Generative AI has arrived!
Now what?
Generative AI is a hot phrase today. A branch of Artificial Intelligence (AI), it can learn from and mimic large amounts of data to create everything from music to art. The media touts the excitable aspects – creating digital images, AI composed music that sounds like the Beatles, solving math problems or even developing code. The technology has been thrust into the public domain by OpenAI’s ChatGPT, Google Bard and Microsoft Bing AI. Since ChatGPT was introduced in November 2022, 1.6B users are accessing it every month. There is ample reason for excitement as the possibilities are endless.
Generative AI has the biggest potential to reshape businesses as we know it today – in every way and shape imaginable. The power of this technology can be harnessed to impact every silo in a business:
Risk reduction – AI can be disruptive in the Financial Services industry because it will democratize access to critical data. Financial institutions normally had to wait weeks to get access to data that then had to be parsed and cleaned before management could evaluate it and act. Even smaller institutions will be able to run scenarios predicting market crashes or defaults on loans.
Customer engagement – Outreach to customers can be made so much more effective if the message hits a specific need. By analyzing behavior patterns (predicting), AI can provide tailored products and services to specific customers and completely change their experience.
Efficient operations – AI has the capability to generate reports targeting specific areas of business, it can take over rote tasks and open capacity for staff to address critical business needs. Responding to specific questions from customers can be handled by AI.
Innovation – Just imagine sitting in a meeting and brainstorming creative ideas. AI can introduce innovative ideas and even create a prototype for participants to evaluate.
Compliance – Compliance has always been daunting. Generative AI can take over data management, due diligence, and ensuring accuracy for financial institutions.
Above, are a few of the use cases for Generative AI. The innovation that will come from continued usage and its application in many domains has yet to be defined. The application of systems like ChatGPT will lead to exceptional opportunities and innovations we haven’t even dreamed of. Conversely, this poses a real threat to established companies and can roil industries. While the excitement around AI is understandable, there should be an equal measure of caution and care as this is an undefined universe with few rules, guidelines, and goal posts.
As Generative AI moves into the corporate world, the need for control and governance becomes immediate. In the near future, the concern for most organizations should be the inadvertent exposure of customer or sensitive data to the AI algorithm. Organizations must establish policies and procedures governing the use of Generative AI. Samsung engineers using ChatGPT to correct a problem with their source code, unwittingly uploaded confidential data and meeting notes related to their hardware. There have been other examples of staff feeding proprietary information into the algorithm and this only serves to confirm that guardrails must be in place before an organization sets out on the AI path.
While exposing customer or proprietary data is a huge risk, utilizing Generative AI presents other risks and potential pitfalls. Some of the major points are outlined below:
Lack of transparency – Models like ChatGPT are opaque in how they provide the answer you receive. It is critical to understand how the model works. The adage trust but verify doesn’t necessarily hold true here. You can hope to trust the output, but verifying how the answer was derived becomes very complicated.
Accuracy – These models can sometimes produce inaccurate or fabricated responses. The system after all digests vast amounts of data to present us with answers that may be “hallucinations”. Sometimes the data is incomplete, sometimes the analysis is flawed. For these reasons, there should always be an evaluation of that output … caveat emptor!
Bias – All AI systems are only as good as the data they have been trained on. This presents a scenario where, if the data is biased, then the output from AI will be too.
IP and copyright – Today there are no data governance and protection assurances regarding confidential information. All users of AI systems should assume that data or queries they enter into systems like ChatGPT will become public information.
Cybersecurity and fraud – Organizations have to guard against malicious actors using Generative AI systems for fraud attacks. In the case of AI where data is critical, hackers may manipulate the underlying models and create outcomes that are skewed.
The promise of machines that can perform work has been in our imagination for many decades. Today with Generative AI, we have the beginnings of that and a whole lot more. The path is undefined, but it has the potential to shape the future of humanity. Perhaps the harder stage is still in front of us because we have yet to define the policies around usage and the mechanisms for control. Our ability to put in place those policies will be critical. We will discuss the policy component in the next article.
To start your AI journey and to empower your organization with the correct governance mechanisms, please contact us at 414.232.3622 or visit us at www.ProcessArc.com.