- Inclined.ai
- Posts
- 🤖 AI Is Outsmarting Captchas
🤖 AI Is Outsmarting Captchas
PLUS: Sometimes Chatbots Need A Bad Influence
What's up? You're reading Inclined AI. A newsletter with more WPM than an Aaron Sorkin script.
FADE IN:
AI companies look to see what hackers can do with chatbots
CAPTCHA tests are not holding bots back
Google will add AI tooling to ChromeOS
Weather forecasting with AI is a whole new ballgame
Let’s hack this story together.
At the recent DEF CON conference in Las Vegas, hackers gathered not to plan digital heists or conspire in dark alleys of the cyber world but to stress-test AI chatbots and image generators from giants like Google, OpenAI, Nvidia, and more.
It was like a massive game of digital Jeopardy, with hackers challenging chatbots and the winners walking away with Nvidia graphics cards and bragging rights.
Now, you may ask, why unleash hackers on unsuspecting AI models?
Think of it as sending chatbots to boot camp, where they'll face the toughest drill sergeants in the business.
The stakes here aren't just for fun and games; they're about making AI systems safer, more robust, and less biased. Through "red teaming" exercises, hackers probe, prod, and manipulate AI models, identifying errors and biases.
They're not cracking code or infiltrating servers; they're using natural language to trick AI models.
It's a dance between human and machine, a tango where one misstep by the chatbot might reveal a weakness that genuine bad actors could exploit.
Companies' participation and the White House's watchful eye make this more than a weekend hacking spree. They compile the results into a massive database that companies will use to fine-tune and fix the glitches, all with a comprehensive report due next February.
And all of this, by the way, without costing the companies a dime. Talk about a win-win.
But what does this mean for us, the users of this fantastic technology?
Simply put, it's about safety and reliability.
By putting these AI systems through rigorous testing now, we benefit from more secure and efficient technology later.
Generative AI has landed in the data domain with a bang.
Einblick is the AI-native data notebook that can write and fix code, create beautiful charts, build models, and much, much more–all with as little as one sentence. (pretty cool right?) This is all made possible by their AI agent, Einblick Prompt.
We all know the data analytics and data science process can be a real slog. From debugging syntax errors, to searching through documentation and Stack Overflow threads…it can be a real pain. That’s all before you even get to extracting meaningful insights from your data!
Luckily, Einblick Prompt removes all the tedious parts of the process by leveraging generative AI. Since Prompt is built into every data notebook, you don’t have to copy-paste any code, or switch between tabs. The code is generated straight into the notebook, and you can run it and test it immediately.
And the best part?
Prompt is context-aware. That’s right. You don’t have to type paragraphs of information about your data to get decent results. Prompt will infer which table to use, which libraries to import, and what data types are in your dataset. Plus, if there are any errors, you can debug instantly with the “Fix with Prompt” button.
Don’t take our word for it though. Try it out yourself–it’s free. (You don’t even need your own OpenAI API key–Einblick’s got you covered.)
a robot takes a captcha test --ar 2:1 --s 400 --chaos 25 --v 5.1
Hold on to your keyboards, digital citizens, because the tables have turned:
Bots are now acing CAPTCHAs, those annoying online tests designed to keep them at bay.
You know, those twisted text challenges that leave us squinting and guessing while bots zip through with a remarkable 99.8% accuracy? Yep, that's the reality we live in now.
Researchers at the University of California, Irvine, and the University of Kent have painted a picture that's less "robots taking over" and more a reflection of the outer ripples of more intelligent AI.
While humans take 9 to 15 seconds to solve these CAPTCHA problems with a mere 50% accuracy, bots cruise through them in under a second.
But don't let the CAPTCHA apocalypse get you down.
The study shows it's time to innovate, with experts suggesting more dynamic, intelligent algorithms that study behavior rather than throw pixelated puzzles at us.
So wave goodbye to the old-school CAPTCHA and say hello to a future where bots being good at something might make our online lives more manageable.
Move over, Turing test; the bots have graduated, and they're coming to class with answers in hand.
🌐 Data centers are thriving under the new demand in AI circles
🌦️ Weather forecasting is changing thanks to AI
🧑🔧 Microsoft’s Copilot can now deploy field techs for you
👀 Look at the potential for AI to change how we communicate
📌 Tech guys can’t be the only ones who get a say in what AI becomes
💦 Companies are thirsty for more of your data to train new AI models
✍️ ChromeOS is rumored to be getting AI updates that include a writing tool
⎯ The FEC sets a line in the sand about whether or not AI can be added to political ads
Check Out Our Sunday Edition
You can dive into more AI news and topics with us every week by subscribing to our premium edition.
We’ve written about the following:
If you’re not already subscribed, that’s okay. We’re offering a free 7-day trial so that you can read this one. That’s how excited I am to post it.
WorkbookPDF - personalized language workbooks (link)
MeMemes - turn yourself into memes (link)
MindPal - dump all your files & chat w/ them (link)
PodcastGPT - instant insights from podcasts (link)
Audit Landin Page - audit your landing page copy like a behavioral scientist (link)
Co-Founder - supercharge your startup idea (link)
Notion to Chatbot - turn Notion documents into AI bots (link)
Nutrition Buddy - stop counting calories, start enjoying life (link)
Drunk Founding Fathers
This takes Drunk History episodes to a whole new level.
- That’s it for today. I hope you enjoyed the latest edition of inclined.ai - Davis.