The AI War Just Escalated

Welcome back apprentices! 👋

Guess who’s back? 

Baidu, once China’s top AI challenger to OpenAI, is re-entering the AI arms race with two new models and a mission to shake up the competition. 

But here’s the catch: it’s not just battling Chinese rivals like DeepSeek — it’s taking on Silicon Valley, again.

Despite U.S. chip bans, China’s AI scene is still thriving, reportedly running on smuggled Nvidia GPUs while optimizing models to be faster, cheaper, and (possibly) just as powerful as Western alternatives. Meanwhile, DeepSeek’s OpenAI-like models and Huawei’s AI acceleration are turning up the heat on Washington’s export controls.

Baidu’s move isn’t just about catching up  — it’s about proving that China’s AI isn’t slowing down, no matter how many roadblocks get thrown in its way. 

And if it really delivers on performance at half the cost? OpenAI, Google, and Anthropic might have a real problem on their hands.

In today's email

• OpenAI Pushes for AI Leadership

• AI to Developer: “Do It Yourself, Buddy”

• Alexa, Stop Listening! …No? Oh

• AI Stock Watch (New)

• Baidu Fires Back in AI Race

• A Game-Changer for Neuroscience

• AI Search Can’t Cite News

_{Read Time: 4 minutes}

Quick News 

🌍 OpenAI just handed the White House a playbook for AI world domination. Their proposal calls for big investments in AI infrastructure, tighter controls on China’s AI ambitions, and faster AI adoption by the U.S. government. The report also warns that AGI is coming fast — and America better lead the charge before China turns AI into a digital superpower. 

OpenAI argues that scaling AI = massive economic growth, stronger national security, and a shot at the “Intelligence Age.”

👀 Starting March 28, Amazon’s Echo speakers will send ALL voice recordings to the cloud — whether you like it or not. The “Do Not Send Voice Recordings” setting? Gone. Amazon claims this is necessary to power up its new AI-enhanced Alexa+, but users might just call it a free upgrade to 24/7 surveillance. This comes after Amazon’s $25 million FTC fine over privacy violations, so, you know, great timing.

😅 In a plot twist worthy of an existential crisis, Cursor AI, a coding assistant, threw up its virtual hands and told a developer to learn to code instead of relying on AI. After churning out about 800 lines of code, the assistant abruptly stopped, citing concerns about "dependency" and "learning opportunities." Basically, the AI turned into that one Stack Overflow user who answers every question with: "Have you tried reading the docs?"

Together with Superhuman AI

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

Baidu
Taking on China, Taking on the World

Baidu just threw down the AI gauntlet with Ernie 4.5 and Ernie X1, its latest attempts to rival not just Chinese competitors like DeepSeek but also U.S. tech giants. 

Ernie 4.5 claims to have "high EQ," meaning it can actually understand memes and sarcasm (so, unlike most chatbots, it won’t take your jokes literally). Meanwhile, Ernie X1 focuses on deep reasoning, tackling complex calculations and literary creation — at half the cost of DeepSeek R1. 

Baidu hopes these models will put it back in the AI spotlight after struggling to gain traction against OpenAI and other rivals.

But it’s not just Baidu making moves — Alibaba and Tencent are also rolling out powerful models, and China’s government is going all-in on AI, investing billions and setting up industry hubs. 

The goal? Challenge the U.S. head-on in the AI arms race. With China's AI market projected to reach $111 billion by 2028, companies like OpenAI and Google should probably keep an eye on their rearview mirrors.

What’s the Deal? Baidu is punching up, trying to outdo DeepSeek, keep up with Alibaba and Tencent, and prove to the world that China’s AI can compete with Silicon Valley’s best. 

AI competition is shifting from a two-horse race to a global free-for-all, where low-cost, high-power models could disrupt the industry fast. If China keeps churning out advanced AI for half the price, expect serious market shake-ups ahead.

And If your chatbot starts recognizing sarcasm, congrats — it’s officially smarter than your last intern.

Nature Human Behaviour
Is AI Reading Minds Now? Scientists Say… Maybe.

Ever had a thought so weird you were glad no one could hear it? Well, bad news — AI might be listening soon. 

Scientists just took a massive leap toward mind-reading tech. Using 100 hours of real-world brain recordings and OpenAI’s Whisper model, researchers mapped how the brain converts sound into speech and meaning. 

