Baidu Just Launched Ultra-Cheap AI

Hola Decoder😎

If someone forwarded this to you and you want to Decode the power of AI and be limitless, then subscribe now and Join Decode alongside 30k+ code-breakers untangling AI.

🤖 Baidu Unleashes Two Powerful AI Models at Unbeatable Prices
Insights from Reuters

Baidu just launched ERNIE 4.5 and ERNIE X1, two advanced AI models designed to rival top competitors at a fraction of the cost. These models aim to push the boundaries of reasoning, language understanding, and multimodal capabilities.

The Decode:

🔹 ERNIE 4.5: Smarter, Faster, and Cheaper - Baidu claims ERNIE 4.5 surpasses GPT-4o in benchmarks while costing just 1% of its price. It features enhanced EQ, better logic, improved hallucination prevention, and stronger coding skills.

🔹 ERNIE X1: Baidu’s Deep-Thinking AI - The ERNIE X1 model matches DeepSeek R1's reasoning power at half the cost, excelling in complex problem-solving, document analysis, and structured "thinking" approaches.

🔹 Multimodal Advancements & Real-World Applications - Both models support text, video, image, and audio processing, allowing for more human-like interactions, better understanding of memes, and cross-format content generation.

🔹 A Bold Play in the AI Pricing Wars - With ERNIE 4.5 priced at just $0.55 per million input tokens and $2.20 per million output tokens, Baidu is aggressively undercutting Western AI models, signaling a major shift in AI affordability.

China is rapidly scaling AI at industry-shaking prices, forcing global competitors to rethink pricing strategies. Baidu’s move could drive AI costs down worldwide, making cutting-edge AI more accessible and accelerating AI adoption across industries.

Together with Poplar

How Kettle & Fire leveraged direct mail to win back customers

Email and digital ad performance aren’t what they used to be, and brands are looking for new, efficient ways to reengage their customers and drive repeat purchases.

That’s why Kettle & Fire turned to direct mail with Poplar to reach valuable customers who simply weren’t engaging with their email campaigns.

Using Poplar’s easy Klaviyo integration, Kettle & Fire targeted lapsed customers who didn’t open their emails with a 1:1 mailer that fit into their existing Klaviyo workflows. Not only did this immediately boost the response rates of that group by 1.4%+, it drove over 15x ROI.

Don’t take our word for it, hear it from them directly: “We use Poplar for our win-back campaigns, and they’ve consistently shown a super high ROAS with personalized messaging and content.”

Want to see how direct mail can drive more conversions for your brand?

Book a free demo today!

🚀 Autonomous AI Agents: A Paradigm Shift in AI Operations

Microsoft Copilot Studio now enables developers to build Autonomous AI Agents—systems that self-initiate tasks based on external triggers, unlike traditional AI agents that rely on user prompts.

How It Works

1️⃣ Trigger Activation – The agent is activated by an external event (e.g., an email or system update), initiating its response.

2️⃣ Processing & Decision-Making – It analyzes the input using generative AI, instructions, knowledge sources, conversation structure, actions, and past interactions.

3️⃣ Response or Execution – The agent either replies to the user or autonomously performs a task, such as updating a system.

4️⃣ Feedback & Optimization – Interaction data is logged for context, while performance insights are sent to analytics for future refinements.

The Key Difference

🔹 Regular AI Agents – React only when prompted by a human.

🔹 Autonomous AI Agents – Act independently, initiating workflows based on triggers like emails or system events.

Why It Matters

Autonomous AI agents revolutionize workflows by reducing human intervention, enabling proactive problem-solving, real-time automation, and seamless enterprise operations. With Copilot Studio, businesses can now deploy AI that works in the background—anticipating needs, executing tasks, and continuously improving. 

🎭 AI Search Has a Citation Crisis
Insights from Columbia Journalism Review

AI search tools are replacing traditional search engines but struggle with citing sources accurately. A study found they frequently misattribute content, fabricate links, and bypass publisher restrictions, creating a false sense of reliability that misleads users and harms news organizations.

The Decode:

• Rampant Inaccuracy & False Confidence – AI search tools got citations wrong in over 60% of cases. Grok 3 had a shocking 94% error rate, while ChatGPT misidentified 134 articles but admitted uncertainty only 15 times. Premium AI models were even more confident but less accurate than free versions.
• Ignoring Publisher Controls & Fabricating Links – AI tools bypassed robots.txt restrictions, accessing content from sites like National Geographic despite explicit blocks. Over half the URLs provided by Gemini and Grok 3 were completely made up, with Grok 3 delivering 154 broken links out of 200 queries.
• Redirecting Credit & Traffic to the Wrong Sources – Instead of citing original publishers, AI search tools are frequently linked to Yahoo News and AOL—even when the original source had licensing deals with the AI company. This diverts traffic and ad revenue away from content creators.

AI chatbots borrow credibility from trusted news brands, but when they miscite sources, they mislead audiences and erode trust. Publishers lose traffic, revenue, and control over their content. 

⚡Tools you Cannot Miss:

📊 Neurons – Predict ad performance before launch and boost engagement with AI-driven insights. Book a free demo today!

📝 AI Auto Blogger – Generate high-quality blog posts effortlessly with AI. Save time and keep your content fresh.

💬 Whimsey – Analyze customer feedback with AI and turn insights into action. Improve experiences and boost satisfaction.

⚡ SQLFlash – Optimize SQL queries instantly with AI. Eliminate manual tuning and improve database performance.

📱 PhoneCaseAI – Instantly design and create stunning custom phone cases. AI-powered personalization at your fingertips.

🚀 Quick Hits

❓ What if TV ads could be your next big growth channel? Boost Oxygen turned to WonderAds when digital ads plateaued—and discovered that TV wasn’t just for big brands. With strategic linear and streaming placements, they drove more traffic, boosted sales, and built brand trust. Ready to tap into TV’s power? See how WonderAds can help!

🛍️OpenAI and Google are lobbying the US government to allow AI training on copyrighted content, arguing it's vital for maintaining America's competitive edge in AI. They claim restricted data access could hinder innovation and advantage international rivals like China.

🔍 Google has integrated Chirp 3, its sophisticated speech-to-text and HD text-to-speech models, into its Vertex AI platform. This update, which supports 31 languages, plans to enhance voice AI applications, such as voice assistants and audiobooks, while ensuring responsible usage with new restrictions.

🤖 Google's Gemini AI model is increasingly being utilized to remove watermarks from images, including those from prominent platforms like Getty demonstrating the model's advanced technology.

🚩 Y Combinator CEO reported that nearly a quarter of their current startups now have 95% of their code written by AI, enabling small teams to reach $10M in revenue with dramatically fewer engineers.

Thanks for Decoding with us🥳

Your feedback is the key to our code! Help us elevate your Decode experience by hitting reply and sharing your input on our content and style.

Keep deciphering the AI enigma, and we'll be back with more coded mysteries unraveled just for you!