A new report by the Government Accountability Office explores the risks and opportunities associated with generative AI. The technology behind generative Artificial Intelligence is examined, as well as the many possible uses of these systems.
The first of many GAO reports that will explore the world of generative AI is entitled Generative AI Technologies: Their Commercial Applications .
Future reports are expected to establish best practices and examine the social and environmental impacts of using generative AI. They will also look at federal development and adoption and the use of generative AI technology.
This first part provides an overview of generative Artificial Intelligence and how it differs. It also gives examples of its application in various industries such as software engineering and medicine.
The report begins with a question: Why is it important? The report goes on to explain that generative AI has been adopted by over 100 million users due to its enhanced capabilities and increased interest.
A level so high that it has sparked debate about the potential of this growth to revolutionize industries like healthcare and education versus the risks to national security, the environment, and the potential for spreading misinformation.
Key Takeaways from GAO Study:
-
Generative AI is different from other AI systems because it can create new content, requires large volumes of data for training and has complex models.
-
Generative AI systems employ several model architectures, or underlying structures. These systems, which are called neural networks, are loosely based on the human brain. They can recognize patterns in data.
-
Commercial developers have created AI models that generate text, code, images, and videos. They also offer products and services to enhance existing products, or refine models for customers. For many applications, their benefits and risks remain unclear.
-
The rapid development of generative AI was enabled by a combination of factors. The availability of large datasets and the refinement and enhancement of deep learning algorithms as well as computer capacity were key factors.
-
Training generative AI models often requires large amounts data. This is usually obtained from freely available information on the Internet, which may include copyrighted material.
-
Commercial developers use a process known as reinforcement learning based on human feedback to provide more meaningful and purpose-fit responses. Humans evaluate outputs and rank them, and the models then mimic the human preferences.
-
Training large AI models that generate data can take months and cost several hundred millions of dollars.
Click here to read the full report