ChatGPT: Public vs. Private—where do we go from here?

To say ChatGPT has blown up in the last 8 months would be an understatement. It has not only sped up the pace of digital transformation; it has revolutionized search. A space that was previously dominated by a long list of hyperlinks—now has the capability to interact with results and beyond.

While Generative AI has captured the attention of the masses, questions emerge on how this technology can be infused into the apps employees use at work.  And what are the ramifications and policies related to working with Generative AI?

As we contemplate this, we need to understand that there are two major avenues for Generative AI. One is Public Generative AI which works with the mass of public data—think Bing, Bard, and ChatGPT. The second, Private Generative AI is a very similar technology that can be deployed inside of a company’s current applications and works with the data your company owns or licenses.

The policies, benefits, and use cases are very different between these public and private applications.

Public ChatGPT:

Open AI’s ChatGPT is trained on vast amounts of publicly available text from the internet. They have been fine-tuned to generate creative responses, provide information, and engage in open-ended conversations. Public ChatGPT models excel at a wide range of tasks, from answering questions to providing recommendations and even generating human-like content.

What has captivated the public’s attention is the actual interactive experience they have with it. Instead of simply typing a keyword or asking a question users can now chat back and forth or give commands to complete a human-like task. For example, a query could look like this: “Please write a 2000-word essay on the origins of the Civil War” and then add in “Can you write this for a 5th grader?”. I have had my fair share of conversations with AI since the launch of ChatGPT, and this is something that until late last year—was purely in the realm of science fiction.

Despite being a publicly accessible tool, there exists a valid justification for enterprise employees to utilize the capabilities of Open AI’s ChatGPT. Just like the need for employees to access Google, leveraging ChatGPT would extend their scope of utility in a manner that surpasses traditional search engines. In order to provide employees with the chance to leverage the capabilities of public ChatGPT, companies need to implement policies and standardized procedures that safeguard the interests of both the organization and the users involved. This is crucial due to the uncertainty surrounding the accuracy and hallucination of data, as well as the ownership of copyrights.

Generative AI in the Enterprise (Private):

Businesses are eager to reap the advantages of this incredible technology. They have already embraced its use as consumers, so why not enable users to harness its potential in a business context?

By employing Private Generative AI, businesses can efficiently utilize this technology within their day-to-day business applications. This approach provides a higher level of control concerning contextual understanding and data privacy. It offers users the opportunity to enhance their search capabilities exclusively within their organizational data, thereby empowering them to derive valuable insights while maintaining the confidentiality and security of their information.

For instance, consider a market researcher seeking to inquire, “What percentage of our sales in 2020 comprised Generation Z?” With Private Generative AI, the researcher would receive answers solely based on data available within the company, ensuring confidentiality and limiting the scope to internal information.

Private Generative AI provides three essential attributes: accuracy, safeguarding copyright for the generated content, and the exclusion of data sharing or training of extensive language models. This ensures that the generated output is highly accurate, addresses concerns related to copyright ownership, and upholds data privacy by refraining from sharing or training the models with external data sources. These attributes collectively contribute to a more controlled and secure AI environment for businesses.

Conclusion:

As companies formulate AI policies to govern the utilization of this transformative technology, it becomes crucial to acknowledge the inherent disparities between public and private sources and applications. Embracing a one-size-fits-all to generative AI is not the answer. Instead, a nuanced approach that recognizes the distinct characteristics and challenges of public and private implementations of these tools is essential. By adopting tailored policies that align with the specific needs of their organization, businesses can navigate the intricacies of AI deployment, fostering responsible and effective utilization while safeguarding privacy and maximizing the benefits derived from both public and private AI sources.