ChatGPT and Generative AI, the New Era of Text-to-X

12 min readJun 19, 2023

Image generated by DALL-E: “A 3D rendering of ChatGPT, artistic”

Executive Summary

Beyond the hype around ChatGPT and generative AI, lies a great potential beyond creating text, images, videos, and code: the potential to automate repetitive tasks that have little value and to transform databases and archives into knowledge. This article presents how to integrate ChatGPT and generative AI to take advantage of these two disruptive opportunities.

Introduction

Artificial intelligence (AI) has evolved significantly from its early days of hand-coded knowledge-based rules to modern deep learning neural networks, with generative AI being the latest news generating excitement. By analyzing patterns in existing data and learning how to reproduce them, generative AI can create brand new content in the form of data, text, image, code, and video, making it a transformative technology for primary activities like design and innovation, research and development, and marketing and sales[1] as well as support activities such as finance, legal, and human resources.

In this article, we’ll explore how generative AI has more to offer than content creation. Through real-world use cases, we’ll examine how it can automate low-value repetitive work in all the creative and productivity tasks. We will also see how it can unleash the potential of data and documents that today sit in silos across businesses in various industries, turning them into actionable information and knowledge. As more organizations adopt generative AI, other use cases will emerge, as seen with business digital transformation, data science, and machine learning. Concretely, every organization should consider integrating this cutting-edge technology into their strategy to stay relevant and thrive in the future. Failure to do so puts the organization at risk of missing opportunities to improve business operations or, worse, losing competitive advantage.

Generative AI in simple terms

Put simply, generative AI uses deep learning neural networks to identify patterns in large training datasets. It can produce original material, including text, code, and visual media, with similar characteristics as the training data. The networks are pre-trained on vast amounts of sources, such as books, articles, and websites. Once pre-trained, they can be fine-tuned for specific tasks, such as generating content in a particular style or topic. This involves training the model on a smaller, specific dataset to generate tailored content.

ChatGPT: OpenAI’s masterpiece, now iconic for generative AI

OpenAI has made a significant contribution to the development and deployment of generative AI, even though they were not the first company to do so. They achieved something unique that Google, Microsoft, and Meta did not: building a minimalistic, distraction-free user interface and promoting it globally. Since then, ChatGPT has become iconic, akin to 3M’s Post-It.

The go-to-market strategy of OpenAI is remarkable. Relying on millions of users to improve its product, the company orchestrated meticulous communication to create buzz around releases. When OpenAI makes an announcement, it overshadows all other communications, even those from tech giants like Google, Microsoft, and Meta. This stunning example of marketing strategy and mastery is reminiscent of the iPhone’s release in 2007.

OpenAI has demonstrated exceptional innovation and strategic thinking in the development and deployment of its products, setting new industry standards. While Microsoft sees ChatGPT as an extension of its Bing and Office businesses and Google envisions Bard as the next generation of its search engine, OpenAI’s ambition goes beyond ChatGPT. The company aims at building what appears to be the next-generation operating system that AI applications — not just generative AI — will rely on. This bold move positions OpenAI at the forefront of the AI industry, providing a unique opportunity to shape the future of AI. OpenAI’s contributions have not only benefited the business, but also the future of generative AI beyond content creation — as a technology to transform dormant data and documents into actionable insights and knowledge, and as the foundation of the next generation of software that we can interact with in natural language.

However, OpenAI is just the beginning and not the end of generative AI. Open source projects and new startups emerge daily, offering either complementary[2],[3],[4] or competing[5] solutions. OpenAI’s early lead provides no guarantee of long-term success, as demonstrated by the rise and fall of Netscape, Lycos, AltaVista, and Yahoo. Adding to this uncertainty are varying regulations in different countries, ranging from loose governance to full bans. Only time will tell if OpenAI will maintain its lead and shape the market, or if tech giants like Google, Microsoft, and Baidu will take over. But one thing is certain: whatever happens to OpenAI, its products will remain a good bet to build — in the words of Bill Gates — the digital nervous systems of tomorrow.

