--- title: "🎙️ AI Daily Podcast — Model Wars, The AI Solo Show, and the US-China Rivalry" description: "Hey everyone, welcome to \"Mark's Tech Insights\"! I'm Mark. Wow, the AI news this week has been absolutely explosive; it feels like the entire industry's direction has shifted again. Today, we're going to dive into a few super exciting topics: Is Meta giving up on open source? Google and China's DeepSeek are in the midst of an open-source model war; meanwhile, the wave of job losses and wealth concentration caused by AI is becoming increasingly apparent. Ready? Let's get started" canonical_url: "https://blog.markkulab.net/en/tech-news/ai-daily-podcast-2026-04-18" author: "Mark Ku" author_url: "https://blog.markkulab.net/en/author/mark-ku" site: "Mark Ku's Tech Notes" date_published: "2026-04-18 10:00:00 +0800" category: "What's New in Tech" tags: ["ai-daily", "podcast", "tech-news", "AI模型", "Meta", "MuseSpark", "Google", "Gemma4", "DeepSeek", "開源模型", "閉源模型"] language: "en" license: "CC BY 4.0" license_url: "https://creativecommons.org/licenses/by/4.0/" attribution: "when reusing or quoting, credit the author and link back to the original" --- # 🎙️ AI Daily Podcast — Model Wars, The AI Solo Show, and the US-China Rivalry ## Opening Remarks Hey everyone, welcome to "Mark's Tech Insights"! I'm Mark. Wow, this week's AI news has been absolutely explosive. It feels like the entire industry's direction has shifted again. Today, we're going to dive into some incredibly exciting topics: Is Meta giving up on open source? Google and China's DeepSeek are in an open-source model war; and at the same time, the wave of AI-driven job losses and wealth concentration is becoming more and more apparent. Ready? Let's get started! ## Today's Top Stories ### 1. Meta Abandons Open Source? Debut of its First Closed-Source Model, Muse Spark - **Source**: TechCrunch ([https://techcrunch.com/2026/04/08/meta-debuts-the-muse-spark-model-in-a-ground-up-overhaul-of-its-ai/](https://techcrunch.com/2026/04/08/meta-debuts-the-muse-spark-model-in-a-ground-up-overhaul-of-its-ai/)) - **Summary**: A major strategy shift from Meta! They've released Spark, the first product in their new closed-source model series, Muse. This signals that Meta is no longer focusing solely on the open-source path with Llama but is starting to develop its own proprietary AI technology, preparing to compete directly with OpenAI and Google. - **Taiwanese Perspective**: For many Taiwanese startups and developers who rely on the Llama ecosystem, this is a warning sign. If Meta pivots its most powerful models to be closed-source in the future, everyone may need to re-evaluate their tech stack. - **Discussion Points**: - Why is Meta shifting to a closed-source strategy now? Is the open-source business model not working, or is it to catch up with OpenAI? - What impact will this have on the entire AI open-source community? - How will Meta balance the Llama open-source community and the Muse closed-source business model in the future? - **Script Suggestion**: Man, this news is a real bombshell. Do you all remember? For the past few years, Meta has been playing the role of the "benevolent patron of open-source AI," building a huge following with its Llama series and giving many smaller companies and academic institutions a chance to play with top-tier models. And now, they've suddenly done a complete 180, launching the closed-source Muse Spark. It feels like your favorite all-you-can-eat buffet suddenly announcing it's switching to à la carte, and the most expensive A5 Wagyu is no longer available for takeout. I think there are two possible reasons behind this: First, they found that while open source earned them a great reputation, it didn't earn them money. Second, they were spurred on by the commercial success of OpenAI and Anthropic and felt that if they didn't create their own "secret sauce," they'd fall behind in the top-tier competition. For developers in Taiwan, it's really time to start thinking about "not putting all your eggs in one basket." ### 2. Google Gemma 4 Goes Fully Open Source, Challenging Closed-Source Hegemony with Full Multimedia Support - **Source**: Google Blog ([https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/](https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/)) - **Summary**: Just as Meta is pivoting to closed-source, Google is doing the exact opposite by launching the incredibly powerful open-source model series, Gemma 4. It not only supports text and images but also audio and video, with a context window of up to 256K. It's clearly aimed at challenging the closed-source models from OpenAI and Anthropic. - **Taiwanese Perspective**: