From Igor @ AIadvantage.
This week we're sharing with you a Google game that trains your image prompting skills, a free web app that helps you build perfect ChatGPT prompts.
Why you should care? This free app from Google Labs gives you a hands-on, challenging, genuinely fun way to learn AI image-prompting.
It's more like a game than anything else—you're given an AI-generated image and are challenged to guess what prompt was used to create it. Once you input your guess, it generates an image based on your prompt and compares your image with the original, giving you a score from 1-100 based on how closely the two images match.
Say What You See is way more challenging than it looks. It's difficult to get any score higher than a 60-70% match on images.
News That Matters: Are AI Labeling and Watermarks Actually Harmful?
Why it Matters? Most experts agree that AI won't be creating killer viruses or launching nukes any time soon. The current destructive potential for AI lies in its ability to create realistic text, images, and video that are intended to deceive people.
The biggest players in the AI space are keenly aware of this issue, and they're actively taking steps to reduce the negative impacts of AI.
On Tuesday, Meta announced new plans to combat the growing threat of disinformation. Meta is in the process of working with industry partners to develop technical standards for identifying AI content. Once these standards are established, all Meta platforms (like Facebook and Instagram) will automatically identify and label AI-generated images.
While it's good that companies like Meta are working on putting both visible and invisible watermarks on AI images, it does not solve the issue. In fact, it may make it worse. Why?
For this approach to work, Meta needs to ensure that 100% of all AI-generated images are tagged. There can be no room for error.
If you're told that all AI-generated images on Instagram will be flagged as such, you get conditioned to simply "look for the watermark" instead of viewing things with a critical eye. If an image slips past Meta's detection algorithm, you'll be completely unprepared to catch the fake. "No watermark" would become an implicit seal of authenticity on images.
Meta's approach will hurt our ability to view things with a critical eye. Society is in a great place right now. We're being cautious about everything we read, see, and hear online. In a way, the threat of AI-generated disinformation has trained us to be more aware than ever before. Watermarking and labeling will walk all of that progress back, and we'll become dependent on corporations and governments to tell us what is and isn't real.
So, what do we do?
View everything with a critical eye. If you sense deception, call it out. It's up to us to create the world we want to live in!
Prompt of the Week: Perfect Prompt Builder
We're giving you something slightly different this week. Instead of a copy/paste prompt, we're giving you a web app that helps you build perfect prompts in seconds! It's a great way for beginners to learn how to build high-quality prompts that give the best results.
How to Use:
- Click on each item in brackets to fill out the builder. For some like "Select a Format", you get a drop-down menu with options to remind you of all the outputs ChatGPT is capable of generating. For others like "Enter an Example" you'll need to enter in your context manually.
- Hit the "Example" button at the top to get an idea of how it's supposed to be filled out. The "Clear" button will reset the builder.
- Once you're satisfied with the prompt, hit "Copy to Clipboard" and you're ready to use the prompt!
While it's labeled "ChatGPT Prompting", this format works well for all LLMs. This includes Gemini, Claude, and any chatbots you run locally. Keep this bookmarked right next to your favorite chatbot to save a huge amount of time crafting prompts!
That's all for this week...
See you next Thursday. Much Love, Igor
-->
Was this email forwarded to you?
Sign-up for the weekly AI Advantage newsletter to stay up to date on AI tools and prompts that actually matter.
-->
Follow The AI Advantage: