Let’s get one thing straight, peeps.
When it comes to AI, it seems like everyone and their mothers is obsessed with the big shots.
OpenAI, Meta, Anthropic, and even Apple…they hog the headlines, including mine.
Togetger they are valued trillions of dollars, and invest billions in new Foundational Model research, and every with it, they get the last ounce of media attention. But while these biggins are busy building the flashiest AI with huge budgets, there is a quieter, a lot scrappier movement happening behind the scenes.
This, my friends, is the open-source community.
It’s a ragtag bunch of developers, researchers, and everyday tech nerds who are building an AI universe for people with, you guessed it, normal-sized wallets. I always think of them as the Robin Hood of AI, if Robin Hood also happened to enjoy coding in his mom’s basement. And like RH, they sometimes steel cash from the rich.
The open-source movement doesn’t do splashy launches or corporate boardrooms. No one here is pushing the media to be the next big tech CEO or make headlines.
They are more interested in building communities, sharing knowledge, and, yes, giving back to the AI ecosystem. Because for them, it’s about creating tools everyone can use, even if they don’t have the budget of a small country. But before I talk about what makes this movement tick, let’s take a step back and look at what open source even means.
Before we start!
If you like this topic and you want to support me:
- Comment on the article; Google appreciates that and it will really help spread the word 📢
- Connect with me on Linkedin 🔗
- Subscribe to TechTonic Shifts to get your daily dose of tech 💉
The open source software revolution
Let’s analyze it a little.
Open source is a software “movement”. That is actually a term the community might actually cringe at. It’s about creating free software.
Free to use, free to modify, and free to share.
When I visit github, or huggingface, it’s an all-you-can-eat buffet, for software code.
There’s no VIP section, no pay-to-play scheme. The goal is simple: make the software as accessible as possible so anyone can join in and make it better.
This isn’t some new trend, either.
Open source has been around since the early 1950s.
It’s the same concept that gave us the internet, the World Wide Web, and basically every tool running modern digital infrastructure. Without open-source tools, we would all still be using dial-up, and praying the phone line doesn’t ring while we’re online.
So next time you’re streaming Netflix or tweeting about your coffee, remember: open source paved the way.
Stable Diffusion and the open-source AI wonderland
If you need a prime example of open source in action, look no further than Stable Diffusion which was released in 2022.
This text-to-image tool is the holy grail of open-source AI.
And since it launched, developers worldwide have jumped in to create new image tools, new capabilities, and a whole ecosystem of innovation that didn’t require billion-dollar budgets or a mega-corporation’s approval. Stable Diffusion is the AI equivalent of a DIY project on steroids, and it is inspiring an entire generation of developers to experiment, create, and improve.
Other magic happens on platforms like Hugging Face. That is a site which is hosting over a million open-source models and demos. It is a massive digital playground where developers work together, and make AI accessible to the masses. No pricey subscriptions though, the open-source community is creating gorgeous AI-driven images, videos, and utilities that anyone can use. And they’re doing it for free, or close to it, anyway. Hugging Face and others isn’t about pretty pictures and video effects, though. The open-source community is responsible for a flood of AI utilities that keep flying under the radar. The funny thing with Hugging Face though, is that it is funded by Microsoft, Amazon, and Google… These players are not betting on one horse.
A good example of how people can use open source AI is the Ollama platform, which is quietly revolutionizing how AI is run on user devices. With Ollama you are not relying on mega-clouds like Google or Microsoft. Ollama lets you run smaller AI models directly on your computers. No mega-servers required, just a modest setup on your home machine. It is perfect for hobbyists, privacy advocates, and people who don’t want to fork over cash to tech giants just to try AI. When I meet companies who would love to use ChatGPT or Claude, but their privacy (or shadow AI) policy doesn’t allow for them to use it, I advise the use of locally installed Ollama.
The power of community-driven innovation
If you’re still thinking that open-source AI is just for basement geeks with too much time on their hands, you are mistaken my friend. The truth is that, if AI ever finds its way into your smartwatch or becomes as common as your phone’s flashlight, it will likely be because of the groundwork which was done by open-source developers.
This people aren’t hobbyists in home offices… it’s universities, research labs, and even small companies that are contributing to the open-source pool. And let’s face it, an open-source community is the tech version of a flash mob. When something needs fixing or a new feature is needed, the crowd comes together and makes it happen faster than any corporate R&D team ever could!
