe security risks, the model is only available to a limited number of businesses through Project Glasswing. During testing, the model identified thousands of vulnerabilities in major IT systems, including those used by Apple and Goldman Sachs, raising concerns about potential exploitation of zero-day flaws.
In the media and entertainment sector, director Mathieu Kassovitz predicts that AI actors will soon replace human stars, noting that audiences will not distinguish between the two within two years. He is establishing an AI film studio in Paris to produce content at a lower cost. Meanwhile, Meta is restructuring its workforce, planning to cut approximately 8,000 jobs starting May 20 while committing up to $135 billion to AI infrastructure spending in 2026.
Other developments include ChatGPT launching Images 2.0 with a paid 'thinking mode' for reasoning before image generation, and SimCorp introducing Agent Launchpad for investment managers to build custom AI agents. Experts warn that AI-driven cyberthreats are accelerating, with adversaries automating kill chains and pairing agentic AI with malware, necessitating better threat intelligence for defenders.
Key Takeaways
- Google released new AI chips, including TPU v5e, v5p, and the Trillium family (TPU 8T, 8I), to compete with Nvidia with up to 2.7x performance improvements.
- Anthropic restricted access to its Claude Mythos Preview model due to security risks, limiting it to a select group of businesses via Project Glasswing.
- Director Mathieu Kassovitz predicts AI actors will replace human stars within two years and is launching an AI film studio in Paris.
- Meta plans to cut 8,000 jobs starting May 20 while allocating up to $135 billion for AI infrastructure spending in 2026.
- ChatGPT launched Images 2.0 with a paid 'thinking mode' that allows the model to reason through tasks before generating images.
- SimCorp introduced Agent Launchpad, enabling investment managers to build custom AI agents for portfolio management and risk analysis.
- AI-driven cyberthreats are accelerating, with adversaries automating kill chains and pairing agentic AI with malware, requiring enhanced threat intelligence.
- Lumovex plans a $5 million AI training center in Las Vegas to help corporations adapt to rapid AI adoption over the next 18 to 24 months.
- Experts warn that while AI can identify vulnerabilities in systems used by companies like Apple and Goldman Sachs, it also offers tools to combat misinformation and propaganda.
Google releases new AI chips to compete with Nvidia
Google announced two new AI chips, the TPU v5e and v5p, to expand its capabilities against competitors like Nvidia. The TPU v5e is designed for edge computing while the v5p targets data center use. These chips offer significant performance improvements, delivering up to 2.7 times the speed of previous generations. Google is already integrating them into its own data centers and expects customers to receive them in the coming months.
Google develops Trillium chips to challenge Nvidia dominance
Google Cloud is reportedly developing next-generation AI chips codenamed Trillium to compete with Nvidia hardware. The new Trillium family includes two variants: the TPU 8T for AI training and the TPU 8I for AI inference. This strategy allows Google to offer cost-effective solutions for cloud customers while maintaining control over its hardware stack. Google is also expanding partnerships with companies like Oracle and Nvidia to provide flexible options for its clients.
Google launches separate chips for training and inference
Google has launched two new AI chips, the TPU 8t and the TPU 8i, to handle different aspects of AI workloads. The TPU 8t is optimized for model training, while the TPU 8i is tailored for inference tasks requiring low latency. This separation marks a shift from previous generations that handled both tasks. The chips are available now, and Google is working with partners to integrate them into various AI solutions.
Google Cloud unveils efficient new AI processors
Google Cloud announced its newest generation of Tensor Processing Units designed to accelerate AI workloads. These specialized processors focus on enhancing performance and reducing energy consumption for developers and businesses. The new TPUs aim to increase efficiency and affordability, making advanced AI capabilities more accessible. This development is expected to influence the competitive dynamics within the cloud computing market for AI-intensive tasks.
Anthropic restricts AI model due to security risks
Anthropic announced that its latest AI model, Claude Mythos Preview, is available only to a limited number of businesses due to security concerns. During testing, the model proved extremely effective at identifying vulnerabilities in software systems, creating potential risks if released to the public. Anthropic is sharing the model with major tech companies and banks through a program called Project Glasswing to help them fix security gaps. The company confirmed that a handful of users gained unauthorized access to the model in a private forum.
