Why we choose to work with open source technology
Open-sourcing AI technology seems counter-intuitive, as both the development and computing costs can be quite high for larger models. Despite this, closed sourced models such as GPT-3 are exceptions in the AI landscape. The staggering million dollar cost to train a GPT-3 neural network from scratch might have something to do with that. So, why on earth are other models and algorithms being open-sourced by the companies that develop them?
Why we are experiencing an open source AI community
Google, Meta (Facebook), Microsoft, OpenAI, IBM, Amazon, etc. have all open sourced major technology components in the past 10 years. Most go even one step further and are actively publishing their research in the (academic) community. While the answer to why they are doing this is not clear cut, a large part of their decisions are motivated by the fact this strategy allows for a fast growth of the AI community as a whole and helps to build global trust in the technology. Leading companies want their technology to be the foundation on which innovations of the future are built. And don’t forget… you can’t use AI algorithms without having the corresponding data, which is the real treasure that is being hoarded inside the company walls. Lastly, let us not ignore that combining algorithms and data requires training the model, which is done on (cloud) computing infrastructure.
The open-source AI landscape can as such not be viewed as a purely philanthropic concept, however it does allow for (smaller) companies and people worldwide to adopt and create technology that would otherwise be beyond their reach. This is where we come in.
The disruption created by open-sourcing TensorFlow
A few years ago, Google open-sourced the software library for TensorFlow. Though TensorFlow wasn’t the first open-source AI software out there, it is widely regarded as one of the most advanced AI libraries in the world. Thus Google’s move to make TensorFlow open-source marked an unprecedented step forward, which its competitors couldn’t resist but to follow.
Google could have maintained a competitive edge by keeping its code to itself. Still, they realized their software would be even more productive if unaffiliated programmers could toy with it, add, and share their developments. Fellow tech giants would share this sentiment. Open source is a concept that gained traction in the ’90s with the success of Linux; it works on the same underlying principle that spurred the creation of the internet: the idea that everyone should freely share information and be available to all who want access. When it comes to AI, open-source technology is all about high-speed innovation. If everyone kept the knowledge of the algorithm inside a closed system, it would stifle innovation. Placing fundamental building blocks in the hands of the open-source community has created a valuable feedback loop for all involved.
At ML2Grow, we believe that open source allows businesses to experiment with AI more easily and builds greater trust in the technology.
Benefits of open source (AI)
Transparency and trust in AI
Introducing AI in a business process is a difficult hurdle for many companies as in most cases it involves taking away part of the decision process away from operators, managers or other specialists and trusting ‘the machine’. Although the decisions of AI algorithms and especially complex neural networks can at times be difficult to grasp, it surely helps that the software code is open and can be viewed by everyone to build that much-needed trust.
Flexibility and no vendor lock-in
If your AI software is not flexible, you’re going to get left behind by the competition. Open-source enables technology agility, typically offering multiple ways to solve problems. Open-source helps keep your AI capabilities from getting blocked because a particular ability isn’t available from a vendor. Instead of waiting for the vendor to deliver that capability, you can create it yourself.
Speed and experimentation
A significant advantage of open-source is the ability to experiment fast and begin to deliver value right away. This allows you to get the best of both worlds: flexibility, agility, and the ability to get started quickly and inexpensively, with the ability to mature to a large scale, fully supported implementation. And you don’t have to go over proprietary licensing hurdles to get there.
Cost-effectiveness
Open-source is generally much more cost-effective than a proprietary solution. Not only are open source solutions typically much more inexpensive in an SME environment for equivalent or superior capability, but they also give SMEs the ability to start small and scale. Provided that AI projects are often budget-challenged, it just makes financial sense to explore open source solutions.
Solid security
The responsiveness of the open-source community and vendors relative to information security problems has been excellent.
ML2Grow and open source AI
ML2Grow is a specialist in creating tailor-made AI solutions in many verticals. To be able to offer the best solutions to our clients at every moment in time for a wide variety of problems, we rely on the power of the open source AI community. ML2Grow is a vendor-neutral player in the market that is innovating on top of the most powerful and recent open source technology.
Open source stimulates transparency and, therefore, open competition. This way, you can always enjoy the most high-performance software made by the knowledge of many. With AI models, one can only speak of these advantages if, together with the software, the associated data is also ‘open’ and ‘accessible’. After all, an AI algorithm is unusable if it cannot be trained and later fed with the relevant data. Therefore, it is not surprising that the call is becoming increasingly louder to open and accessible data. This is to break through monopolies and ecosystem dependencies that arise from the owners of (large amounts of) data and the development of innovative services by players who do not own them. A serious problem that is hampering innovation globally.
However, on a positive node, many solutions only require internal or publicly available data to be developed. Therefore, companies are able to combine their own data with open source AI technology to create powerful problem-solving systems.
Interested to know more? Contact us for some free advice on how open source AI can help your business.
