AWS has announced new tools and services that make it easier for almost anyone to build generative AI apps. And there’s now Auto-GPT, a generative AI tool that can create and perform subtasks with little to no human intervention. As generative AI becomes more pervasive and more powerful, its potential to put your enterprise’s critical data at risk grows as well. Here are some recommendations to help you and your enterprise understand and respond to these recent developments.
New Generative AI Tools For AWS
AWS, Amazon’s market-leading cloud computing platform, announced earlier this month several new tools designed to ease and speed the creation of generative AI applications.
- Amazon Bedrock is a new service that makes the so-called “foundation models” used to train AI deployments easier for application builders to use via a simple application programming interface (API).
- Amazon Trn1n is a new AWS instance optimized for large-scale generative AI applications, with higher throughput and lower latency than its predecessors and support for large-scale distributed inference.
- Amazon Inf2, another new AWS instance, is optimized for low-cost, high-performance generative AI inference, promising costs up to 70 percent lower than predecessors.
- Amazon CodeWhisperer uses machine learning (ML) to help developers write code more quickly and accurately by suggesting code completions, variable names, and other relevant information as they type. CodeWhisperer can also generate code in any of multiple programming languages in response to natural language descriptions.
Relevant Risks: Generative AI is frequently deployed by non-technical “citizen developers” with little to no experience with or awareness of enterprise cybersecurity or data protection protocols. In addition, many of those deployments happen with no review, approval, or even awareness by IT or security teams. These new AWS offerings seem poised to increase and accelerate these risks at many enterprises, despite efforts by AWS to deliver secure and well-managed offerings.
Auto-GPT: Autonomous Generative AI
On March 30, game developer Toran Bruce Richards announced Auto-GPT, an open-source application that uses OpenAI’s GPT text-generating models to perform “autonomous” tasks. In response to a natural language description of a goal, Auto-GPT will break that goal into logical subtasks, then execute those tasks with no additional human intervention required. “Those tasks” can range from answering complex questions to creating presentations and other content to writing code. Auto-GPT is also already being used to create new types of AI applications.
Relevant Risks: Both open-source and generative AI deployments present risks to the quality and security of critical enterprise data and personal information. Given that Auto-GPT is open-source generative AI that can write and run complete programs, it combines and expands those risks, making enterprise guardrails and clear use policies absolutely essential.
What You Need to Do Now
These developments and those almost guaranteed to follow underscore a critical requirement. Your enterprise must implement technologies and processes that eliminate “unknown unknown” technology deployments. To ensure your data is kept accurate, consistent, and secure, you must be able to know about everything that consumes, manipulates, or contributes to that data. That must include all generative AI and other low-code/no-code deployments, as well as all open-source tools.
Your enterprise must also be able to identify and challenge or refuse connections to critical data requested by any applications or services that threaten data quality or security, including but not limited to generative AI deployments. And given the increasing autonomy of generative AI technologies, you need to implement protections that are highly automated and easily and quickly updated as new features appear. You may even need to craft and select solutions that use some of the very technologies creating more urgent needs for faster, more flexible, and more comprehensive data protection.
How Incisive Can Help
Incisive Software is focused on helping organizations build a strong foundation for success based on accurate and trustworthy data, especially in the face of new and growing risks spawned by generative AI and other low-code/no-code technologies. Incisive offers Incisive Analytics Essentials, a solution that enables you to gain managerial control over generative AI and other low-code/no-code deployments while making them available to authorized users. The Concourse platform, the heart of the Incisive solution, provides consolidated, comprehensive abilities to know what you have, know what changes, and effectively manage, protect, and trust your business-critical data across your entire enterprise.
To learn more about Incisive Analytics Essentials or to arrange a demo or free trial, visit https://www.incisive.com, email [email protected], or call 408-660-3090.