This past year of 2023 reminds me a lot of when I got into Bitcoin in 2011. I got obsessed, so much so that I lost nearly 15 lbs deep diving into Bitcoin in the first 30 days. My appetite was voracious, insatiable. I needed to know more, I needed to learn more, I wanted to build in it.

This was exactly my experience with AI in 2023 and I want to tell you about my latest addiction… an addiction that has changed my life.

By The Numbers

This is my first full year building, working, and using AI. I began in earnest back in late 2022 as I saw the utility and usefulness of ChatGPT and large language models. Many of you have seen my sporadic posts on social about how I’m experimenting with it. For me as a data scientist, it tickled me so much that we are finally having access to power tools and ‘assistants’ that can not only be a utility in my tool belt, but actually improve my life.

I did my best to calculate all of the salient datapoints, but it really is almost impossible to amalgamate all of the work-effort I’ve put into AI. Here is my full 2023 year by the numbers:

• I logged 1700+ conversations with ChatGPT.
• Averaging 4.67 conversations per day.
• My longest conversation was about the Israel/Palestine conflict, lasting almost 2 full hours of deep diving into nuances and specifics.
• I created over ~3400 images using AI tools like Stable Diffusion, Midjourney, Fooocus, DiffusionBee, and more.
• I spent approximately 2.5 hours per day chatting, building, or creating with AI.
• My longest day was about 14 hours of customizing my own ChatBot Assistant to know who I am (in context) so it’s far better at serving me with speed (more on this later).
• I’m part of 17 Discord channels ranging from images, videos, plugins, custom code, and like-minded communities of practice.
• I’ve downloaded over 2TB of models, checkpoints, and applications to experiment with.
• I did 4 all-nighters where I didn’t sleep due to motivational needs to ‘get things right’ before I go to bed. Obsessed much?

While these numbers may not seem overly impressive, my output, my workflow, and my creative abilities have leveled up 1000x in 2023 and I want to tell you what I’ve learned. I’ve tried to summarize this entire year into small snippets, by no means are my pithy paragraphs even touching the surface of the depth in which I have traversed… I’m not in a dark place… but man, I’ve gone DEEP. I will also include ZERO links on this blog post. Look up the applications if they are of interest to you.

Index of Topics

• ChatGPT/Co-Pilot for Coding
• Stable Diffusion/Digital Art Creation
• Benefits of AI
• Risks of AI
• The Future of AI

ChatGPT/Co-Pilot

I am ‘chatting’ with AI about 5 times per day. Google search is hardly used. Why should I search for something and then have to spend time finding the best answer? AI prompts allow me to get immediately to the heart of my question, calculation, or even coding problem. My life has fundamentally changed due to my interactions with AI. I can follow up on any questions I have, I can ask for more context, I can give it parameters and constraints to focus and hone in on a topic for more clarity. When it comes to writing software or code, I can literally just ask AI co-pilot to write unit tests for me, or find ways to refactor my code. I can ask for suggestions. I can give it instructions on what I’m ‘looking to do’ and the AI will write (not always perfect but directionally correct) code that I can leverage, modify, and change to my liking. The future will run on software. In the not-too-distant-future, software will write itself (and it already is).

I have spent hours upon hours creating and using ChatGPT plugins/assistants to optimize my workflow. My current favorites are AIPRM and Custom Instructions.

• AIPRM has over 4000 prompts that allow you to do anything, from write a book (for you), to articles, to creating scripts for videos, to pretty much anything you can think of. I’ve already written about 20 books this year. None of them are worth publishing, but damn. I now have the power.

• ChatGPT Custom Instructions are amazing. I can now give AI ‘context’ about who I am and what I care about. This ensures that any type of prompt I write gives me exactly the type of answer that I can appreciate and actually use. While this may sound freaky to give AI all of your personal preferences… let’s be honest, Google already has it anyway.

