ChatGPT: Where Are We Now and the Opportunities and Challenges Ahead

In this Insight we will address basic questions about the technology, the current flaws, the competitive and political landscape and the advancements on the horizon for various target audiences, including businesses, professionals and educational institutions. This article is intended as a summary only and not a deep dive into the details of the technology, nor is it a substitute for the creativity that will come from technology providers and end users. Despite legitimate concerns, when approached ethically, the use of AI bots has the potential to shift how we learn, live and work, as AI will inevitability co-exist with human contributions.

ChatGPT is an AI model that interacts in a conversational way (i.e. “Chat”). GPT is an acronym for Generative Pre-Trained Transformer. The program can research questions, write articles, computer code and poems (or at least poems that follow iambic pentameter) and lyrics (albeit without the intangible heart and soul of human authors and artists), solve scientific, tax and mathematical problems and so much more. The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises and reject inappropriate requests. In fact, quotes in italics throughout this Insight are ChatGPT-generated responses to queries about the AI model.

ChatGPT was developed by OpenAI, a for-profit research laboratory that has a nonprofit parent. Founders and donor/investors include Elon Musk, Silicon Valley’s Peter Theil and Microsoft. While “Big Tech,” such as Microsoft, Google and Amazon, with their computing infrastructure and deep pockets, will be the clear winners, it will be their licensees who develop the AI powered services for niche markets that will have the most impact on our lives.

The ChatGPT platform has been free to users since its launch in November, as the company gathers feedback on the platform, but a paid subscription version of ChatGPT called ChatGPT Professional already has a waiting list according to ZDnet.com. In February OpenAI announced ChatGPT Plus, a $20 per month paid subscription version. Other paid service tiers to monetize the substantial investment are on the horizon as new pricing packages are being updated regularly on the OpenAI website.

While ChatGPT has been most prominently in the headlines, it is certainly not the only player in the space. Other companies and organizations developing and improving language models include Google's BERT, BARD and LaMDA , Facebook's RoBERTa, and the Allen Institute for AI's GPT-3. Each language model will continue to innovate and improve its technology, while also expanding its capabilities and potential uses. Additionally, each will need to add and maintain strong partnerships and collaborations with other players in the space.

Current Weaknesses and Concerns

OpenAI’s Privacy Policy as of March 14, 2023

OpenAI’s Privacy Policy can be found here. Interestingly, however, when ChatGPT is asked to provide its Privacy Policy, its response is more abbreviated – a prime example that ChatGPT cannot be relied upon to be 100% accurate even with its own information! While most of the terms are standard across privacy policies, particularly in the technology space, a couple of provisions are worth noting.

Communication Information: If you communicate with us, we may collect your name, contact information, and the contents of any messages you send (“Communication Information”).

Personal Information: We Collect Through Our Social Media Pages: We have pages on social media sites like Instagram, Facebook, Medium, Twitter, YouTube and LinkedIn (“Social Media Pages”). When you interact with our Social Media Pages, we will collect Personal Information that you elect to provide to us, such as your contact details (“Social Information”). In addition, the companies that host our Social Media Pages may provide us with aggregate information and analytics regarding the use of our Social Media Pages.

Personal Information We Receive Automatically From Your Use of the Services: When you visit, use, and interact with the Services, we may receive certain information about your visit, use, or interactions (“Technical Information”), including the following:…

Usage Data: We may automatically collect information about your use of the Services, such as the types of content that you view or engage with, the features you use and the actions you take, ….

In certain circumstances we may share your Personal Information with third parties without further notice to you, unless required by the law, including without limitation in the situations below:

Vendors and Service Providers: … we may share Personal Information with vendors and service providers, including providers of hosting services, cloud services, and other information technology services providers, event management services, email communication software and email newsletter services, advertising and marketing services, and web analytics services.

*Emphasis added

OpenAI’s Terms of Use as of March 14, 2023

Again, many of the provisions in OpenAI’s Terms of Use are standard, but in summary, the user bears the ultimate responsibility for using ChatGPT, so proceed at your own risk.

Restrictions. You may not (i) use the Services in a way that infringes, misappropriates or violates any person’s rights; (ii) reverse assemble, reverse compile, decompile, translate or otherwise attempt to discover the source code or underlying components of models, algorithms, and systems of the Services (except to the extent such restrictions are contrary to applicable law); (iii) use output from the Services to develop models that compete with OpenAI; (iv) except as permitted through the API, use any automated or programmatic method to extract data or output from the Services, including scraping, web harvesting, or web data extraction; (v) represent that output from the Services was human-generated when it is not or otherwise violate our Usage Policies; (vii) buy, sell, or transfer API keys without our prior consent; or (viii) if you are using the API in connection with a website or application directed at children, send us any personal information of children under 13 or the applicable age of digital consent.