Sure, you might hear about big companies like Meta releasing “open” large language models like LlaMa. But when Meta says “open”, they don’t mean the real open-source deal. True open source means that nothing is locked away, and yet LlaMa is only “open weights”.
Open weights is calling a gated mansion “open housing” because you can peek through the fence. With real open source, nothing is locked, nothing is hidden. And LlaMa’s version of “open” is more like “open-ish,” and the open-source community sees right through it.
Beware the open source bait-and-switch
But not all that glitters is gold.
Recently, there has been an ugly trend in the tech world.
Some startups have figured out that “open source” has a nice ring to it, so they launch “open-source” products as a way to get attention. They roll out their code to the public, make a big deal about being part of the movement, and when they have reeled in enough users, 💥boom💥they switch to a full commercial model.
Do you know who I am talking about?
Yes indeed!
The biggest example of this is OpenAI.
The same peeps who first launched their product as an open-source AI model maker.
They captured opur hearts, minds, and our wallets (3.5 billion of it), only to pull the ultimate bait-and-switch, transforming into a profit-driven powerhouse.
Sneaky…Sam Altman, take a bow.
Why open source might win out anyway
The open-source movement in AI is thriving. Just try to count the number of ML models on Hugging Face. Bet you will stop when you’ve counted till 350k.
For them, it is not just about money..nooo… it’s about reputation, innovation, and pushing their own personal boundaries.
The ML/NN Python Modules that we are playing with, names like TensorFlow, PyTorch, Keras or scikit-learn, Mistral AI, or BLOOM may not be plastered all over billboards, but they are the backbone of the AI industry. Open source is where reputations are being made, not with flashy announcements but with good ol’, steady improvements.
Take Meta’s release of Llama, for example. Though not really 100% open source, but when they opened it up to the community, countries with fewer resources gained a unique opportunity. And all of a sudden, you got new languages and cultural data that would never see the light of day in proprietary systems.
The most recent example of open-source power came from Genmo’s Mochi-1 AI video model.
This state-of-the-art tech required a network of four H100 GPUs, which only the elite could afford. But hours after it was released, the community stepped in, and by sunset, a fine-tuned version was up and running on a regular gaming PC.
Now that is the power of open source! It is making the impossible accessible to anyone.
Research backing up the open-source impact
For anyone who thinks this is just a fad, meet Dr. Elizabeth Warner. She is a leading researcher in AI ethics and governance.
She wrote a paper about “Democratizing Artificial Intelligence Through Open Source”. And in it, she discusses how open source is leveling the AI playing field. And according to Warner, open source makes advanced technology available to more poeple, regardless of their money. And that is reducing the gap between large corporations and smaller organizations and that is because it allows for smaller players to innovate without being drowned out by corporate monopolies.
Hear hear!
👂
And then we have Dr. Eric Sangford who is basically saying the same with his study, “The Sustainability of Open-Source Development in Artificial Intelligence”. Open-source AI projects let everyone have access to new models, and help create more sustainable innovation. But what he mentioned next, is quite interesting, “Progress in closed systems, often stalls due to corporate priorities. But in open-source environments, the crowd can step in quickly, and produce solutions at a pace that is impossible in traditional R&D”. In other words, corporate agendas slow you down, while in open source, the community creates solutions much faster than in your corporate for research teams.
Top 10 open source AI
I have chosen the ten open-source AI tools below, because they each bring something unique to the table. They are, one after the other, practical, and often just as powerful as the expensive, closed-source alternatives. I have used nearly all of them in the past, from building a chatbot that doesn’t drive people mad, or analyze mountains of data without an army of data scientists, or turn my laptop into an image-recognition machine… All the tools below make this possible.
1. Acumos AI
Acumos AI is basically AI for people who don’t want to code every single line.
A.k.a. me.
It’s an AI drag-and-drop for dummies.
A.k.a. me.
If you want to build an AI models, you just throw in a few pre-built piece and voila. Your masterpiece is ready. And you don’t need (a lot of) experience in computer science to use it. Acumos lets regular folks like y’all play with AI and you won’t have to call IT support every five minutes.
2. ClearML
ClearML is a personal assistant for data scientists. What it does, is that it tracks your experiments so you don’t end up with 50 “final_final_final_v2” files.
Recognize that?
