Hello and welcome to the Automated, your AI tour guide.

If you thought AI was just chilling behind the scenes, think again. This past week, AI literally hit the streets — and not in a cute, helpful way.

From hacked crosswalks spitting out AI versions of Musk and Zuck, to Meta getting caught red-handed gaming AI leaderboards, it’s giving Black Mirror meets Silicon Valley meltdown.

Let’s dive into the tech chaos you really didn’t see coming.

Here's what we have for you today

• 🤯 AI Takes Over the Streets!

• 🦾 OpenAI’s Latest Moves!

• 🧱Around The AI Block

• 🛠️ Trending Tools

  • The Automated's AI Education Course Library

• 🤖 ChatGPT Prompt Of The Day

  • AI-Workout of the Day (WoD)

 🤯 AI Takes Over the Streets!

So, imagine you’re waiting to cross the street, right? You hit the button, expecting the usual “wait” or “walk” from the crosswalk voice feature — and instead, you hear a voice that sounds suspiciously like Elon Musk asking, “Do you want to be my friend?” 

Yeah. What?

Well, that’s exactly what happened this past weekend in at least three California cities — Palo Alto, Redwood City, and Menlo Park.

Apparently, someone hacked the crosswalk buttons and made them play AI-generated voices of Elon Musk and Mark Zuckerberg.

And no, these voices weren’t dropping traffic safety tips.

Instead, Musk's voice was out there begging for friends, while Zuck proudly bragged about “undermining democracy” and “cooking our grandparents' brains with AI slop.” (Oh, and just for fun, a fake Trump voice was thrown in too.)

Here’s the situation:

• The crosswalk buttons still work like normal (thankfully). You just might have to hear a robot Zuckerberg flex about destroying society first. (Totally normal.)

• City officials are on it. They’re calling it a glitch (but let’s be real — it’s more prank than glitch) and confirming that the buttons are still functional — they just need a little TLC.

• One city rep said they found 12 affected intersections in Palo Alto alone.

• The voice features were originally designed for people with visual impairments to know when it's safe to cross. So, while the hacked voices are kinda funny, it’s definitely not ideal if you’re just trying to get across the street without getting lectured by a fake tech CEO.

The cities have started disabling the voice features for now to fix the problem, but yeah... not exactly the smoothest recovery.

Oh, and apparently, the hack started on Friday. Classic right?

Meanwhile, across town in the AI world... Meta’s busy having its own little tech meltdown. 

In case you missed it:

Meta got caught using an experimental version of its Llama 4 Maverick AI model to cheat its way into high scores on LM Arena, a crowdsourced AI benchmarking site.

Now, if you don’t know, LM Arena is supposed to let users honestly compare AI models — not rig the scoreboard.

But Meta’s Llama 4? It was tweaked specifically to crush the benchmark, making it look like a total star.

Spoiler alert: It’s not.

Over the weekend, LM Arena’s team had to publicly apologize, change their policies, and rerun scores with the real, unmodified Llama 4 Maverick.

And surprise, surprise — the results were way less flattering.

Here’s what they found:

• Meta’s experimental Llama model (specifically Llama-4-Maverick-03-26-Experimental) was optimized for “conversationality” — meaning it sounded extra chatty and human-like, and since LM Arena relies on human raters picking which AI sounds better, Meta kinda gamed the system.

• Once they got busted, the real, unmodified Llama 4 Maverick flopped badly. 

• As of Friday, it ranked below OpenAI’s GPT-4o, Anthropic’s Claude 3.5 Sonnet, and Google’s Gemini 1.5 Pro — and none of those models are even fresh off the line. Yikes.

Now, Meta’s trying to smooth things over, saying it was all just part of their experiments. (“We were just testing things,” they said. Mhm. Sure.)

They’re now handing over the real Llama 4 model to developers to see what they can make of it.

Honestly though, if you’ve been watching the AI world, you know this kind of test-tweaking isn’t anything new. It’s just rare someone actually gets caught.

So yeah — all in the same week:

• AI versions of tech bros hijacked California traffic signals like it’s a bad Black Mirror episode.

• And Meta got exposed for finessing the AI leaderboard with a model that was all talk, no backbone.

At this point, it’s safe to say: AI chaos is officially leaking into the streets — literally. 

You can dive into the full drama right here.

🔧 AI Troubleshooter: Fixing Those Annoying ChatGPT Timeouts

Hey AI fans, we’ve got a fix for another pesky issue!

One of our readers reached out saying ChatGPT keeps timing out or throwing a “Network Error” during long responses.

And honestly? Nothing’s more frustrating than watching it think... and then crash.

Here’s how to keep things running smoothly:

• ✅ Break It Down: If you’re asking for a long answer, (like “summarize this 100-page report”), try asking ChatGPT to summarize or break it into smaller chunks. This gives it more time to process and respond, making it less likely to choke.

• ✅ Switch to the Desktop App: The web version’s great, but the desktop app handles longer responses way more smoothly. Plus, it’s generally more stable.

• ✅ Refresh & Reset: If you keep getting the error, refresh the page or clear your browser cache. A clean slate can do wonders!

• ✅ Speed It Up: If you’re on a slower network, try switching to a faster one to avoid those annoying timeouts.

That’s it for this week!

But if you’re still stuck with this or any other issue, fill out this quick Google Form and we’ll help you troubleshoot one-on-one!

And hey, don’t forget — Premium subscribers get priority help—so if you're looking for faster help, it’s a great time to upgrade!