How generative AI tools like ChatGPT can change businesses

Generative AI will revolutionize the way businesses operate across different functions:

In design and innovation, it has the potential to create new, surprising, and valuable visual designs in less time. This appeals to multiple industries such as automotive, electronics, furniture, toys, fashion, luxury, and cosmetics[6].
In research and development, ChatGPT can help developers write code and documentation, generate synthetic data, and test scenarios. It can even identify anomalies and defects in the development process, significantly reducing the time and effort required to develop new products and services[7].
In marketing and sales, it can assist in creating marketing and sales materials that incorporate text, images, and videos. It also helps organizations generate product user guides. Furthermore, it aids in analyzing customer feedback and enhancing the efficiency of customer support[8].
In finance, legal, and human resources, generative AI can streamline the drafting and review process for financial statements, annual reports, and legal documents. It can automatically summarize large bodies of regulatory documents, answer questions from a large amount of legal documents, and even create interview questionnaires for candidate assessment, saving HR teams significant time and effort[9].

Overall, generative AI will bring a new level of efficiency and automation to businesses, enabling them to automate the low-value repetitive tasks. From design and innovation to research and development, to marketing and sales, generative AI has opened new possibilities. These possibilities are currently being experimented with by companies across various industries, including finance, legal, and human resources.

Morgan Stanley is one of the early big users of OpenAI’s GPT-4 model. Its wealth management content library includes over 100,000 pages of knowledge across various internal sites and formats. Previously, client advisors had to search for information when selling investments or responding to client requests. Complex requests had to go through multiple experts who authored the documents, leading to high levels of risk and time-consuming communication[10]. Now, an internal-facing chatbot uses OpenAI’s pre-trained GPT-3 and GPT-4 to provide advisors with instant access to executable knowledge from internal documents. Knowledge databases have become more than just a pile of documents and advisors now have access to the expertise of the most knowledgeable person in wealth management.

Bloomberg developed a large-scale model called BloombergGPT, which has been trained on financial data for natural language processing (NLP) tasks in the financial industry. BloombergGPT will be used for existing financial NLP tasks, such as sentiment analysis, named entity recognition, news classification, and question answering, and unlock new opportunities for data analysis[11].

GM is currently developing a virtual personal assistant that can provide access to information on how to use vehicle features commonly found in an owner’s manual. Additionally, they are exploring the potential of incorporating programming functions, such as the ability to set a garage door code[12]. By leveraging state-of-the-art technology, GM aims to enhance the overall user experience and make vehicle operation more intuitive and convenient.

In a similar vein, Coca-Cola has taken a leap into the realm of artificial intelligence. By experimenting with OpenAI’s ChatGPT and DALL-E, the company is delving into the world of personalized advertising. Through these AI models, Coca-Cola can craft ad copy, images, and messaging that resonate with individual consumers, forging deeper connections and delivering more impactful marketing campaigns.

CarMax, a renowned used car retailer, is leveraging the power of ChatGPT to improve its customer experience. With thousands of customer reviews to analyze, CarMax utilizes ChatGPT to summarize this vast amount of feedback. This summary serves as a valuable resource for potential customers, enabling them to quickly and efficiently choose a used car that best aligns with their preferences and requirements[13].

Mattel, the famous toy manufacturer, has embraced the creative potential of DALL-E. By utilizing this AI model, Mattel can generate innovative ideas for new toy cars. This collaboration with DALL-E enables Mattel to explore a wide range of possibilities and concepts, fostering the development of exciting and imaginative toy car designs that captivate the hearts and minds of children and collectors alike[14].

Challenges of ChatGPT and generative AI for businesses

As the previous use cases have demonstrated, generative AI holds immense potential and has received significant attention. However, it also brings forth a set of challenges and risks that demand thorough examination and analysis. While a comprehensive discussion would necessitate a full paper, I will outline a few key aspects that merit consideration.