The arrival of Gemma 4 is a huge boon for Taiwan's content creation, media, and education industries. A powerful open-source model that can handle audio and video can significantly lower the barrier to developing multimedia AI applications. - **Discussion Points**: - Why did Google choose to double down on open source just as Meta is tightening up? Is this a competitive strategy? - What killer applications might emerge from multimodal open-source models that support audio and video? - How attractive is Gemma 4's Apache 2.0 license for commercial applications? - **Script Suggestion**: See, this is where it gets interesting! Meta zigs, and Google zags. Gemma 4 is a really generous offering this time, open-sourcing its multimodal capabilities, especially audio and video processing. It's like LEGO not only giving you an infinite supply of bricks but also throwing in motors and sensors so you can build actual moving robots. For Taiwan, we have many talented content creators and video platforms. Previously, the barrier to entry for AI video generation or analysis was extremely high. Now, with Gemma 4, it's like being handed a legendary weapon. You can imagine startups using it to auto-generate short videos, intelligently edit podcasts, or even create real-time interactions for virtual idols. Google is playing this move brilliantly. It's telling developers worldwide: "Hey, don't just look at Meta and OpenAI. I'm the one with the real arsenal!" ### 3. China's DeepSeek R2 Appears Out of Nowhere, Delivering a Double Shock of Price and Performance - **Source**: Decode the Future ([https://decodethefuture.org/en/deepseek-r2-explained/](https://decodethefuture.org/en/deepseek-r2-explained/)) - **Summary**: Chinese AI company DeepSeek has released a 32B open-source model called R2 with incredible performance, achieving a stunning 92.7% on math reasoning tests. Even more terrifying, it's small enough to run on a single consumer-grade graphics card (RTX 4090), and its API price is 70% cheaper than top Western models. - **Taiwanese Perspective**: For Taiwanese AI developers and businesses, this is both a "threat and an opportunity." The threat is the astonishing speed of China's technological catch-up; the opportunity is that we now have another powerful, highly cost-effective model to choose from. - **Discussion Points**: - How did DeepSeek R2 achieve such high reasoning capabilities with a smaller parameter size? - How much of an impact will this price disruption have on the existing AI API market (like OpenAI, Google)? - How should the Western world respond to this challenge of high-performance, low-cost AI models from China? - **Script Suggestion**: If Gemma 4 is like airborne support from the regular army, then DeepSeek R2 is like a ninja squad that appears out of nowhere—fast and deadly. A 32B model that can run on a consumer GPU and has frighteningly strong math skills... this isn't just "overtaking on a curve," it's like driving an F1 car on a go-kart track. The most crucial part is the price. What does 70% cheaper even mean? It's like getting the performance of a specialty pour-over coffee for the price of a Starbucks latte. This is incredibly tempting for any company that needs to use AI APIs at scale. This also teaches us a lesson: the AI arms race is no longer just a game between American giants. Chinese players are entering the fray with completely different cost structures and market strategies. The battle is about to get much more interesting. ### 4. Snap Lays Off 16%, AI Has Taken Over 65% of Coding Work - **Source**: TechCrunch ([https://techcrunch.com/2026/04/15/snap-is-cutting-1000-jobs-16-of-its-workforce/](https://techcrunch.com/2026/04/15/snap-is-cutting-1000-jobs-16-of-its-workforce/)) - **Summary**: Snap has announced it's laying off 1,000 people, or 16% of its total workforce, citing the rapid advancement of AI. Reportedly, AI is already generating 65% of the company's new code and handling over 1 million customer service issues per month. This restructuring is expected to save $500 million in annualized costs. - **Taiwanese Perspective**: This is a wake-up call for Taiwan's tech industry, especially for software engineers. The belief that coding is a stable job is being challenged. As AI-assisted programming tools become more powerful, an engineer's value must shift from "writing code" to "system design" and "solving complex problems." - **Discussion Points**: - Is AI replacing programming jobs faster than we expected? - What new skills will future software engineers need to avoid being replaced by AI? - After customer service and translation, will programmers be the next white-collar profession to be massively