It keeps your datasets organized so you don’t lose your mind. And yes. It even syncs with cloud servers because, be honest people, nobody wants to carry all that data around on their laptops. Actually ClearML keeps your AI projects looking neat and tidy while you pretend to have everything under control. That’s why I ended up using it
3. H2O.ai
Most data scientist who are starting know the feeling of being drowned in data and just want some answers, H2O is your lifeboat. This thang hussles big datasets. And it does all the heavy work without whining. Unlike you, my friend. With H2O.ai, you can get insights that make you look like they know what you’re doing.
4. Mycroft AI
Ever feel like Alexa and Siri are just a little too ….umm… dumb.
Check out Mycroft AI. Who was the character behind Mycroft… (lookup…), oh yeah, the older brother of Sherlock Holmes! He was depicted as more intelligent than Sherlock, but he is not interested in … doing any work? Hmmm… Ok. Mycroft is supposed to be the open-source voice assistant that actually respects your privacy. You can teach it to do whatever you want without worrying about it listening in on your midnight ummm….. conversations. Oh, and it’s customizable too. It’s ethical. And it’s the voice assistant that won’t rat you out to (the cops) Big Tech.
5. OpenCV
OpenCV is a tool for all things visual. If you want your computer to recognize your face, or track objects, or spot your dog in a crowd of people, you go to OpenCV. This thang is a Swiss multi-knife for computer vision. And unlike that overpriced piece of camera you call a security device (ding dong… ), OpenCV doesn’t charge a monthly fee to detect your neighbor stealing your garden gnome.
6. OpenNN
OpenNN is …. a neural network tool (duhh). And it’s fast ! It does its thing while you sit on the couch eating your popcorn. This thang has library which is all about purrrrrrformance. It is actually built for industries where processing speed actually matters. So if you’re crunching numbers in say finance or you want to impress your buddies, OpenNN won’t let you down.
7. PyTorch
If you are starting out with Python and Machine Learning, you get to work with PyTorch. It is the popular kid in the deep learning world. Everyone loves it. Facebook made it and researchers can’t get enough. Because it’s simple. It’s flexible. And it doesn’t make you want to throw your laptop out the window. Dynamic computational graphs for instance…Yes please.
8. Rasa Open Source
Rasa lets you build chatbots that understand what people are saying. No more “I don’t understand that” responses every time you ask something even slightly unusual. With Rasa, you can create bots that actually help your customers instead of driving them to scream at their screens. You know.. the standard out-of-the-box-kinda rules-based bots. And it’s open source. Build. Customize. Repeat.
9. TensorFlow
Like PyTorch, you set your first babysteps into the world of AI with TensorFlow. It is Google’s gift to mankind. Or maybe it’s their way of making sure everyone depends on them just a little more. Either way, it’s a powerful tool for machine learning. You can build just about any AI model with TensorFlow. From pet detectors to medical diagnostics. And with all those high-level and low-level options, it’s another Swiss knife of AI. If you can dream it, TensorFlow can handle it.
10. Tesseract OCR
Like the name !! With Tesseract you are having a superpower. It reads text in images. Scans documents. Turns blurry pictures into actual words. So, if you need to turn a stack of PDFs into text, turn to Tesseract. It’s free. It’s powerful. And it doesn’t care if you use it to digitize your **** collection.
And now, finally the answer to the question:
Why open source could be AI’s secret weapon
Open-source AI has a few things that corporate AI just can’t replicate.
The best thing is that there’s flexibility. If you want to adapt a model for a specific need, say a specific branche, you can go in with open source and make those changes. Good luck trying that with a locked-down corporate model. Then there’s transparency. Open-source projects are, well, open. Anyone can examine the code, suggest improvements, or point out flaws. Because more and more systems will have some form of AI in it, from hiring to healthcare, transparency is a necessity.
And of course, there’s the sheer speed of innovation. Well, to be fair to closed source, with the money that the big seven poor into AI development, even the open source community as a whole cannot compete, but that aside, open source is the fast lane of AI development. Think about it: instead of waiting for a single company’s R&D cycle, the community steps in and accelerates things overnight. Genmo’s Mochi-1, for example, went from needing an expensive setup to running on consumer-grade hardware in a day, thanks to the crowd’s efforts. That’s the open-source difference.
Signing off – Marco
Well, that’s a wrap for today. Tomorrow, I’ll have a fresh episode of TechTonic Shifts for you. If you enjoy my writing and want to support my work, feel free to buy me a coffee ♨️
Think a friend would enjoy this too? Share the newsletter and let them join the conversation. LinkedIn appreciates your likes by making my articles available to more readers.
Leave a Reply