I was listening to a podcast called Everyday AI and the host made it clear that if you work in front of a computer, AI can help you. There are ways to train GPTs (Your own custom Chat GPT-like thing) that can be built to help you do your work.
He specifically said, whatever you spend 80% of your time on is probably something redundant that can be automated. In my case that would be researching lists on different topics. For today’s example we can use music.
What I realized is the hours and hours I spent building my Playlist could be handled by a well-trained AI in the future.
One of my 2024 goals is to learn how to train custom GPTs for specific needs for people.
What is a GPT?
The acronym stands for Generative Pre-trained Transformer. A GPT is an algorithm that focuses on using regular language to deal with large amounts of data. Basically it lets you chat with the AI bypassing a need for coding language.
It relies on LLMs (Large Language Models) which are algorithms focused on interpreting language. Think of it like an LLM is an algorithm for interpreting what you mean and a GPT is an algorithm that actually performs the tasks you ask for.
Each of the companies out there use different LLMs to create their own unique GPTs. With the right skill-set you can even use the available solutions to go ahead and train your own GPT to do exactly what you need.
To visualize this a little better, when you start using ChatGPT to ask questions, it answers broadly. It assumes whatever role it wants and you leave things open to interpretation. By creating guidelines and rules, you can narrow the focus of ChatGPTs responses to have personality and care about the things you care about. You can feed it customized data and get results that are super specific to you.
I know this was a pretty gross simplification, but you can train a GPT to be your personal assistant at work with the right know-how.
Anyone can learn how to build prompts
I know that my custom GPT journey will force me to get involved with Python and learn some basic coding. I’m also aware ChatGPT and other variants can generate the code I need. They can also explain each line of code I will be using so I can understand the logic and framework behind it.
Then, the copy/paste experimentation campaign will truly be underway. I will need to create a series of datasets and complicated prompts, that when requested in a specific order, will create the GPT mind I want to work with. As time goes on, I will be able to train it to a point where it can do things like research upcoming festivals for me.
Or it can keep me in tune with any rapper that is trending or has a viral moment in my city. I see the potential in how learning to train the robot will actually get me around some of the blockers I’ve had finding people to work with. You can even teach the AI to have a bit of a personality to make the experience more enjoyable.
I will need to evaluate a bunch of tools and get a sense for the AI landscape. But once I understand the right 2 or 3 companies to give my money to, I can already see some crazy pipeline opportunities.
This will definitely help my podcasting
There are AI transcription tools that can analyze the podcasts and verbatim capture what we say. Once a conversation is in text form, AI can scan that and create summaries. It can write episode descriptions, titles, thumbnail suggestions.
AI can write out the timestamps and generate the best possible keywords. It will be able to tell me which part of the episodes should be made into clips. Then as long as I am diligent about recording the performance of all the content in an AI-friendly way, I will be able to use the same GPT to learn from the past. It will then use the clip performance results to pick better clips in the future.
I am certain the deeper I get into it, the deeper my reliance on AI will become. Given I will keep writing this blog in my own voice, I can even train the AI to write like me. It will have thousands of words to scan in different tones. I can teach it to identify my voice in the podcasts to hear my oral vs written.
By the end of the year I’d love to create a Holden-bot that could battle me via song.
I’m excited to get started on the AI learning. I got frustrated by a problem with my day job and I really think I will long term be able to provide real value to people by learning how to automate with machines.
If this interests you, let’s start a club where we build together.
Live Long and Prosper Everyone