impacted by AI? - **Script Suggestion**: This news sent a chill down my spine. We used to joke about AI coming for our jobs, but Snap just used layoff numbers to tell us, "This is not a joke." 65% of new code being generated by AI—that number is insane! This means a feature that might have required 10 engineers in the past can now be handled by 3-4 engineers with AI assistance. This doesn't mean engineers will be out of a job, but the requirements for being an engineer have completely changed. In the future, "code monkeys" who just bury their heads and write code will be in a very precarious position. But engineers who can command AI, define problems, and design architecture—the "AI commanders"—will become incredibly sought-after. This is a challenge for both Taiwan's education system and working engineers. We need to quickly upgrade our skill sets and learn to collaborate with AI, not compete against it. ### 5. The M-Shaped AI Economy: 75% of Gains Captured by 20% of Companies - **Source**: PwC ([https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-performance-study.html](https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-performance-study.html)) - **Summary**: A new research report from PwC reveals a harsh reality: the economic benefits of AI are highly concentrated, with 75% of the profits being captured by the top 20% of companies. These leaders aren't just using AI to save money; they're using it to drive entirely new growth models, causing the gap between the "AI rich" and the "AI poor" to widen. - **Taiwanese Perspective**: Taiwan is dominated by small and medium-sized enterprises (SMEs). This report reminds us that if we only dip our toes into AI, we are likely to be pushed out of the market in the future. Business owners must adopt an "AI-first" mindset and embrace AI at the core of their strategy. - **Discussion Points**: - Why are the economic benefits of AI so concentrated? Is it the technical barrier, data moats, or a talent issue? - How can SMEs cope with this "winner-take-all" situation? - What role can the government or industry associations play to help more companies cross the AI divide? - **Script Suggestion**: This PwC report is basically the white paper on the "M-shaped society" of the AI era. We often hear about the massive economic value AI will create, but now we're finding out that most of those benefits are going into the pockets of a few giants. It's like a grand banquet where 80% of the people can only eat plain rice with pickles on the side while watching the other 20% enjoy a full imperial feast. The report says that the leaders are thinking about how to use AI to make more money, while the laggards are just thinking about using AI to save a little money. This gap in mindset is the key. For the large number of traditional industries and SMEs in Taiwan, this is an existential issue. If a boss still thinks AI is just "a tool to lay off a few customer service reps," then that company is in serious danger. The real winners are the companies that use AI as an engine to develop new products and create new markets. ### 6. AI Has Its Limits? Top Human Scientists Outperform AI on Complex Tasks - **Source**: Nature ([https://www.nature.com/articles/d41586-026-01199-z](https://www.nature.com/articles/d41586-026-01199-z)) - **Summary**: A study in the top journal *Nature* is pouring some cold water on the hype. The research found that when it comes to complex, multi-step scientific research tasks, human scientists still perform far better than the most powerful current AI agents. This study raises serious questions about the reliability of the currently trending "AI scientists" or autonomous agents. - **Taiwanese Perspective**: This gives Taiwan's high-tech R&D sectors (like semiconductors and biotech) some breathing room and also points the way forward: AI is a powerful "co-pilot," but the "main driver" who truly holds the steering wheel and can handle unknown challenges is still the human expert. - **Discussion Points**: - In which areas do current AI agents still fall short of human experts? Is it creativity, common-sense reasoning, or the ability to handle unexpected situations? - How should we view the role of AI in scientific research? As a replacement or an augmentation tool? - Will this research cool down the overheated expectations for AGI (Artificial General Intelligence)? - **Script Suggestion**: After hearing a ton of news about how amazing and powerful AI is, this *Nature* study is like a glass of ice water, helping us cool down a bit. It tells us that AI is great at solving problems with "standard answers" or "clear paths," but as soon as it encounters truly complex