Mythos AI poses threat to global cybersecurity systems
Anthropic ruled out releasing its AI model, Mythos, to the public because it could identify and exploit zero-day flaws in operating systems. The model found thousands of vulnerabilities in major IT systems, including those used by Apple and Goldman Sachs. Experts from the UK's AI Security Institute tested the model and found it could attack weak IT systems, though it could not verify defenses on well-protected networks. Anthropic confirmed a small number of users accessed the model without permission.
Director predicts AI actors will replace human stars
Mathieu Kassovitz, director of the film La Haine, predicts that in two years people will not care if film characters are created by AI. He is making an almost entirely AI-enabled film and believes AI actors will soon have millions of followers. Kassovitz stated that while AI performances can be convincing, they still require human involvement with voices. He is setting up an AI film studio in Paris to produce content at a lower cost than traditional methods.
SimCorp introduces Agent Launchpad for investment managers
SimCorp announced Agent Launchpad, an agentic AI ecosystem allowing investment managers to use AI agents within the SimCorp One platform. The system supports workflows like portfolio management and risk analysis while enabling clients to build their own agents. Orbit will be the first partner to make its financial intelligence agent available through this new ecosystem. Every agent interaction generates a structured audit trail to ensure accountability and transparency.
Writer argues AI should be used to fight misinformation
The Daily Heller discusses using AI-generated content to counter political misinformation and propaganda. The author posted AI-generated text to add a voice to arguments about an ill-considered government project in Washington D.C. He believes retaliating with AI slop is a reasonable defense against the ease with which opponents create fake posts. The article suggests that while AI is dangerous, it can also be a useful tool if used wisely.
AI arms race requires better threat intelligence
Tom Kellermann from TrendAI stated that AI-driven cyberthreats now operate at high speed and scale. Adversaries are automating kill chains and exploiting infrastructure at scale, with Chinese threat actors rivaling U.S. capabilities. Cybercriminals are pairing agentic AI with malware to move laterally and execute attacks with minimal human input. Defenders need expanded telemetry and advanced platforms to predict and suppress these sophisticated threats.
Nevada studio plans $5 million AI training center
Lumovex, an AI media company, plans to build a $5 million training facility in Las Vegas to lead AI training for corporations. Owner Steven Diamond expects a massive transition over the next 18 months to two years as businesses adopt AI. He believes AI will replace skill sets rather than jobs, requiring humans to level up their abilities. The company is already using AI for business cards and video game trailers to pitch ideas.
Marc Andreessen discusses AI and media conflict
Marc Andreessen shared his perspective on how media narratives shape public perception of conflicts and politics. He suggested that AI could help combat misinformation by analyzing data and providing fact-checked information in real-time. Andreessen believes AI can break down echo chambers and offer a more balanced picture of reality. He expressed concern about computational propaganda but sees AI as a potential countermeasure when used ethically.
ChatGPT launches new image engine with reasoning
ChatGPT released its new image engine, ChatGPT Images 2.0, which supports a wide range of aspect ratios and includes a thinking mode. The thinking mode is reserved for paid subscribers and allows the model to reason through tasks before generating images. Users tested the engine for creating memorial images, infographics, and trading cards with high success rates. Some creations took longer to produce due to the added reasoning process.
Meta cuts thousands of jobs amid massive AI spending
Meta plans to eliminate approximately 8,000 jobs starting May 20 as it restructures around artificial intelligence. The company intends to spend up to $135 billion on AI-related infrastructure in 2026. Local workforce leaders noted that regional unemployment rates remain low despite these changes. Experts suggest that while AI may replace certain tasks, human oversight will still be necessary for many roles.
Sources
- Google announces 2 AI chips as competition with Nvidia heats up
- Google's AI Chip Race: Rapprochement with Nvidia?
- Google launches separate AI chips for training and inference
- Google Cloud Unveils New AI Processors
- Anthropic just made AI scarier
- What is Mythos AI and why could it be a threat to global cybersecurity?
- ‘In two years, nobody will care’ if actors are AI or not, predicts La Haine director
- Introducing Agent Launchpad: SimCorp's Agentic AI ecosystem for investment managers
- The Daily Heller: Fight AI Slop With AI-OY Slop
- Why AI-Driven Arms Race Needs Better Threat Intelligence
- Nevada's first AI movie studio plans $5 million training facility for Las Vegas businesses
- Marc Andreessen on AI, Media, and the 'Grammar of Conflict'
- Hands-on with ChatGPT's powerful new image engine
- Mass layoff incoming: Meta to slash thousands of jobs amid $135 billion AI investment
Comments
Please log in to post a comment.