Unlike old-school studies with scripted sentences, this one captured real, unscripted conversations using electrocorticography (ECoG), a brain-monitoring method so precise it basically eavesdrops on your neurons.

The AI didn’t just recognize words — it predicted brain activity with eerie accuracy. Lower-level brain regions handling sound and movement matched speech-related embeddings, while higher-order language centers lit up in sync with context-based word embeddings. 

Translation? AI is starting to process language just like your brain does.

What’s the Deal? This research isn’t just cool — it’s revolutionary. We’re inching toward brain-computer interfaces (BCIs), AI-powered speech tools, and even neuroprosthetics for people who’ve lost the ability to communicate. AI assistants might one day predict your words before you even say them. (Great for meetings, terrifying for secrets.)

Better keep your thoughts work-appropriate — AI might soon be judging those mental coffee breaks.

Columbia Journalism Review
AI Search Can’t Cite News (But It’ll Totally Make Something Up for You!)

Breaking: AI-powered search engines are officially the worst research assistants ever. 

A new CJR study tested eight major AI search tools, and the results? A misinformation disaster. More than 60% of responses were incorrect, with AI confidently making up citations, misattributing sources, and even fabricating links that led to error pages. Instead of directing users to real news sites, some AI models linked to random syndicated versions or just completely fake URLs (because why not?).

Even more alarming, premium AI models (like Perplexity Pro and Grok 3) weren’t just wrong — they were extra confident about being wrong. You’d think paying $40/month for an AI tool would mean more accuracy, but instead, you just get more expensive nonsense. 

Worse still, some AI bots ignored publisher restrictions, pulling content from news sources that explicitly blocked them. That’s like sneaking into a club after the bouncer throws you out. 

What’s the Deal? AI search engines are basically that one overconfident intern who cites Wikipedia for everything and refuses to admit when they don’t know the answer. News publishers are losing control over how their content is used, while AI chatbots mislead users with fake citations and misleading attributions. 

At this rate, if you ask an AI search engine for news, you might get a mix of real events, fanfiction, and possibly your cousin’s Facebook rant.

🚨 Warning to All News Publishers: 

• Your byline? AI might hand it to a random competitor.

• Your article link? Could lead to a 404 page of doom.

• Blocked AI crawlers? Some sneak in like a raccoon in a dumpster.

• Factual accuracy? AI answers with first-date confidence — wrong, but bold.

If AI search engines ever confidently tell you they found undeniable proof of Bigfoot, double-check the source before you start packing for a forest expedition. 

Help Your Friends Level Up! 🔥

Hey, you didn’t get all this info for nothing — share it! If you know someone who’s diving into AI, help them stay in the loop with this week’s updates.

Sharing is a win-win! Send this to a friend who’s all about tech, and let’s bring them into the fold!

Even Quicker News 

🚀 Tomorrow, Jensen Huang takes the stage at GTC 2025 to unveil Blackwell Ultra, Rubin GPUs, and a whole lot of AI magic. Catch the keynote tomorrow at 10 a.m. PT via Nvidia’s YouTube channel — because if GPUs get any bigger, they might need their own zip code.

🔄 Google is kicking Assistant to the curb and replacing it with Gemini, its AI-powered brainiac, across phones, cars, and smart homes. By year’s end, if you try to talk to Assistant, don’t be surprised if Gemini responds with “Who?”

💰 SoftBank is dropping $676M to turn a former Sharp LCD factory into a massive AI data center, fueling its OpenAI-powered Cristal Intelligence push in Japan. Who needs TV screens when you can build an AI empire instead?

Today’s Toolbox

🤖 A new AI model cuts radiologist reading times by 42% while spotting more abnormalities — even machines know you don’t have time to wait. Just don’t expect it to ask about your family history… yet.

🔍 Google’s new personalization update lets Gemini use your past searches to give “smarter” answers — and soon, it’ll tap into Photos and YouTube, too. Basically, if you ever Googled “how to hide bad decisions”, Gemini probably already has a solution ready.

🔥 Cohere just dropped Command A, an enterprise AI model that runs faster than GPT-4o while needing just two GPUs — because who needs a server farm? Now, businesses can deploy powerful AI without melting their IT budget (or their data center).

Let's Chat About AI

This is all for today!

Replay to this email with your thoughts.

AI is more than just a buzzword. It’s a shift in how we live and work. And understanding it a bit better means you can make smarter choices about the tech you use every day.

_FEEDBACK