Hey all, I want to create something big and helpful using AI. Can you suggest some ideas so I can combine them?
1 Like
Let me know when you figure it out.
The entire industry is trying to come up with a use for AI (beyond writing uninspired letters to sports celebrities.)
AI produces a façade of convincing intelligence that makes you think there are applications waiting to be built, but the unpredictable nature of what it produces (and that AI companies won’t leave the pretraining and qualities alone), the lack of separation between instruction and data, and the open input it can take, makes it a challenge to find a useful niche that can perform 100% reliably across every scenario you might have for AI in a limited domain.
For example, a vile AI use reveals its instructions on-demand, contrary to the guidance within, then ready for the provider to fingerprint this system prompt and ban the API account (and this app can be repurposed and realigned in an instant)
2 Likes
It sounds like you are putting the cart before the horse.
The good first question should be (should always) be:
What big problems do you have that you want to solve to be helpful?
Then, and only then, should you ask:
How does AI help me do that?
AI, ML, and LLMs are tools.
They are nothing more.
Sometimes they are the best tool for the job - but many times they are not. There are a great many tasks that are much better done with traditional programming tools.
Now… sometimes LLMs work well to supplement those tools.
In particular, I like to think of LLMs being a good “translation” layer between the analog, human, world and the digital, computer, world. We see this translation when we look at features that modern LLMs provide such as tool/function calling, where the LLM can identify what functions will do tasks that can answer the user’s question. Or in embeddings and RAG solutions, where human language is turned into a set of numbers that can be used for other vector math.
So find the problem you want to solve…
…then find the tools that will solve it.
3 Likes
why not ask gemini?
Me: I would like to develop an app that will show case the power of AI. Can you give me ten ideas for apps
10 AI-Powered App Ideas to Showcase AI’s Power
Creative and Engaging Apps
- AI Art Generator: Users input keywords or images, and the app generates unique pieces of art using various AI styles.
- AI Storyteller: Users provide a starting point, and the app generates a captivating story, allowing users to choose different plot directions.
- AI Music Composer: Users select a genre or mood, and the app composes original music pieces.
Practical and Helpful Apps
- AI Personal Assistant: Goes beyond basic tasks, offering intelligent suggestions, learning user preferences, and proactively assisting with daily life.
- AI Mental Health Companion: Provides personalized support, coping strategies, and mood tracking using AI-powered algorithms.
- AI Nutritionist: Analyzes dietary habits, offers personalized meal plans, and provides real-time food recommendations.
Innovative and Futuristic Apps
- AI Fashion Stylist: Creates personalized outfit recommendations based on user preferences, body type, and occasion.
- AI Home Interior Designer: Visualizes different design styles and furniture arrangements based on user input.
- AI Language Learning Tutor: Provides interactive lessons, real-time feedback, and conversation practice using AI-powered language models.
Educational and Informative Apps
- AI Science Experiment Simulator: Allows users to conduct virtual experiments and explore scientific concepts with AI-driven visualizations.
These ideas offer a diverse range of possibilities to showcase the capabilities of AI in different domains. Each app can be further developed and refined based on specific target audiences and user needs.
Would you like to explore any of these ideas further, or do you have a specific area of interest?
@threatthriver , you are either a real 10x developer or late to the party. 
I am fast, and I needed approximately 12 days with testing and feedback. I mean, I could have done it in one week with intense crunching and a clear UI/UX schema already in place. And I still have to do the video. 
But I guess the real challenge, and the reason for this challenge at all, is to find that idea. Sea LLMs or AI, as the marketing says, are brilliant but with very limited fields of use. There is a huge hype going on, and companies are struggling to find good ways to use AI—ways they can make money and create products that are actually good. Of course, there are some out there, but they are already tackled (picture generation, translation, search engine replacement, info filtering, bad content creation ;-)).
I think the best bet is a new generation of search engine that is up to date. Like improvement of the google standard search.
The suggestions from above from the LLM have one thing in common: excuse my wording, they *****. All businesses with them will fail. I guess the challenge is more about finding an idea for commercially using a paid LLM API than about the app itself.
So we all have a very nice and shiny tool in our hands, but we need to figure out a way to use it. And it could well be possible that there is no new way. I am proud that I spotted one that is a little bit useful at least, but this one I’ll keep to myself until release and I am thrilled to see what the other entries come up with. Not so much the apps themselfs.
1 Like
Web Developer Quality Tool (SaaS with App)
The Problem:
Errors in web development can often go unnoticed because they don’t always produce error messages. This makes them difficult to spot. That’s where our tool comes in.
Example:
Consider the Google Gemini AI competition page. Switching the language to German, for instance, reveals some issues.
The App:
The user registers and enters their webpage.
The webpage is scraped.
After every update, and automatically every hour, the page is scraped again and compared to the previous version.
If there are significant problems (e.g., page down, slow loading), the user is notified.
Differences in generated text/HTML/CSS/JS are checked by Gemini AI with enough context and prompts such as: “Will this generated code look good to human readers?”
If the AI spots mistakes, the user is notified.
The Solution:
Mistakes can be spotted even if there are no error messages. The Web Developer Quality Tool would have prevented this. The developer would have received an email with a message from Gemini that the page contains probably mistakes.
1 Like