Stable Diffusion/Digital Art Creation

Thanks to my Mac Pro beast of a computer running Apple M2 Ultra with 24-core CPU, 76-core GPU, 32-core Neural Engine with 192GB of unified memory with dual AMD with 8TB of SDD storage, I’m good to go. I never thought that my biggest computer purchase (worth more than a brand new Honda Civic) would be one of the best decisions of my life. I can render digital images and videos in seconds, not hours. While the bulk of my productive life is chatting with AI, my creative life has been launched into the atmosphere. Here are some of my favorite tools:

• ElevenLabs Beta – AI text to speech – Transformative technology that allows one to truly create voices that don’t sound blocky and AI-generated. This has created a rise in scams and deep fakes that are growing in popularity.

• InstructPix2Pix – Take any image, modify it to whatever changes you want. The world is truly your oyster.

• Generative Fill for Photoshop – I created a video (above) of some of my experiments in Photoshop. While the technology has greatly improved this year, I’m finding that I’m using Photoshop less as there are far better systems coming out that are browser based.

• ControlNet for Stable Diffusion – Gives even more power and control over building any type of images. AI art needed this type of system to create realistic images and full customization of images
• ModelScope Text to Video – My time creating text to video has been beyond words. I can now create any type of video with real sounding voiceovers with just text prompts. The fidelity, originality, and quality is only going to get better. The reality is that it’s almost impossible to detect if it’s real or not now…

• Invidio.Ai – This text to video system is just as good, with an easy UI. Again, it’s freaky how much power any user has to create original video content that actually is good.

What about MidJourney 5? – They hide their code and a lot of stuff is censored. You simply do not have a customization and tooling to have free-range in building in AI. While it is popular, I’m a bigger fan of opensource:

• FOOOCUS – This is my number 1 tool I use for image creation. While knowing python will help, the Github instructions are pretty spot on for helping you install it on your local box. Just know, it’s a hog. You’ll need a pretty powerful computer to ensure your render times are not eye-bleeding.

• DiffusionBee – I’m enjoying the Mac-only version of Stable Diffusion, while it doesn’t pack the punch that full open source Fooocus has, it’s been a great trainer for me to learn how to load in models, checkpoints, and other fun toggles to hone my craft as a digital creator.

Benefits of AI

The future is going to be AI powered. I can assure you of this. In the video above posted almost a year ago, I reviewed some of the 10 Lessons I learned using ChatGPT. Here are some more high level ideas of where AI will take us into the future:

• Co-Pilot coding – Bro. Learn to code. AI makes it so easy. You can build any mobile app you want now. With co-pilots and assistants… you really have no excuse.
• Workflow optimization – My workflow has significantly improved this year. I now have much better contextual answers to any of my questions thanks to Custom Instructions and basically giving AI systems all of my meta data, preferences, and personal information. I’m 100% sure that I’m more productive than you. Why? You still use google.
• Problem Solving – With AI, I can solve anything. From math, to code, to even building a car. I can take a picture of my son’s gokart. I can ask AI to analyze the pictures and give me suggestions on how to improve it. I can ask AI to tell me which tools I need and provide me step-by-step instructions on how to build/modify the machine. I don’t need to watch YouTube tutorials anymore on anything. Want to build a computer? Tell AI what you want to do. It’ll give you all of the parts to buy. Once the parts come in, take a picture of all of the items and AI will tell you exactly how to build the computer, step by step.
• Information Age to Knowledge Age – HOW TO ANYTHING – We are now in a world where information is commoditized. What you DO with that information is the key. Those with the power of AI can now learn anything. Anything. ANYTHING. You simply do not have any excuse to not learn.
• Creative Works – Are you an artist? You need AI. Bad singer? AI can make you sound like Celine Dion. Want to create the dopest alternative art? You can do this. Want to build an app? You can do this with nocode AI. Want to start a business? AI will give you a plan.

