From Google to ChatGPT — How AI is Shaping Our Web Experience
Remember Your First Google Search?
Do you remember the first time you used the Google search engine? Before then, web search was often a frustrating experience using platforms like Alta Vista or Ask Jeeves, delivering mixed-quality results. When Google launched in 1997, it revolutionized the search experience. The clean, ad-free interface was simple, and the results were surprisingly relevant. It was a groundbreaking moment in web user experience.
Google’s founders, Sergey Brin and Lawrence Page, detailed their creation in the paper The Anatomy of a Large-Scale Hypertextual Web Search Engine. Google’s innovation lay in how it ranked pages based on the number of links pointing to them, harnessing collective human intelligence to gauge relevance. This was one of the early steps in shaping Web 2.0—the social web. Google’s search engine dominated the digital landscape until recently, when Artificial Intelligence (AI) emerged to reshape our online interactions.
ChatGPT: An Elaborate Autocomplete Application
Enter ChatGPT, the generative AI chatbot developed by OpenAI. Unlike traditional search engines, ChatGPT can engage in natural language conversations, whether about poetry, business, or debugging code. It is a reading and writing machine capable of tailoring responses to any length, format, style, or level of analysis.
Ethan Mollick, in his book Co-Intelligence: Living and Working with AI, traces the development of AI toward the seminal 2017 paper “Attention is All You Need” by Google researchers. This paper introduced the now-famous transformer architecture, which revolutionized deep learning by using attention mechanisms to process and understand language. Though initially created for machine translation, the paper laid the groundwork for large language models (LLMs) like ChatGPT, which perform tasks such as question answering, text generation, and more. This architecture has since become a cornerstone in the rise of generative AI.
ChatGPT indexes a vast amount of text from online sources, learning patterns in how language is structured. When asked a question, it generates a coherent response based on likely word sequences. In a sense, it functions as an advanced autocomplete application.
AI at Work: A Moment of Realization
AI’s practical use became clear to me during a team meeting. We were stuck in an hours-long discussion, struggling to produce a cost-benefit analysis for a new product. Pressed for time, I turned to ChatGPT for help, requesting a generic analysis of similar products. Within seconds, it provided a solid draft that we quickly tailored for our presentation. I was sold on the power of AI.
Mollick imagines how LLMs can supercharge our productivity. He describes a system in which every aspect of our work is monitored and controlled by AI. It would track activities, behaviors, outputs, and outcomes of workers and managers. It would set goals and targets, assign tasks and roles, evaluate performance and dispense rewards. “AI’s ability to act as a friendly adviser could sand down the edges of algorithmic control, covering the Skinner box in bright wrapping paper. But it would still be the algorithm in charge. If history is a precedent, this is a likely path for many companies.”
AI, please take my job before this dystopian nightmare becomes a reality. ChatGPT has become integrated in my work, and I am grateful for the productivity it gives me. I have no intention, however, to be subject to an algorithm, repurposing my liberated time for maximum efficiency. The benefit is the freedom from low-level cognitive work. I can finally escape the keyboard, go for a walk outdoors, and engage in creative and holistic thought, inventing new work for the AI to execute. Consider too the accessibility improvements, allowing everyone to participate in a workplace that does not require years of training to perform mundane information tasks. That’s where I want to work.
The Role of AI in Writing and Judgment
I pointed ChatGPT to a collection of my online writing and blushed to hear its assessment of my writing style. “Your writing style is reflective, intellectual, and often philosophical, exploring deep questions about art, culture, and human experience. You blend personal insight with broader commentary, using precise language and a contemplative tone.” It went on with its praise. No doubt an AI hallucination. As Mollick observes, “LLMs work by predicting the most likely words to follow the prompt you gave it based on the statistical patterns in its training data. It does not care if the words are true, meaningful, or original. It just wants to produce a coherent and plausible text that makes you happy.”
Hallucination is often cited as a flaw in AI systems. However, misinformation has always plagued traditional information systems, summed up by the phrase “garbage in, garbage out.” Given the enormous size of its training set, the web, the accuracy of ChatGPT is impressive. More problematic is how its answers blend truth and falsehood, and its limited ability to give references. It answers like a human.
Mollick stresses that despite AI’s growing role in cognitive work, we must remain “the human in the loop,” exercising judgment over AI-driven outputs. While we can delegate tasks to AI, the ultimate responsibility for decisions stays with us.
I’m Feeling Lucky
The Google Search page traditionally had a second button, entitled, “I’m Feeling Lucky.” It invited our trust in it to return the right answer with one click. It was a gamble, simply returning the first result of a Google search, usually not a bad choice, but often insufficient. The button was rarely used and variously dropped from the interface.
The I’m Feeling Lucky button is what AI claims to deliver, a complete and satisfactory answer with one click. It could be revolutionary for user experience. Endless clicking, reading, analyzing, and summarizing are cognitive tasks currently performed by humans at desks with computers, keyboards and monitors. We could make all the technology and drudgery disappear into the background. We could just ask the computer the answer, like the Star Trek computer. Will it work? I’m feeling lucky.
Subscribe to my Essays on Substack
Published on September 29, 2024
Updated on September 29, 2024