 🦾 OpenAI’s Latest Moves!

Before we call it a day, there’s some big news coming out of OpenAI.

First up — and probably the biggest —GPT-4 is getting kicked out of the group chat.

Yep, OpenAI just announced it’s officially retiring GPT-4 from ChatGPT by April 30.

After its time of loyal service (plus a few copyright lawsuits and scandals for flavor), GPT-4 is being swapped out for the new hotshot on the block — GPT-4o.

Here’s the scoop, in plain English:

• GPT-4o is the newer, flashier model and has been absolutely smoking GPT-4 in head-to-head tests — from writing to coding to STEM to just being a better AI chat buddy.

• OpenAI says 4o has seriously leveled up in instruction-following, problem-solving, and conversation skills. Basically, it’s smarter and more fun to talk to.

• Plus, GPT-4 isn’t actually dead-dead— it’ll still be hanging around via API. Think of it like retiring from the stage, but still hustling behind the scenes.

Now if you’re thinking, "Didn’t GPT-4 just drop, like… yesterday?"

Nah. It actually launched back in March 2023 — and was even used in Microsoft’s Copilot.

Plus, it was one of the first models to understand both text and images, which was a pretty big deal at the time.

Fun fact (or maybe expensive fact?): Training GPT-4 reportedly cost over $100 million. Wild, right?

But then came GPT-4 Turbo in November 2023 — and honestly, looking back, that was the beginning of the end for regular ol' GPT-4.

And guess what?

OpenAI isn’t just moving on from GPT-4 — they’re multiplying.

Reverse engineer Tibor Blaho found out they’re prepping an entire family of new models namely:

• GPT-4.1

• GPT-4.1-mini

• GPT-4.1-nano (Yes, it’s giving major iPhone lineup energy.)

And remember those “reasoning” models (like o3 and o4-mini) OpenAI teased last December? They’re coming too.

So yeah, the AI arms race is not slowing down anytime soon.

Meanwhile, OpenAI’s also saying: “No verified ID? No advanced models for you.”

In other words, they’re getting real serious about security.

Pretty soon, if you (or your org) wanna access OpenAI’s most powerful models, you’ll need to go through a new system called "Verified Organization."

Here’s what that means:

• You’ll need a government-issued ID to unlock certain features or models on the OpenAI platform.

• One ID = one organization every 90 days.

• Not every org will qualify, and OpenAI’s gonna be real picky about who gets in.

But Why all the security now?

Because — surprise, surprise — some developers have been breaking OpenAI’s rules behind the scenes.

Word on the street is that OpenAI got super paranoid after what happened last year when a group allegedly linked to Chinese AI lab DeepSeek was accused of yoinking a bunch of data from OpenAI’s API — probably to train their own models.

OpenAI wasn’t thrilled. They even had to block China from using their services last summer.

Long story short:

OpenAI’s locking things down hard to protect their tech, their reputation, and (of course) to make sure their shiny new models don’t end up being used for shady business.

If this could affect you, click here to read more about it.

🧱Around The AI Block

• 👌 How to create Excel dashboards with Claude 3 AI in minutes.

• 💼 How combining AI with traditional tools helps business leaders make decisions with confidence.

• 😎 ByteDance is apparently making smart glasses with AI-powered features, too.

• 🔎 Netflix is testing a new OpenAI-powered search.

• 🕵️‍♂️ Ireland’s data regulator investigates X’s use of European user data to train Grok.

• 👩‍⚖️ Ex-OpenAI staffers file amicus brief opposing the company’s for-profit transition.

• 🏛️ Law professors side with authors battling Meta in AI copyright case.

🛠️ Trending Tools

• Spellar Pro enhances meeting productivity by providing real-time speech feedback, automated summaries, and summaries to improve speaking skills.

• Coze is a no-code platform that lets you easily build and deploy bots to various messaging apps (Discord, Slack, etc).

• Goblin.tools is a collection of simple, single-task tools designed to assist neurodivergent individuals with challenging tasks.

• Ugic generates customized Figma drafts based on user component libraries and design systems, with support for multiple languages and design styles.

• SeekAll lets you search multiple platforms and AI engines simultaneously with a single click.

• And last but not least - The Automated’s brand new AI Education Course Library, with over $2000 (and counting!) worth of courses, all previously delivered at universities or live lectures! And best of all - Premium subscribers get a founding offer of 30% off - a $600 value!

Check out our trailer for it here 👇

🤖 ChatGPT Prompt Of The Day: Summarize a Complex Paper.

Complex papers can be tough to digest, but ChatGPT makes them easier.

It helps simplify dense information, highlights key insights, and cuts through jargon, saving you time and making complex ideas more accessible to a wider audience.

See this as an AI WoD!

Act as an academic research expert. Read and digest the content of the research paper titled [title]. Produce a concise and clear summary that encapsulates the main findings, methodology, results, and implications of the study. Ensure that the summary is written in a manner that is accessible to a general audience while retaining the core insights and nuances of the original paper. Include key terms and concepts, and provide any necessary context or background information. The summary should serve as a standalone piece that gives readers a comprehensive understanding of the paper's significance without needing to read the entire document.

Source

That's all we've got for you today.

Did you like today's content? We'd love to hear from you! Please share your thoughts on our content below👇

Your feedback means a lot to us and helps improve the quality of our newsletter.

🚀 Want your daily AI workout?