scientific puzzles full of unknowns and surprises, the AI starts to "get lost." It's like how an AI can be great at following a recipe, maybe even better than most people, but it can't invent a completely new, world-changing dish from scratch. This is actually good news for us. It emphasizes the importance of uniquely human abilities like creativity, intuition, and cross-disciplinary integration. So, to all the scientists and engineers out there, you don't have to worry about losing your jobs just yet. Your "brain" is still the most advanced processor in the known universe! ### 7. US-China Tech War Heats Up: Three Major US Firms Join Forces to Block Chinese "Model Stealing" - **Source**: Bloomberg ([https://www.bloomberg.com/news/articles/2026-04-06/openai-anthropic-google-unite-to-combat-model-copying-in-china](https://www.bloomberg.com/news/articles/2026-04-06/openai-anthropic-google-unite-to-combat-model-copying-in-china)) - **Summary**: OpenAI, Anthropic, and Google—once rivals—have now joined forces! Through the "Frontier Model Forum," they are sharing threat intelligence to jointly combat "model distillation attacks" from China. This type of attack can "steal" the capabilities of top US models at a very low cost. Anthropic alone has recorded 16 million unauthorized calls from Chinese companies like DeepSeek. - **Taiwanese Perspective**: As Taiwan is on the front lines of the US-China tech war, this news highlights the geopolitical risks in the AI field. Taiwanese AI companies need to be more careful about compliance and information security when using or interacting with models from different camps. - **Discussion Points**: - What is the principle behind "model distillation" attacks? Why has it become a serious security threat? - To what extent can the alliance of US tech giants curb this behavior? - Does this signal a future where the AI field will see a more pronounced confrontation between a "US tech camp" and a "Chinese tech camp"? - **Script Suggestion**: This is basically the AI version of "The Avengers" assembling! OpenAI, Google, and Anthropic, who are usually at each other's throats, have now decided to put aside their differences to form a united front against a common enemy: model copying from China. The so-called "model distillation," to put it plainly, is using a small model to constantly ask a large model all sorts of questions and then mimicking its answers. It's essentially stealing the "intelligence" of the large model to train up your own small one. We just mentioned earlier how DeepSeek's model is both powerful and cheap, and now this news might offer a possible explanation. This is no longer just business competition; it has escalated into a national-level technological battle of offense and defense. Taiwan is caught in the middle and really needs to tread carefully. In the future, when choosing technology partners, we might have to consider not just performance and price, but also which "camp" they belong to. ## Closing Paragraph Alright, today's news was incredibly dense! From the strategic battles between Meta and Google, to the huge impact of AI on jobs and wealth distribution, and the tech rivalry between the US and China, we can see that AI development is entering a more complex and critical phase. This is no longer just a technological race; it's a full-blown game of strategy, economics, and politics. Thanks for listening, and we'll see you next time! ## Keyword Tags #AIModels #Meta #MuseSpark #Google #Gemma4 #DeepSeek #OpenSourceModels #ClosedSourceModels #AIJobImpact #SnapLayoffs #PwC #WealthConcentration #AIEthics #Nature #USChinaTechWar #ModelDistillation #FrontierModelForum --- ## About this article and its author Originally published on [Mark Ku's Tech Notes](https://blog.markkulab.net/en/tech-news/ai-daily-podcast-2026-04-18) License: [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) — when reusing or quoting, credit the author and link back to the original ### About the author **[Mark Ku](https://blog.markkulab.net/en/author/mark-ku)** — Software Solution Provider - 10+ years senior software engineer, now an AI Builder - Focused on large-platform architecture — North-American e-commerce, AI SaaS subscription billing - Combining AI Agents and automation to build evolvable product foundations ### Free tools built by the author All of these are free to use: - [Free PDF Sign Tool](https://blog.markkulab.net/en/tools/pdf-sign): Online PDF sign tool — draw, type, or upload a signature, then drag, resize, and download. Everything runs in your browser; nothing is uploaded. - [VS Code Refactory](https://blog.markkulab.net/en/tools/refactory): Refactory is a VS Code refactoring extension: 