Ex-Google CEO Warns West To Bet On Open-Source AI Against China

Let’s dive into the world of AI and check out the latest trends.

Former Google CEO Urges West To Prioritize Open-Source AI Against China

Eric Schmidt is raising alarms. The West needs to go all-in on open-source AI. If not, China will take the lead. DeepSeek proved that last month. Their R1 model hit the scene. It’s open-source, powerful, and built better than its U.S. rivals—more efficient than OpenAI. This is a real shake-up.

Schmidt, now a big name in tech investing, sees the problem. Most of the top AI models in the U.S. are closed: Google’s Gemini, OpenAI’s GPT-4, and Anthropic’s Claude. Only Meta’s Llama is open-source.

This isn’t just about competition. Without open models, universities lose access, scientific progress stalls, research slows down, and costs increase.

At the AI Action Summit in Paris, leaders took the stage. U.S. Vice President JD Vance promised the U.S. would stay on top. No backing down.

Sam Altman, OpenAI’s CEO, had a change of heart. He admitted that OpenAI bet on the wrong side. They need a new strategy, one that involves moving away from expensive, locked models.

Still, Big Tech isn’t convinced. Google, Amazon, and OpenAI are pouring billions into closed systems, betting on profits over openness.

Schmidt disagrees. He says the future is both open and closed AI. “Altman moved too fast,” he said. The best path is a mix.

The U.S. built the strongest AI models first. But Europe still has a shot. Schmidt says the secret is in applications. Businesses need to build on top of AI. Use it. Integrate it. Scale it.

“Europe needs to wake up,” he said. “The application layer is where the power is”, he added.

He’s putting money behind his words. Schmidt just launched a $10M AI Safety Science program. Through his nonprofit, Schmidt Sciences.

Some big names are on board. Yoshua Bengio, a Turing Award winner, is developing AI risk mitigation. Zico Kolter, an OpenAI board member, is diving into AI security threats.

Schmidt also calls for global cooperation. China and the West face the same AI safety risks. He compares it to military rocket testing. Countries share data.

“Why not give them knowledge to make AI safer?” he said. The AI race is heating up. But it’s not just about dominance. It’s about keeping AI in check.

Baidu Gears Up For Next-Gen AI Launch Amid DeepSeek Disruption

China’s Baidu is gearing up to drop the next-gen of its AI model later this year. The race is heating up. Newer players like DeepSeek are shaking things up, making big names rethink their game.

Ernie 5.0 is coming. A “foundation model” with serious upgrades in multimodal skills. It won’t just process text. It will mix and match videos, images, and audio. Think text-to-video, video-to-text, seamless conversions.

These models understand language, generate content, and chat naturally. They do it all. Chinese AI companies are pushing hard. They want to catch up with OpenAI and the U.S. tech giants. DeepSeek? They made waves in January with an open-source AI model that’s cheaper and faster. Global markets took notice.

Robin Li, Baidu’s CEO, spoke at the World Governments Summit in Dubai. He made it clear—AI is evolving at lightning speed. Costs are dropping. More than 90% slashed in a year. Innovation moves fast. Lower costs mean higher productivity.

Baidu was first in China with a ChatGPT-like chatbot, Ernie, in March 2023. It had a strong start. But now, competition is fierce. Alibaba, ByteDance, and others are pushing ahead. Stock prices reflect the battle. Alibaba is up by 33% this year. Baidu is just 6%. Tencent is barely 4%. And ByteDance is private but making moves.

Baidu isn’t backing down. Ernie powers AI across its ecosystem—cloud, content creation, and business tools. Wenku, Baidu’s AI-driven doc creator, hit 40 million paying users by the end of 2024. That’s a 60% jump from 2023. AI-generated presentations? Already rolling out.

Ernie 4.0 launched in October 2023. A “turbo” version came in August 2024. No official word on Ernie 5.0’s release yet. But it’s coming. Meanwhile, OpenAI’s GPT-4o landed in May 2024. And Sam Altman is keeping quiet on GPT-5. No public timeline. The AI race isn’t slowing down. Baidu knows it. Everyone knows it.

What Is Goose AI? Jack Dorsey’s Open-Source Bet On The Future of AI

Goose isn’t just another AI tool. It’s a bold move. Developed by Dorsey’s company, Block, it gives developers full control. No locked systems. No hidden data pipelines. Just pure, flexible AI integration.

What is Goose?

Goose is a free, open-source AI framework. It makes building AI agents easier. Works with the biggest models: DeepSeek, OpenAI, Google, and Anthropic.

Use it how you want. Deploy on-premises. Run it in private clouds. Keep it within your own infrastructure. That’s huge for industries like banking, healthcare, and government. They need top-tier security. Now, they get AI without compromise.

Why this will change everything?

AI used to be a game for giants. Big budgets. Secretive models. Locked doors. Goose kicks that door open. Dorsey wants AI for everyone. Startups, indie devs, small businesses—now they can play, too.

Kevin Henrikson, an AI and blockchain entrepreneur, gets it. He says many developers want to work with AI but don’t have deep machine-learning expertise. Goose removes that barrier. AI becomes easier. More accessible.

Dorsey dropped a hint on X (formerly Twitter) back in January. “Open source everything.” That’s the plan. Goose is here to change the rules.

Open-source magic

Open source isn’t new. Linux did it. Android did it. Now, AI is next. Goose could push AI adoption into overdrive. But it also raises concerns. Security risks. Ethical challenges.

What can Goose do?

