Artificial intelligence (AI) is a set of tools or devices that allow computers to solve problems in ways that mimic or exceed human behavior by ‘learning’ from very large amounts of data. They were first called ‘perceptrons’ in the 1950’s and 1960’s due to the idea that they would perceive similar to the human mind.
Since then, the technology has grown to a level where we now use and interact with AI in our daily lives. Search engines such as Google use it to customize search results for our individual needs. Insurance companies use AI to determine credit scores and premiums based on data generated by millions of people with different profiles.
AI tools are basically software and so they fall under the two major categories of software development space, proprietary and open source. Proprietary software is any software that is copyrighted and bears limits against use, distribution and modification that are imposed by its publisher, vendor or developer. This means that one must pay a licensing fee to either use or modify the software.
This was the traditional model from the 1960’s until the 1990’s when GNU/Linux was released as a free to use and modify computer operating system. From there, the open source movement was born with the idea that a community can develop software together, share and modify it for free to the benefit of all. Therefore, open source as an ideology is a contrast to proprietary.
Open Source AI Balances Profit with Humanity
As companies take up AI to boost productivity and revenue, several concerns about privacy and use of personal data need to be addressed. This is pertinent in sectors such as in banking, insurance, health care, pharmaceutical and medical products, as well as in the public and social spaces where bias can lead to loss of access to services and even lives.
Communities of open source AI developers create an opportunity to validate algorithms and training data without being motivated by profit.
One of the most controversial uses of AI in recent years has been predictive policing. The ability of AI to recognize images with accuracy has grown from less than 70% average to over 95% average, making in better than the human eye.
However, issues of racial bias in police facial recognition systems used in the USA have been raised in the media (The Atlantic, 2016; The NY Times, 2018). These systems are not required to go through public scrutiny or testing. Open development in an area such as law enforcement has the ability to create public trust.
Industry Best Practices
Organizations and companies dedicated to unlocking the potential of open source AI have focused on publishing charters and industry best practices so that ethics play a central role in development.
For example, the OpenAI Charter reflects the company’s commitment to cooperation, distributed benefits and long-term safety of their technologies. Similarly, h3O.ai; an open source AI development company, prioritizes the “democratization of AI” so that licensing fees do not hamper the testing and progress of AI.
Making AI development open does not mean it cannot be profitable. The development of algorithms and training data validation are critical to good AI and making these stages open source is very valuable.
New technologies and tools are being released every few months, without tangible use cases in the beginning. Collaborating with the open source community to test and develop such systems in terms of training data management, methods of validation and algorithm tweaking benefits both sides.
Companies such as Amazon, Oracle and Cloudera have platforms that are built on open source but wrapped in proprietary tools. These hybrid technologies mean that companies remain very profitable while still using community trusted tools at their core.
Is your company leveraging the power of open source AI?
Let the experts at ASB Resources show you how to identify and take the available opportunities to improve your IT systems using open source AI solutions. Schedule a call with one of our experts today!