34 actions plus a 37-rule code-smell inspection layer with a Code Health dashboard, across 18 languages, backed by 534 tests. It learns your repo's conventions: where interfaces live, where DI is registered, whether 'use client' belongs. It ranks files by git churn × complexity so you know what to fix first, and hands any smell to the Claude Code already on your machine. Free to use, and your source never leaves your computer. - [DB-Kit Database Manager](https://blog.markkulab.net/en/tools/db-kit): DB-Kit is a lightweight, cross-platform database manager built with Tauri + Rust + React. Manage MySQL, MariaDB, PostgreSQL, SQL Server, Oracle, SQLite, MongoDB, Redis, Kafka, Elasticsearch and RabbitMQ from one consistent interface: passwords encrypted in the OS keychain, SSH tunnels, full CRUD, a visual query builder, stacked multi-statement result sets, cross-connection data transfer and compare/sync, Excel / CSV import & export, visualized execution plans, ER diagrams, scheduled backups, SQL stress testing with p50–p99 latency percentiles, a 15-rule SQL review engine, Kafka message browsing with monitoring & alerts, a bilingual UI (Traditional Chinese / English), a built-in AI assistant (natural-language SQL, AI review and tuning advice) and the dbk CLI. Free and open source (MIT), with installers for Windows, macOS and Linux. - [VS Code Super Mermaid](https://blog.markkulab.net/en/tools/super-mermaid): Super Mermaid is a VS Code extension for beautiful Mermaid diagrams out of the box: auto-colored live preview, mouse pan & zoom, high-res PNG / SVG export, 21 templates and multiple themes. Free and open source (MIT). - [React Super Mermaid](https://blog.markkulab.net/en/tools/react-super-mermaid): react-super-mermaid is an open-source React component library: render beautiful Mermaid diagrams with a single , with built-in colorful / sketch themes, pan & zoom, in-diagram search, and high-res SVG / PNG export. Lightweight, SSR-safe, fully typed. Free and open source (MIT). - [Jira / Confluence Super Mermaid](https://blog.markkulab.net/en/tools/jira-super-mermaid): An Atlassian Forge app: write Mermaid syntax directly inside a Jira issue or a Confluence page and get flowcharts, sequence diagrams, state machines and Gantt charts. 11 diagram types, SVG / PNG export, light and dark themes, full CJK support. Runs on Atlassian: your diagrams live in your own site and the app calls no third-party service. Free, coming soon to the Atlassian Marketplace. - [Mermaid Live Preview](https://blog.markkulab.net/en/tools/mermaid-preview): Write Mermaid in your browser, see it render instantly, and share the whole diagram as a single link. No sign-up, nothing uploaded to a server, and mermaid.live share links work as-is. - [React Intl Phone Number](https://blog.markkulab.net/en/tools/react-intl-phone-number): react-intl-phone-number is an open-source React component: framework-agnostic and antd-free, with E.164 in/out, a searchable flag / country-code dropdown, configurable validation levels (strict / mobile-strict / loose), themeable CSS, and i18n — phone logic powered by google-libphonenumber. Lightweight and fully typed. Free and open source (MIT). - [Uptime Kuma Cluster](https://blog.markkulab.net/en/tools/uptime-kuma-cluster): Turn single-node Uptime Kuma into a highly available cluster: OpenResty + Lua smart load balancing, shared MariaDB state, health checks and automatic failover, plus cluster-management REST APIs. One Docker Compose command to start. Free and open source (MIT). - [Special Education](https://blog.markkulab.net/en/education): Learning materials crafted for special education students ### Daily podcasts - [Mark's Tech Insights — Daily AI News](https://blog.markkulab.net/en/category/tech-news): Daily curated AI and tech trends. Catch the latest developments via audio summaries — covering AI applications, software architecture, DevOps, and engineering practice. — RSS: https://blog.markkulab.net/feed.xml - [AI股市蝦聊](https://blog.markkulab.net/en/category/ai-stock-chat): Every trading day, an AI-analyzed take on the Taiwan stock market, delivered as a two-host conversation covering the session and the next-day outlook. — RSS: https://blog.markkulab.net/ai-stock-chat/feed.xml - [開源好物週報](https://blog.markkulab.net/en/category/open-source-weekly): A weekly two-host pick of free open-source tools surfaced from real Hacker News, GitHub, and Reddit buzz — what pain they solve and the fastest way to get started. — RSS: https://blog.markkulab.net/open-source-weekly/feed.xml ### Newsletter [Subscribe to the newsletter](https://blog.markkulab.net/en/subscribe) — Be the first to know about new posts. No spam, unsubscribe anytime.