Content. As between the parties and to the extent permitted by applicable law, you own all Input. Subject to your compliance with these Terms, OpenAI hereby assigns to you all its right, title and interest in and to Output. This means you can use Content for any purpose, including commercial purposes such as sale or publication, if you comply with these Terms….. Other users may also ask similar questions and receive the same response. Responses that are requested by and generated for other users are not considered your Content. ...

Accuracy. Artificial intelligence and machine learning are rapidly evolving fields of study. We are constantly working to improve our Services to make them more accurate, reliable, safe and beneficial. Given the probabilistic nature of machine learning, use of our Services may in some situations result in incorrect Output that does not accurately reflect real people, places, or facts. You should evaluate the accuracy of any Output as appropriate for your use case, including by using human review of the Output.

Disclaimer. THE SERVICES ARE PROVIDED “AS IS.”

*Emphasis added.

OpenAI has been relatively transparent on its website about limitations of the current version of the program. The unknown is how many of these flaws are correctable in near- and long- term updates or whether are they inherent in AI bots.

As an AI language model, ChatGPT relies on patterns and data available at the time of its training to generate responses to user inputs. While it constantly learns from new data, its responses may contain errors or be outdated, especially if the information has changed since its last update in 2021. It is important to appreciate that ChatGPT's responses should at best be used as a starting point for research and not as a definitive source of information. Users should always verify the accuracy of the information provided by ChatGPT, provide citations for any quotes or paraphrases and seek additional sources to supplement their research. As the technology advances, the frequency of updates and improvements to AI language models will increase, reducing the potential for errors and outdated information.

Second, while the program is “trained to decline inappropriate requests” and “may occasionally produce harmful instructions or biased content,” adult supervision is always recommended.

Copyright Issues

The current version of ChatGPT denies that it has copyright liability for its use of third party material. When asked directly ChatGPT-3 (2023) responds:

… it's important to note that ChatGPT may generate responses that include copyrighted material, particularly if a user inputs copyrighted text as part of their question or conversation. In such cases, ChatGPT's responses would be considered derivative works and subject to copyright laws.

It is the responsibility of the user to ensure that any material they input into ChatGPT does not infringe on any copyright laws. It is also important for ChatGPT to be used ethically and responsibly, in accordance with OpenAI's terms of service, which prohibit using the model to create or distribute content that infringes on third-party intellectual property rights.

While it is impossible to know with certainty, much of what ChatGPT currently uses to create its database is believed to be from copyright free sources such as Wikipedia and U.S. government publications. That is likely to change as more data is added. Chat GPT’s Terms of Use and individual contracts with corporations and professional users will need to address copyright liability in more detail in future iterations.

As of this writing, there is no known litigation surrounding copyright infringement of the chat features of ChatGPT under the Copyright Act, 17 U.S.C. § 501, or the Digital Millennium Copyright Act (“DMCA”), 17 U.S.C. § 1201-1205. This is almost certain to change. The most relevant litigation to date is a class action suit alleging copyright infringement, filed on January 13, 2023, Sarah Anderson, et. al. v. Stability AI LTD, et. al., U.S. District Court, Northern District of California, Civil Action No. 3:23-cv-00201. While this case may provide some guidance on these issues, it Is critical to note that the subject matter involves the use of images, not text, and both the Copyright Office and the courts have focused to date on AI images. OpenAI is also the developer of DALLE-E, a learning model that generates digital images from natural language descriptions, so they will certainly be following this case closely.

Privacy

The Children's Online Privacy Protection Act (COPPA) is applicable, despite ChatGPT’s Terms of Use provision to the contrary, since ChatGPT is collecting data from users under the age of 13 - the user’s IP address, user interactions and the country where they are located. COPPA requirements also include obtaining verifiable parental consent before collecting personal information from children under 13. There is no indication that this is being done either by Chat GPT or its “operators.” (“Operators” is so far an ambiguously defined term in OpenAI’s literature; it may include AI partners, schools, businesses and even end users.) It is important for all operators of ChatGPT to ensure compliance with applicable data protection laws to protect user privacy and avoid legal liability.

Ethical Concerns: Biases, Plagiarism and Attribution

Biases

One concern which the developers are focused upon is to fine tune the diversity data because, to quote ChatGPT-3 (2023), it has the potential to “perpetuate or amplify biases that are present in the data it was trained on. For example, if the training data contains stereotypes or discriminatory language, the model may generate similar language when making predictions.

In a similar vein it can generate “fake news and manipulate public opinion” mirroring the kind of malicious activity already pervasive on the Internet. The pending U.S. Supreme Court decisions in two cases against Twitter and Google with respect to Section 230 of the Communications Decency Act (which protects platforms from liability for hosting their users’ posts) may tangentially give some indication on how they might rule in future cases regarding AI-generated content.

Plagiarism

Commentators differ as to the legal, ethical and practical implications of ChatGPT. One of the most cited criticisms of ChatGPT is the potential for plagiarism.

