AI is pretty powerful - but how can you leverage that power to enhance your own day to day interests?
Artificial Intelligence as we know it has been around for a while, but it’s really taken off this year, as I bet you’ve seen on Twitter.
AI can up your game whether you’re a startup founder, an artist of pretty much any kind - or just a regular person who likes building stuff. This note will help you understand the tools out there and how to level up for your own purposes.
Here's what we'll cover in this note
Stack (for developers)
Workflow Automation tools
Stack at a glance:
Pricing: $10 (Basic), $30 (Standard)
Sometimes you just need some inspiration and fast. Midjourney helps with that. It’s a place that lets you easily use stable diffusion through a Discord bot.
It’s in beta right now but is already being applied to several different use cases, from landing page inspo to creating beautiful artwork through prompt engineering.
You can start using the Midjourney beta right now by following along with these steps:
Option #1 - Official Midjourney Discord (free to try & test!)
Go to the Midjourney website and click on “Join the beta”
Join the official Discord server
Enter a channel that starts with newbies, such as newbies-1
Use the /imagine command
Type your prompt in the box and hit enter
Wait for your images to generate!
Option #2 - Midjourney bot in a server where it has already been set up
If you’re in a server where there is already a Midjourney bot, follow the steps starting from 4 in option #1.
Pricing: Free (up to 100 images), $10-$60/month
Lexica also lets you leverage AI for image generation. It’s a web app where you can simply enter a prompt and click generate to get some images that match what you’re looking for.
You can even search through the images that have been created with stable diffusion in the past!
It also has a custom-trained photorealistic model called Aperture. This enables you to create super realistic stock images for your business or marketing needs. Highly recommend — check out some aperture generations below!
This is more of a category of AI tools, rather than a specific one.
Basically this is software that uses AI for generating synthetic voices that you can use for things such as ads, videos, tiktoks, podcasts and more!
There’s a lot of these and many more in development. I'll cover some cool ones here:
Pricing: Free, $49 (pro)
This software is basically a writing machine. You can use AI to write killer blogs, emails, tweets, posts, etc. The possibilities are crazy and it's already being used by countless companies, studios and agencies.
Think of it critical tool for any creative, influencer, or student.
It’s also super easy to get started, simply go to their landing page, click on the button that says Get Started - It’s Free, answer a couple questions, and immediately create a project.
Pricing: Free (experimental beta)
If you’re reading this note then you’ve most likely heard of this popular AI app. It essentially lets you have a conversation with an AI, simulating a text convo with a real person.
You can literally ask it to do anything, from creating a song about flowers in the style of Drake to generating a list of startup ideas, all in a way that feels natural and intuitive. In the image below, we asked it to write a resume using markdown formatting.
Rather than going in depth of this, just read the full note below. Your mind will be blown.
Pricing: Free trial, various pricing plans
This product is similar to copy.ai in that it’s also an AI copywriter.
However, the thing it really shines at is generating content that is optimized for SEO and marketing content. You can use it to create SEO-optimized and original content for your blogs, ads, emails, socials and landing pages.
Cool! That marks the end of incredible AI tools that you can start using today. But if you’re a dev, you’re probably saying to yourself: “I already know about these tools, but how do I make one?”
Well, there are two types of devs here:
Those that want to build the underlying AI models
Those that want to build interfaces that allow non-technical people to use those AI models
Maybe both categories apply to you — I’ll be mentioning resources that apply to both groups here.
The AI Stack
The following is a diagram of the stacks used when working with AI that I got from this Medium blog.
These stacks are used at the model-layer, not really at the application-layer, which is why I’m putting it in this section.
If you’re trying to build neural networks or ML models, then Python is definitely the way to go.
It’s the most AI-friendly language because there are so many Python libraries optimized for data science and ML.
Python is also really easy to pick up if your dev skills aren’t the sharpest. The following languages are all used for working directly with ML models, but R and Python are the most applicable.
If you’re working with ML in Python, the most widely used libraries are TensorFlow, Keras, PyTorch, Numpy, Scipy, Scikit-learn, Theano, Pandas, and Matplotlib.
They all bring different things to the table. For example, you can use PyTorch to create computational graphs or TensorFlow for numerical computation.
I won’t go in-depth here as it depends on what your goal is. These are the heavy hitters below.
The most specialized IDE for machine learning with Python is definitely Jetbrains PyCharm.
Jupyter Notebook is also good for data science and ML, and would be a better choice if you’re not using Python.
Another option is simply using Visual Studio Code, as its very versatile and supports many languages.
Special shoutout to Github Copilot. It supercharges any of the above IDEs. Try it out!
Most ML engineers, data scientists, and AI hackers use some sort of visualization tool to get a clearer picture of what the model is doing or how it’s performing.
Some options include MATLAB, Seaborn, or Facets, and some IDEs even have these tools integrated!
Workflow automation tools
Some of these tools apply to developer productivity across the board.
Others are more optimized for working with AI.
You and your team can use GitHub or Gitlab as general tools for DevOps.
There’s also AzureDevOps or Anaconda (esp for data science) to ship stuff faster.
The above stack will mostly be used by devs looking to work directly with machine learning. However, the following tools are geared more toward devs that want to use AI at the application layer.
If you’re building a web app/mobile app, the most common way to integrate AI is to hit an API.
In order to build Productivity Machine, I hit OpenAI’s API in order to use the user’s prompt to fetch a response from OpenAI’s GPT-3 model and display it on the app.
Some other APIs you can hit include:
Google’s Natural Language API
Microsoft’s Azure Cognitive Services (a suite of AI APIs from computer vision to NLP similar to AWS in structure)
AWS AI Services
Rev.ai, which focuses only on speech-to-text but does it really well.
Sometimes it’s better to hit the same API twice. I did this for my app when I took the response from one call and used it as the input for a second call.
This is a process that’s known as prompt-chaining when used with GPT-3.
This is a super quick overview of the AI Stack, but if you’re feeling a little overwhelmed with this new info you’ve learned I suggest you should simply start building things in AI that you find interesting, learn as you go, and consult the vast amount of resources on the internet or even use chatGPT if you ever get stuck.
A lot of my learnings came from the AI Writer project, so make sure to do it if you haven't yet!