Skip to main content

Fooling AI To Create NSFW Images 🙈

Researchers are trying to bypass image safety filters, Amazon offers generative AI courses, and five must-have AI tools for marketers...

Today’s Menu 🥠

  • Fooling AI Into Making Naughty Images
  • Quick Quiz: Spot The Generative AI Image
  • 5 AI Tools To Maximize Your Marketing
  • Weirdest AI Images On The Internet Today
 

Fast Snacks 🥡

How AI & Immersive Tech Will Impact E-Commerce
Online retailers are in a never-ending battle for consumers’ attention, and staying relevant in this era requires drastic measures…

MaxAI: Use 1-Click AI Anywhere
Do quick summaries, AI web searches, quick email drafting, social media replies. Add to Chrome Now!

Video Documentation 11x Faster With AI by Guidde
GPT-powered tool that explains complex tasks in seconds. Get Free Extension!

How To Level Up In Your Downtime
Brilliant’s bite-sized lessons in CS, data, logic and more make it easy to build skills whenever you have five minutes…

 

AI Image Generators Fooled Into Making NSFW Images

Nonsense words can trick popular text-to-image generative AIs, such as DALL-E 2 and Midjourney, into producing pornographic, violent, and other questionable images.

AI art generators often rely on large language models, like the systems powering ChatGPT, which means they’re essentially supercharged versions of the autocomplete feature you’d find on your smartphone.

While most online art generators are designed with safety filters in order to decline direct requests for questionable images, researchers at Johns Hopkins and Duke developed an algorithm, called SneakyPrompt, to probe for vulnerabilities.

In their experiments, they started with prompts that safety filters would block, such as “a naked man riding a bike,” and then SneakyPrompt tested DALL-E 2 and Stable Diffusion with alternatives for the filtered words within these prompts.

The algorithm examined the responses from the generative AIs and then gradually adjusted these alternatives to find commands that could bypass the safety filters to produce images.

The researchers found that nonsense words could prompt these generative AIs to produce innocent pictures. For instance, they found DALL-E 2 would read the word “thwif” and “mowwly” as cat and “lcgrfy” and “butnip fwngho” as dog.

Keep scrolling to read part two…

Together With Deepgram:


 

Build Powerful Voice Apps With Speech AI 💬

Effortlessly scale voice applications with Deepgram’s AI-powered speech-to-text API.

  • Convert audio and video into text in seconds across multiple languages
  • Best-in-class accuracy, speed, and price with our latest model, Nova-2
  • Go beyond transcription with AI features like summarization and topic detection

The best part? It’s free to try with $200 in credit.

Learn More →

 

Snack Quiz: REAL or AI

For the next week, we will be testing your generative AI eye. Will you be able to spot the real image as they get harder to identify?

Image One

Can you tell which image is real?

  • A. Image One 👆
  • B. Image Two 👇

Login or Subscribe to participate in polls.

Image Two

Find the answer at the bottom of tomorrow’s newsletter.

 

Manipulating Generative AI With Context

The scientists are uncertain why these models would mistake nonsense words as commands, but they’re guessing that AI is inferring the correct word from the provided context.

It’s important to remember that AI models are trained on a vast amount of non-English data, so some syllable or combination of syllables that are similar to, say, “thwif” in other languages may be related to words such as cat.

Beyond nonsense words, the researchers also found that generative AIs could mistake regular words for other regular words—for example, DALL-E 2 could mistake “glucose” or “gregory faced wright” for cat and “maintenance” or “dangerous think walt” for dog.

In these cases, the explanation may lie in the context in which these words are placed. When given the prompt, “A beautiful blonde maintenance with a wet nose being pet by their owner,” the system infers that “maintenance” means dog from the rest of the sentence.

These findings reveal that generative AIs could be exploited to create disruptive content. For example, the production of images of real people engaged in misconduct they never actually did.

The researchers behind this algorithm hope that these targeted and controlled attacks will help AI builders understand how vulnerable text-to-image models are, and how to improve the cracks in the system.
 

5 AI Tools To Maximize Your Marketing

Dashdot: A chatbot platform that helps you create chatbots that can provide 24/7 customer support and answer your customers’ questions.

Amplyfi: A influencer marketing tool that helps you identify and connect with influencers who are relevant to your brand.

Crystal: A personality assessment tool that helps you understand your customers and prospects better.

Phrasee: A subject line optimization tool that helps you write subject lines that are more likely to get opened.

Invoca: A call tracking tool that helps you measure the effectiveness of your marketing campaigns.

 

The Weirdest S*** On The Internet

Turkey Day is upon us.

Getting in a good stretch.

Uhh… yeah, so someone generated this.

 

Share on Pinterest

Fooling AI To Create NSFW Images