The Challenge of Democratizing AI Development

May 31, 2024

colorful digital art of people and a brain with gears
In a society flooded with AI technologies, many of them controlled by multi-billion dollar companies, calls for democratization are growing louder. But the widely varying meanings of "democratize AI" can make this misleading. Does it mean AI use/access, development, governance, understanding, profits, or something else? For example, in 2023, OpenAI began working on "systems that incorporate public inputs to steer powerful AI models." Back in 2016, Microsoft published their intent to make AI "accessible for all." Stable AI, the creators of the image generator Stable Diffusion, have consistently emphasized "democratization" and keeping their tech open source. Hugging Face similarly focuses on the "open source" meaning. Democratization is a popular word, but we can very different ideas about what it means, and which meanings matter most.
Often, companies mean the use of AI: making a certain AI readily usable for the widest possible audience with a broad demographic range. When AI has an impact on productivity, such as with ChatGPT, accessibility is important for equality; still, the technology remains a product for most users to "democratically" consume. Democratic development is far more difficult to achieve, but has more power to truly shift the status quo. With democratic development, users are not passive consumers but active participants in the development of the AI itself.

Why is this important? It's been clear to the average citizen for at least a few years that AI mastery is the key to the future. In an already grossly unequal world, where income disparities only seem to be growing, those without the ability to effectively use AI are hopelessly lost, while those who can actually develop useful AI are guaranteed success. The scurry to develop AI and infuse it into every aspect of life is the gold rush of the 20's, and most people aren't exactly striking it rich.

There are many open source foundational AI development technologies, such as TensorFlow and PyTorch, but these require programming knowledge, reducing their access to those with special skills. Some recent technologies better capture the spirit of democratizing development. Google's Teachable Machine allows users to provide data sets to train AI to correctly categorize items. This may be more ideal for learning the basics of AI development and pattern recognition than for serious development, but it's still a step in the right direction. This article by Forbes lists a variety of codeless AI development technologies, which are largely split between pattern recognition and data processing. Amazon Sagemaker stands out as a powerful tool to "prepare, build, train, and deploy high-quality machine learning models quickly." However, this service has a limited free tier, and is targeted towards "developers and data scientists." Simpler tools like SageMaker Canvas don't require special knowledge, but are restricted to basic functionalities and prebuilt solutions.
Free or low-cost options for robust codeless AI development are limited, but there is a clear interest to develop in that direction, such that even monopolies like Google and Amazon are making moves to democratize AI development. One may be skeptical, however, if such massive companies are going to provide products that put AI development power wholly in user hands. In December 2023, MIT Technology Review posted an article entitled, "Make no mistake—AI is owned by Big Tech." The article highlights many concerns concomitant with this monopolization, leading us to yet another meaning of AI democratization: shifting power from Big Tech to startups and smaller companies in order to make the field more competitive and diverse.
Happily, there are still smaller companies and research institutions active in democratic AI development. One such example is Obviously AI, who proudly state that their company began in "an AirBNB basement." You can explore their predictive model use cases here. Causaly provides scientific researchers a solution for navigating and analyzing vast bodies of scientific literature. Invideo AI enables users to create videos with text prompts, but, like many, is reliant on OpenAI's technology. The complexity and potential scope of creating AI development tech that enables users themselves to build truly anything they can imagine makes it difficult for smaller companies to compete with Big Tech.
Difficult -- but possible. Creating democratic AI that users can develop from the ground up to create a vast array of behaviors does not have to require massive infrastructure or an outsized wallet; it just needs an intelligent approach. Neuraville's solution is FEAGI (Framework for Evolutionary Artifical General Intelligence), a brain-inspired technology that allows users to codelessly develop brains using digital genomes and neural circuitry. Because of its similarity to the human brain in digital structure, FEAGI is more like a conduit for AI development than a rigid structure. Users can develop artificial brains from genome to maturity, unlimited by set features or presets. We have provided both open source and codeless browser-based versions of this technology, and encourage users to explore and give suggestions as we further develop FEAGI's vast potential. (You can learn much more about how FEAGI works in our blog post here.)

Democratizing AI in key areas is essential to regaining and ensuring equality on a global scale. If you are a developer, we encourage you to explore existing solutions and get involved in open source development; if you are not, we encourage you to seek out burgeoning and smaller-scale AIs of all varieties rather than limiting your experience to giants of the industry like ChatGPT, and to explore tools that allow you to develop AI yourself. The future belongs to all of us--don't hesitate to make it known!

© 2024 Neuraville Inc.

Discord
ServicesCareersContactNewsBlogPrivacy & Credits