Legal issues: Using third party-owned data to train models can raise legal issues related to intellectual property infringement. If the data used to train the model is protected by copyright, patent, or other intellectual property laws, using it without permission or proper licensing could result in legal action. Already, Microsoft, GitHub, and OpenAI are being sued for allegedly violating copyright law by reproducing open-source code using generative AI[15]. This lawsuit could have a huge impact on the wider world of artificial intelligence. This gives a competitive advantage to companies that own their raw material, such as Adobe, whose Firefly uses the company’s image stock.

Errors and biases: Chatbots contain errors and biases, leading to the dissemination of incorrect information. They have limitations in knowledge and do not access the most up-to-date information. Currently, ChatGPT does not integrate events that occurred after September 2021, potentially limiting its effectiveness in applications. For example, ChatGPT is unaware of events such as the death of Queen Elizabeth II, the World Cup, the end of the Covid pandemic, or the conflict in Ukraine. Retraining would be a solution, but it is expensive. That is precisely why OpenAI does not seem to have a plan to retrain ChatGPT anytime soon, risking depreciation of its value over time beyond language processing and translation.

Fake facts and news: Additionally, generative AI models rely on ingesting data found on the web, including a significant amount of fake news and intentionally misleading information. As a result, chatbots can propagate inaccurate or misleading information to their users, who may trust them as they do with search engines. This is particularly alarming given the potential for chatbots to reach large populations and contribute to the spread of fake news on a massive scale. The consequences of this issue can be damaging to individuals, governments, companies, and media, leading to further mistrust of real information. Bad information tends to travel faster than good information, making it even more critical for developers to implement strategies to prevent the propagation of fake news and ensure the accuracy and reliability of chatbots in providing information to their users.

Declining web quality: The escalating adoption of bots could lead to websites becoming increasingly tailored for interactions with bots, causing a surge in promotional content. This might erode trust in online information and compromise the overall quality of web content. A 2018 Pew Research Center study revealed a striking statistic: 66% of tweeted links to popular websites were shared by bots[16]. Another potential consequence of this trend is the decline in motivation, interest, and perceived value for individuals contributing to non-profit initiatives like Wikipedia and Stack Overflow.

In summary, generative AI has the potential to revolutionize how businesses operate and innovate. However, carefully considering the challenges and overcoming them is vital to ensure responsible and productive use. In the next section, we will see how training tools such as ChatGPT on curated data and documents, both internal and external, can turn these challenges into opportunities.

Overcoming generative AI challenges with curated data and documents

The availability of generative AI tools with user interfaces is currently dominated by self-service online tools such as Chat-GPT, DALL-E, Stable Diffusion, and Midjourney, or as add-ons to existing products such as Microsoft’s integration of ChatGPT into its Bing search engine and Office suite.

However, I believe the true potential of this technology lies in the curation of data and documents. As shown by the use cases of Morgan Stanley and Bloomberg, generative AI can be used to extract insights from large amounts of curated data, providing organizations with valuable insights into their business operations. By connecting chatbots to curated databases and knowledge bases, organizations can enable them to provide more accurate and relevant responses to user queries.

In addition, organizations can replace the time-consuming process of searching and reading through documents by using generative AI to prompt a home-grown ChatGPT who will act as the front-end of the organization’s digital nervous system.

The potential applications of generative AI in curated data and documents are virtually limitless. For example, generative AI can be used to analyze financial data and provide investment recommendations, or to integrate a new regulation into business operations. As organizations continue to invest in generative AI technology, the benefits will only continue to grow.

Unlocking the power of generative AI today

As the examples of use cases presented above show, there are mainly two ways of integrating generative AI in businesses:

Automating low-value repetitive tasks that eat most of knowledge workers’ time in design and innovation, research and development, marketing and sales, finance, legal, human resources, IT, and support functions.
Transforming curated data and documents from both internal and external sources into insights and knowledge that anyone authorized can get access to through natural language interaction.