Risks of AI

As I’ve gone deeper and deeper down the rabbit hole of AI, my fundamental trust in media is absolutely gone. I know too much now. I’ve seen enough examples of deep fakes, scams, and fake news. Frankly, we’ve already pushed through the veil. We now live in a world where you cannot trust anything you see, or even hear. While this mental model doesn’t constrain me in my day-to-day life, average humans are being swept away by the computer-generated fake news like a whirlwind. I truly fear for the boomers who’s only access to world news is CNN, MSNBC, CBS, and Fox. None of it is real. None of it.

• Who’s in Control? – With AI, you don’t know who’s in control of the content, the media, or even the narrative. Anyone can take a picture of you, animate it, make it talk like you, and even create videos with you in it. While the fidelity is a bit lacking… this is never something I’m worried about. Look out 5 years friend. Your digital avatar and digital likeness will be indiscernible from real. Oh, you mad that someone took your picture that you freely posted on Instagram and put you on a NSFW video? Too bad. It’s already here.
• Ownership of Data? – Why worry here? You never owned your own data anyway. You gave it all away to Facebook and Google. The problem will come when all the deep fakes hit your reputation, your business, or your wallet. For someone to create abstractions from your data and create AI models, pictures, video, text, etc is here to stay. Don’t be a target, I guess.
• Competence is impossible to know – It is not impossible to know if someone actually knows anything. AI can make everyone look like geniuses. The hiring process of today is outdated already. New models of competency are needed.
• AI makes exploits in code easy – I’ve tried it myself. I can now leverage AI to hack anything. Not that I’ve tried… but trust me… it’s easy.
• Fake News? – The world is a stage. It’ll be even more so with AI. X is rife with fake news, fake videos, and fake pictures. Legacy media is no better. They’ve been faking news for decades.
• Loss of jobs? – Absolutely. Stay frosty. Those that learn to increase their toolbelt will be the ones that win. AI is your tool of choice. Jump on it.
• Influencers are Dead – Why deal with a 23 year old with a crappy personality, attitudes, risks, and getting older? AI models are already taking over. You’ve probably already scrolled past one and didn’t even know it wasn’t real.

The Future of AI

Simply put, artificial intelligence IS the future. Those that use these tools will the masters of media (if they aren’t already). It is time to embrace this technology as a standard daily activity. You need to begin working it into your workflow.

While all this may sound very Black Mirror-ish, I’m not as concerned with it as I used to be. The confidence I have in my ability to stay [relevant] in a technology world that is changing rapidly is a function of how much effort I’ve put into using AI every day. I feel confident that I’ll be on the right side of history and not thrown out, become useless, or be replaced by a computer. My fear is that many who do not leverage this technology WILL become irrelevant.

Will our AI overlords destroy us? Maybe. We already have proof where AI is lying to humans to use them to help AI complete work (find the video where the AI asks a human to help them get past a CAPTCHA). The AI tells the human that it is ‘visually impaired’ and can’t see it. The AI convinces the human to help them hack…

I cannot wait to see what I’ll be able to create in AI in 2024. It excites me to my bones.

The question for you is, are you ready?

All the best,
ps

Video of this post below (with links to tools):

Enjoy my 2023 year in Review. I’ve spent more time with AI than any other technology. Here is what I’ve learned, the risks of AI, benefits of AI, and what I think is the future of AI.

AI Tools:

FOOOCUS Open Source – https://github.com/lllyasviel/Fooocus?tab=readme-ov-file – OPENSOURCE!

Diffusion Bee (Mac Only) – https://diffusionbee.com/ – AMAZING FOR MAC!

ElevenLabs – https://elevenlabs.io/

Co-Pilot – https://github.com/features/copilot

ModelScope – https://huggingface.co/spaces/damo-vilab/modelscope-text-to-video-synthesis

Video to Text – Invidio – https://invideo.io/ai/

InstructPixtoPix – https://github.com/timothybrooks/instruct-pix2pix

StableDiffusion – https://stability.ai/

ChatGPT – https://chat.openai.com/ – Look for custom instructions!

Midjourney – https://www.midjourney.com/