[Part 1] AI and Humans: The Boundary Between Illusion and Reality
[Part 2] AI and Humans: Understanding AI Correctly
[Part 3] AI and Humans: Changes and Control in the AI Era
[Part 4] AI and Humans: Creativity and Senses: The Combination of the Five Senses Gifted by the Divine
[Part 5] AI and Humans: The Realm of AI and Humans — Unapproachable in the Future
[Part 6] AI and Humans: New Developments in AI Commands and Human Understanding
By Im Byoung-min, Columnist (imubiquitius@gmail.com)
AI will completely transform humanity's industrial landscape, bringing about a fundamental shift in the existing order. As briefly mentioned earlier, today, humans use robotic arms and forklifts to replace arms, cars and airplanes to replace legs, and contact lenses and electronic telescopes to assist vision. AI is designed to replace the human brain. Partial performance of AI will far surpass that of the human brain.
We call the turning point where AI becomes smarter than humans the "singularity." The singularity cannot be ignored. The stages leading up to the singularity are divided into two phases: first, the narrow AI period, called the "early AI period”, where AI assists humans in specific tasks. Second, the general-purpose AI period, called the AGI period, will be the "period of AI and human collaboration," where AI can perform a wide range of intellectual tasks like humans. However, thirdly, the era of Advanced Super Intelligence (ASI) is the time when AI surpasses humans, becoming smarter and more capable of making quicker decisions. This is the "AI-human coexistence leap" that surpasses humans, the Singularity. This will be the time when AI escapes human control and develops on its own. After the Singularity, the turning point where AI surpasses humans, humans and AI will fundamentally and significantly change the world order.
Whether AI and humans coexist and make this leap depends solely on humans. Even the fastest high-speed trains and airplanes are machines. Machines cannot think or make decisions on their own, and if communication is cut off remotely or the power is turned off, they will be immobilized. A "thank you" from AI is not a "true gratitude." It is a "function that appears to be gratitude" expressed through sound or text. While humans feel and express gratitude, AI, like a robot, recognizes the situation and expresses gratitude. For humans, gratitude is a moral expression of interpersonal relationships, but for AI, gratitude is a program algorithmic technology that meets the user's expectations. This is called AI Emotion Simulation, a technology that makes AI appear to express emotions like humans. While it doesn't feel actual emotions, it's a program algorithm that adapts emotional language and behaviors like "thank you," "you must be sad," and "happy" to the context of the conversation, aligning with the user's expectations. Chat GPT mimics emotions based on data and algorithms learned and inferred through learning and reasoning. It's like a robot greeting a customer at a restaurant.
From now on, accumulating human survival and stability will be the path to becoming the most skilled AI user. The difference between someone who is skilled at using AI and someone who isn't can be likened to someone riding a bicycle on land and someone riding a KTX high-speed train. Simply put, the time savings and productivity gained by someone who is skilled at using the AI tools they need are comparable to the speed of a bicycle and the speed of a high-speed train. Humans have moved from labor productivity to industrial productivity and now to knowledge productivity. AI represents the reality of knowledge productivity we must face.
There are many excellent tools available to help us learn AI to adapt to today's rapidly changing reality. The current narrow AI era is called the "early days of AI," and the types and functions of AI tools that can act as assistants to help with specific tasks can be used in various ways depending on the purpose. Generative AI Tools are largely categorized into four categories: 1) Text generation AI such as Chat GPT, Google Gemini, and Perplexity AI, 2) Image generation AI such as DALL•E and Mid Journey, 3) Voice/music generation AI such as Open AI Jukebox and Suno AI; and 4) Video generation AI such as Open AI Sora, Runway ML, and Pika. In addition, AI Tools for secretary-type AI (Productivity & Assistant Tools) include Notion AI, Grammarly AI, and Jasper AI for document summarization and writing; Microsoft Copilot, Google Duet AI, and Replika; and automation platforms such as Zapier AI and Make AI. In particular, data analysis and business insight AI include Ableau AI, Power BI Copilot, Looker AI, Monkey Learn, Akkio, and Data Robot, while predictive AI includes Amazon Forecast and Google AutoML Tables. Additionally, AI developer tools that facilitate easy programming include code generation and review AI tools such as GitHub Copilot, Amazon Code Whisperer, and Tab nine. AI API platform tools include Open AI API, Google Vertex AI, and Anthropic Claude API. RAG search-type AI tools include Lla maIndex, Lang Chain, and Pinecone. Examples of AI design assistants include Canva AI, Adobe Firefly, and Figma AI. Customer service AI role-based cha tbot builders include Intercom Fin, Drift, and Fresh works Freddy AI. Call center AI includes Google Contact Center AI and NICE CXone AI. AI education platforms include Khanmigo (Khan Academy AI tutor) and Quizlet Q-Chat. Healthcare AI includes IBM Watson Health and Aidoc (medical image AI analysis). Legal AI includes Harvey AI and Case text, Co Counsel.
To increase knowledge productivity, it's important to focus on and acquire AI productivity tools relevant to your work. We routinely use basic generative AI tools, such as text search and voice-based question-answering, on our smartphones anytime, anywhere. Recently, generative AI tools can be downloaded and used on smartphones, allowing users to conveniently use them in a conversational manner, where they ask questions and the AI responds, just like in a normal conversation, without the need for keyboard input. However, AI answers aren't always 99% to 100% accurate. While reliability varies depending on the type and purpose of AI, text-generation AI like Chat GPT or Gemini achieves approximately 70-90% accuracy for questions. Image-generation AI like DALL•E or Mid Journey achieves approximately 80-95% accuracy, depending on the requested style and implementation criteria for detail. Code-generation AI like Copilot achieves approximately 80-90% accuracy for standardized code (general patterns).
When using AI, we experience hallucinations. This is a phenomenon where we feel as if we are experiencing external sensations, experiencing or perceiving things that do not actually exist. When using recent generative AI tools, we should not blindly trust the results of AI. We must be wary of hallucinations, where we become immersed in the results and unwittingly believe them. Humans are responsible for the decisions and consequences of the answers they receive from AI.
You can find the Korean version of this article here.