ChatGPT should not be cited for academic purposes i) because, as discussed above, it contains errors and ii) its responses are based on patterns, not facts and data. The output generated by ChatGPT may however still contain language and ideas that are similar to existing content.

While plagiarism checkers are nothing new, GPTZero is a program built by a Princeton University student that purportedly is able to detect AI-generated content. How effective this and other competing programs will be is yet to be determined.

Attribution

One of the more serious limitations of ChatGPT is that currently it is not programmed to generate footnotes and references (though an occasional reference can sometimes be found in its responses.) According to ChatGPT-3 (2023):

While ChatGPT is capable of generating responses that contain factual information and citations, it is not currently programmed to generate footnotes.

However, there are some natural language generation tools and techniques that can be used to generate text with annotations, including footnotes. These tools can be trained to generate text with contextual information, references, and citations, and could potentially be integrated into ChatGPT or other language models in the future.

That being said, it's important to note that the quality and accuracy of the annotations generated by such tools may vary, and the responsibility ultimately falls on the user to verify the accuracy and reliability of any information they receive, whether it includes footnotes or not.

Solutions on the Horizon: State Legislation and School Districts Banning ChatGPT

A Massachusetts state lawmaker has drafted proposed legislation to put guardrails on generative artificial intelligence models like ChatGPT, ironically combining multiple prompts in ChatGPT to draft the legislation. Two important provisions would require all text to be generated with a watermark to prevent plagiarism and to prohibit discrimination or bias based on federally protected characteristics.

In addition, the list of school districts banning the use of ChatGPT and potentially other AI models is growing every week. The following is a partial list of school district announcements restricting the use of ChatGPT on school computers and networks. The Los Angeles Unified School District blocked the site on December 12, 2022 days after ChatGPT’s launch, and the New York City Public Schools, the largest school district in the country banned it in early January. Since then Seattle, Fairfax, Virginia County, Baltimore, Maryland and Montgomery County, Alabama Public Schools have joined them in blocking ChatGPT out of concerns about plagiarism. This is only a temporary solution as students will have access to Chat bots outside of the school resources.

Opportunities to Incorporate AI Chatbots Into the Workplace and the Classroom

Experience has taught us that AI will improve in ways we can’t even imagine today, just as calculators, the Internet, Microsoft Office and cell phones have transformed education and our personal and professional lives. Therefore, in spite of the challenges, here are just a few examples of the possibilities ahead.

Educational Applications

Quickly migrating from the university applications, K-12 educators and their students are already experimenting with how to use ChatGPT creatively and responsibly. For example, ChatGPT can provide detailed feedback on student papers and personalized tutoring; create practice tests and quizzes as study tools and assessments, and help students generate ideas for papers by suggesting topics, outlining potential arguments and providing relevant research sources. It can also support students with learning differences, such as dyslexia, by providing feedback on written assignments and tutoring in an accessible and less intimidating way than traditional classroom settings. At a minimum, educating students and parents about the potential risks of AI and setting parameters for its use inside and outside the classroom is already becoming an additional responsibility for K-12 educators. These early experiences will prepare students to enter a changing workplace with the skills they need to work alongside AI.

Consumer Product Marketing and Other Corporate Applications

Coca-Cola has already announced that it will team with Bain & Company and OpenAI’s ChatGPT and DALLE-E platforms to craft personalized ad copy, images and messaging in multiple languages and to improve business operations and capabilities.

Coca-Cola is not alone. Since ChatGPT’s November release, surveys reveal it's been picked up by almost half of US companies, and 93% of these firms are looking to expand its use in upcoming months. According to tech.co, businesses are already using ChatGPT

  1. to respond to customer service inquiries
  2. to streamline first drafts of communications such as social media posts, blog posts, copyrighting, presentations and emails
  3. as a coding tool
  4. as a personal assistant for a variety of administrative tasks like data entry, email management and daily scheduling
  5. for keyword research/search engine optimization and
  6. to manage meetings using the AI-backed software to summarize meeting transcripts.
Legal, Accounting and Other Professions

ChatGPT has the potential to modify the way professionals approach their work in many of the same ways that corporations are making use of the technology. For example, Casetext has developed CoCounsel, an AI legal assistant, with applications specifically for the legal community. Applications include document review and citations to sources. ChatGPT Version 4 can read the tax code and apply it to individual filers’ unique situations. These features when fact checked can free professionals’ time to apply their skills to make critical strategic and judgment calls.

In Conclusion

We encourage you to try ChatGPT if you haven’t already done so. Despite ChatGPT’s weaknesses, you are certain to find creative ways to incorporate the technology into your personal and professional lives. As with all machine learning, ChatGPT can be enhanced with additional depth in certain subject areas. But what is clear is that AI and ChatGPT are not and should not be a total replacement for doing the hard work.

Lutzker & Lutzker will be following the developments in AI and chatbots and will provide updates as appropriate. “Trust, but verify,” an admonition Ronald Reagan used during his presidency decades ago, remains sound advice in the world of AI.