This thing is powerful. Developers can use it to:

Jackie Brosamer, VP of Data & AI at Block, shared how her team uses it. Goose syncs Google Drive with Slack. It even tracks 40 hours of meetings and delivers key takeaways.

Open AI, open challenges

For businesses, Goose is both an opportunity and a challenge. Small companies get AI superpowers. But it also forces tough conversations. Ethics. Safety. Regulations. Expect debates. Expect changes.

Scotland, the UK, and beyond—tech leaders should take note. Goose could shift AI’s future. Its open nature spreads innovation fast. But responsibility matters, too.

The open-source advantage

Goose runs under the Apache 2.0 license. That means:

• Anyone can use it

• The code is fully transparent

• Development happens in the public

• Innovation is shared

Bradley Axen, AI Tech Lead at Block, puts it simply. “We’re not monetizing Goose directly. It’s free and fully Apache licensed. Instead, we’re focused on building products that work with AI agents.” Jack Dorsey just rewrote the AI playbook.

Meet Open O1, A Game-Changer For AI Reasoning & Performance

The Open O1 project is shaking things up. It’s an open-source push to match the firepower of proprietary models, especially OpenAI’s O1. The goal? Bring cutting-edge AI to everyone. No walled gardens. No hidden code. Just pure, open innovation.

Closed models like OpenAI’s O1 dominate in reasoning, tool use, and math. But they’re locked down. Hard to access. Harder to tweak. Open-source options exist, but they struggle to keep up. Weak data. Less refined training. Efficiency gaps.

That’s where Open O1 steps in. It’s changing the game with high-quality Supervised Fine-Tuning (SFT) for Chain-of-Thought (CoT) Activation. This makes smaller models—like LLaMA and Qwen—think deeper, reason better, and handle long-context tasks once exclusive to closed systems.

How does Open O1 close the gap?

First, it trains on a dataset built to mimic OpenAI’s O1. Every detail is fine-tuned for sharp reasoning and deep understanding. Then, models like OpenO1-LLaMA-8B and OpenO1-Qwen-7B go through an intense SFT process. Hyperparameters are Optimized. Logical flow is sharpened. Adaptive scaling ensures efficient inference, making responses faster and smarter. Finally, deployment is flexible—whether on Hugging Face or local servers.

Performance speaks loud

Open O1 has been put to the test. It outperforms past open-source models in key benchmarks.

• Math (MATH): Sharper reasoning. Stronger problem-solving.

• General knowledge (MMLU): Deeper understanding.

• Complex tasks (BBH): Breaks down logic better.

• Hellaswag: Still trailing slightly but gaining ground.

The results show Open O1 isn’t just competing. It’s catching up.

What’s next?

The team isn’t slowing down. They’re doubling down on AI evolution. Plans include:

AI for all

Open O1 is built on transparency and collaboration. No paywalls. No exclusivity. Just open, ethical AI—available for researchers, devs, and businesses worldwide.

And the best part? It’s 100% open-source.

With constant innovation, tough benchmarking, and a commitment to accessibility, Open O1 is changing the AI landscape. The future of AI is open, inclusive, and for everyone.

YouTube Shorts Gets A GenAI Boost with Veo 2; Here’s What It Means

YouTube Shorts is getting a major AI upgrade. Google DeepMind’s latest video model, Veo 2, is now part of the mix. Creators can use it to generate AI-powered video clips for their Shorts. No need for complicated editing. Just type and create.

This is Google’s answer to OpenAI’s Sora, the text-to-video generator. Before, YouTube had Dream Screen, which allowed users to create AI backgrounds. Now, Veo 2 takes it up a notch. Full video clips. Standalone. Ready to drop into any Short.

Dina Berrada, YouTube’s Director of Product, says these AI tools are faster than ever. More realism. Better physics. Smoother movement. Want a specific lens or cinematic style? Veo 2 can handle it.

AI-generated videos will be watermarked with DeepMind’s SynthID. They’ll also be labeled as AI-made. But that doesn’t erase concerns about misleading synthetic content. AI is powerful. It can be tricky, too.

Using Veo 2 is easy. Open the Shorts camera. Select Green Screen. Tap into Dream Screen. Enter a text prompt. Watch AI do its thing. The feature rolls out this Thursday. It’s available first in the U.S., Canada, Australia, and New Zealand. More countries are coming soon.

South Korea Says No To DeepSeek, Privacy Issues Prompt App Store Ban

South Korea just pulled the plug on DeepSeek. No more new downloads. The country’s data protection watchdog made the call. Officials say the AI chatbot will be back once it meets South Korea’s privacy laws. For now, it’s off-limits on Google Play and the App Store.

DeepSeek blew up fast. Millions in South Korea downloaded it in a week. Topped the charts. Then came the scrutiny. Governments around the world started asking questions.

South Korea wasn’t alone. Taiwan and Australia also blocked it from government devices. Italy followed. Data regulators in France and Ireland want answers. They are looking into how DeepSeek handles user data.

The app’s privacy policy suggests it stores data in China. It may collect personal details like emails and birthdates. That’s raising red flags everywhere. In the US, lawmakers are pushing to ban it from federal devices. Some states already did. Texas, Virginia, and New York moved fast.

DeepSeek’s AI model is shaking things up. It can reason like OpenAI’s O1 but costs way less to train. That’s making people rethink the billions being poured into AI in the US and beyond. For now, South Korea says no new downloads. If you already have it, you can keep using it. Or just head to DeepSeek’s website.

Loved our newsletter?

Stay in touch with us by subscribing to our newsletter to receive weekly pieces and updates.

Please rate our newsletter below. Your feedback matters to us.

Thank you for reading

-Shen Pandi & Team