Today, discussions about AI innovation and How Generative AI are transforming established business processes seem to be happening everywhere. It's not surprising considering that they can do almost everything: generate text, create images, automate routine tasks and more – and that's just the tip of the iceberg. This naturally raises the question: Is there something genuinely special about all of this, or is it just more unjustified hype around Artificial Intelligence?

Generative AI, or just Gen AI, is a neural network trained on a vast amount of text data, enabling it to produce impressively realistic and informative responses to user prompts. The ability to interact with Artificial Intelligence that appears as smart as a person piques our interest and earns our admiration. However, there are several issues associated with the use of Gen AI. Privacy issues, data security and its apparent omniscience play a critical role in this discussion. Should we trust and entrust our private information to a system with such intelligence? Could there be problems with unauthorized access to that data?

Certainly, a bot is a great assistant that can streamline professional tasks (for instance, quickly editing text or brainstorming ideas). But we shouldn't forget about the facts that point to the imperfections and shortcomings of Gen AI. It cannot be trusted to write a Privacy Notice or create a post on a highly specialized topic, and blindly believing in its answers is cautioned against by its own developers.

It is too early to claim how reliable ChatGPT is. It is readily apparent that Gen AI is not infallible. It hasn't reached the stage where we can entirely rely on it, let alone trust it with sensitive information. That said, don’t postpone learning about these new systems even if they aren’t quite ready for prime time. You can still leverage these useful tools without blindly relying on them.

One of the serious issues associated with ChatGPT is what’s called a “hallucination" where the model can generate responses that are not based on reality. This problem arises in part due to the limited dataset on which the model was trained. Additionally, ChatGPT is trained on fictional stories, myths, legends, and social media messages, among other sources. As a result, the neural network can generate responses based on these sources, leading to factual errors or misinformation. To address this issue, developers are teaching the model fact-checking and verification, restricting the types of data it learns from, and developing methods to detect and filter hallucinatory responses.

It can be assumed that over time, the model will improve and that its generated responses will be more trustworthy. However, a deeper problem, from a more philosophical perspective, will persist: even if we can trust ChatGPT’s data and responses, that doesn’t mean we can trust the system itself. In practical terms, the lack of agency in AI raises the question of who is responsible for it or the work it generates: developers, users, or both.

Despite ChatGPT’s popularity, its issues with accuracy have been well-documented since its inception. OpenAI admits that the free version of the chatbot has “limited knowledge of world events after 2021,” and is prone to filling in replies with incorrect data if there is not enough information available on a subject.

However, OpenAI recently announced that ChatGPT Plus subscribers, who have access to the GPT-4 model, will be able to search the web for up-to-date information.

Subscribers will now be able to go beyond the confines of pre-loaded data and search the web for real-time information. Need the latest news, facts, or updates? You got it covered! What's even better is that you can pose multiple queries in a continuous conversational format, allowing you to delve deeper into any topic right within your browser. Conversations are automatically named and saved in the sidebar, making it easy for you to manage, rename, or delete them. Plus, you even have the ability to "hide" specific chats if you prefer a clutter-free interface. With ChatGPT Plus and its web search capability, you can explore and discover an entirely new level of interactive information exchange.

At the same time, the information it possesses may not be sufficient, which affects various areas. For example, the University of Minnesota Law School had ChatGPT take four exams alongside real students. The chatbot ended up with a C+ grade. According to Professor John Choi (Harvard Law School's The Practice, 2023), ChatGPT could cite laws and accurately describe cases in its essays, but it couldn't "spot issues" and provide deep reasoning or analysis. While Gen AI demonstrates proficiency in common statistical methods, it may fall short when confronted with advanced or highly specialized statistical techniques. This limitation raises questions about the adaptability of Gen AI to diverse and specialized domains within data science. Understanding the extent of its statistical acumen is crucial for accurately gauging its applicability across various industries.

As we witness the exponential growth of Big Data, Gen AI faces a formidable challenge in grappling with colossal datasets. Its ability to handle extremely large volumes of data efficiently and effectively may be put to the test. Moreover, the complexity posed by advanced statistical techniques might prove daunting for Gen AI, potentially limiting its utility in domains that heavily rely on sophisticated data analysis. As businesses strive to meet modern data processing demands across diverse domains within data science, it becomes imperative to thoroughly assess whether Gen AI's limitations align with the specific needs and complexities of these fields. This evaluation serves as a vital step in determining Gen AI's viability and potential for success in tackling the ever-expanding frontiers of data science. As we witness the exponential growth of Big Data, Gen AI faces a formidable challenge in grappling with colossal datasets.

While Gen AI shows great promise and generates significant interest, addressing privacy concerns, misinformation, and accuracy is crucial for its successful integration into various industries. Recognizing its limitations and aligning expectations with reality is essential to harness Gen AI's potential while mitigating potential risks. As AI development progresses, ethical standards and societal interests must guide its evolution to ensure a responsible and valuable contribution to our technological landscape. There are other ways to automate tasks in your business without risking the uncertainties of experimental Gen AI. Just like AI can condense text, DeepLook simplifies and visualizes data, making it easily accessible and digestible for wiser decision-making. It's a reliable, human-friendly alternative in the ever-evolving world of automation.

1. Professor Jonathan H. Choi. Generative AI in the Legal Profession. Harvard Law School's The Practice. https://clp.law.harvard.edu/knowledge-hub/magazine/issues/generative-ai-in-the-legal-profession/