Automating repetitive, low-value tasks is a relatively simple undertaking. All an organization needs to do is first list the repetitive tasks, then check which ones could be automated with tools like ChatGPT, DALL-E, or Codex, and finally redesign the workflow around new tools that will automate these tasks.

Retraining ChatGPT using curated internal and external data and documents to customize it to their specific needs and improve its performance on their particular use case is a more complex undertaking that will involve much more some steps. To navigate this complexity, project strategy and structuring become critical[17]. It is essential to determine which business function will benefit from generative AI and to evaluate whether the organization has enough data and documents that are worth the effort.

As with any data initiative, integrating generative AI to activate data and documents requires data governance[18]. But in the case of generative AI, data governance becomes more important as the training may be based on data and documents, inside and outside the walls of the organization, that must be regularly, checked and updated. This requires clear policies and procedures for data management, accuracy, security, and protection to not use, change, or divulge sensitive or confidential information as did Samsung employees who accidentally leaked trade secrets while asking ChatGPT to test their code[19]. This incident highlights the significant consequences that can arise from a lack of data governance in generative AI and emphasizes the need for organizations to implement effective policies and procedures to prevent similar incidents from occurring in the future.

Conclusion

The rise of generative AI presents an opportunity and a challenge for all companies across all industries. It presents a huge opportunity to turn data into actionable insights and documents into activable knowledge leading to improved customer experiences, increased operational efficiency, and better results. Companies in different industries are already experimenting with generative tools such as ChatGPT to build employee and client-facing chatbots. Those who will embark in retraining tools such as ChatGPT to specialize it with their data and documents will gain even further from this transformative technology. However, the road ahead isn’t straight but paved with challenges, calling for new strategies, stronger data and AI governance, and organizational changes.

About The Author

Hassan Lâasri is a consultant and interim executive, specializing in data and AI strategy, governance, and activation. He assists startups, growing ventures, and established firms with developing and implementing strategic initiatives. With over 15 years of experience, he helps clients leverage data, analytics, and AI for marketing, sales, CRM, and supply chain operations across a wide range of industries, including retail, luxury goods, cosmetics, finance, and insurance. His educational and research background includes a PhD, a post-doc, and two patents in AI.

Acknowledgement

This article draws from a wealth of reputable sources, including arXiV, MIT Technology Review, Wired, NYTimes, and WSJ. It also incorporates my firsthand experience with cutting-edge technologies such as ChatGPT, DALL-E, Midjourney, and Stable Diffusion. To further enhance the content, I have drawn from the practical knowledge gained through coding courses provided by DeepLearning.AI.

[1] https://www.weforum.org/whitepapers/creative-disruption-the-impact-of-emerging-technologies-on-the-creative-economy

[2] https://blog.langchain.dev/

[3] https://arxiv.org/pdf/2303.17580.pdf

[4] https://writings.stephenwolfram.com/2023/03/chatgpt-gets-its-wolfram-superpowers/

[5] https://www.wired.com/story/get-ready-to-meet-the-chatgpt-clones/)https://www.wired.com/story/get-ready-to-meet-the-chatgpt-clones/

[6] https://www.technologyreview.com/2022/12/16/1065005/generative-ai-revolution-art/

[7] https://openai.com/blog/openai-codex

[8] https://www.forbes.com/sites/bernardmarr/2023/01/17/how-will-chatgpt-affect-your-job-if-you-work-in-advertising-and-marketing/

[9] https://fortune.com/2023/01/25/future-of-finance-generative-ai-mit-researcher/

[10] https://openai.com/customer-stories/morgan-stanley

[11] https://arxiv.org/pdf/2303.17564.pdf

[12] https://www.cnbc.com/2023/03/10/gm-explores-using-chatgpt-in-vehicles.html