Dependent or intelligent AI?

For AI to become an effective work support tool, it is necessary to build the habit of "learning with AI" rather than "learning thanks to AI".

On August 7, Google introduced Guided Learning on the Gemini platform - an "AI tutor" that can tailor lessons to each person, analyze each step, explain with images, diagrams, videos and review with multiple choice questions, aiming at "active learning", improving the situation where many AI chatbots were complained about "returning homework" too quickly, causing self-study skills to decline.

Many useful features

With Guided Learning, Google upgrades the entire Gemini to support deeper learning, including auto-inserting images, YouTube videos, creating flashcards, and creating courseware based on documents or test results. The AI Pro ecosystem comes with NotebookLM (document summarization, analysis), Veo 3 (video processing), Deep Research (in-depth querying)…

OpenAI has also just launched ChatGPT-5, a version that CEO Sam Altman himself likened to a PhD expert. He said that GPT-5 processes information much more accurately, reducing errors by 45% compared to GPT-4o and 80% compared to the o3 model. In particular, GPT-5 has programming "superpowers", taking less than 5 minutes to create a completely new large language model. ChatGPT not only has Study Mode but also possesses a series of other features such as Canvas Mode (visual editing), Advanced Data Analysis (advanced data processing), Sora integration to create AI videos, and o3 and o4-mini models for research. In addition, Microsoft Copilot has entered Word, Excel, PowerPoint; Anthropic Claude is strong in reading and summarizing long documents; Perplexity AI stands out in precise searching and citing sources.

Not to be outdone, billionaire Elon Musk also launched Grok 4, which he calls "the smartest AI in the world ." Mr. Musk asserted that if it took the SAT - a standardized test used in college admissions - it would get a perfect score every time, even surpassing most graduate students in every field.

Ms. Thu Trang, an office worker in Ho Chi Minh City, experienced ChatGPT's Study Mode to learn more sales data analysis skills. AI guided her step by step, asking her questions so she could find the solution herself, just like studying with a mentor.

Along with that, Mr. Minh Quan (living in Ho Chi Minh City) used Guided Learning of Google Gemini to review his graphic design knowledge. "AI explains very intuitively with images and videos, helping me remember longer. However, when downloading long documents, AI sometimes cuts the content into unreasonable parts, so I still have to rearrange it myself to get the right answer" - Mr. Quan said.

AI is getting smarter but needs to be exploited properly to avoid dependence

Avoid dependence

Using AI tools is convenient, but many users also admit that overusing AI sometimes makes them "lazier". Hoang Hai, a fourth-year student at a university in Ho Chi Minh City, said: "Sometimes I "ask" ChatGPT to help me with my homework. However, if I use it for a long time without controlling it, it is easy to become dependent and gradually lose the ability to analyze for myself."

A new study from the Massachusetts Institute of Technology Media Lab (MIT) warns that using AI can reduce thinking activity. 54 volunteers were divided into 3 groups: using ChatGPT, searching Google, and writing themselves. The results showed that the group using ChatGPT had the lowest level of brain activity, was less effective in both neurological, linguistic, and behavioral aspects, and gradually became dependent, even copying the content processed by AI.

Mr. Nguyen Van Thuc, CEO of Keystone Technology and Training Joint Stock Company, commented that AI is developing at an unprecedented speed, opening up many opportunities in learning and working. The biggest problem is not in the technology but in the way people use it, when many people are gradually becoming dependent and ignoring their own thinking process. According to Mr. Thuc, for AI to become an effective support tool, it is necessary to build the habit of "learning with AI" rather than "learning thanks to AI". That is, users should let AI suggest, analyze and orient but still find the final answer themselves. AI platforms should also design mechanisms to encourage proactive thinking, such as asking reverse questions, asking users to explain or compare answers, instead of just providing complete results. "AI can be the most patient tutor, but if it is allowed to "do homework for" learners, they will lose analytical and creative skills - factors that machines cannot completely replace" - Mr. Thuc emphasized.

According to Mr. Phan Thanh Tung, representative of an AI technology solution provider in Ho Chi Minh City, AI is increasingly becoming an indispensable part of the learning and working process. However, if lacking the skills to filter and verify information, users can easily receive it passively, leading to misunderstanding or mechanical application. In addition to providing guidance on how to use tools, schools and training organizations need to integrate critical thinking and technology management skills into the curriculum. "Learners need to know when to ask for AI support and when to analyze on their own. They must know how to use AI intelligently to promote thinking ability, not taking shortcuts and skipping the process of brain training," Mr. Tung noted.

Data Leak

According to The Telegraph, more than 500 private chats with ChatGPT containing sensitive information such as internal discussions, fraud, cyber attack plans and questions from doctors and lawyers appeared in Google search results. Researcher Henk van Ess said that about 110,000 other conversations were stored on the Wayback Machine - the Internet Archive's web archiving tool. Experts warned that this leak could be exploited by cybercriminals; users should not share confidential data with AI and asked developers to strengthen security to prevent similar incidents.

Source: https://nld.com.vn/le-thuoc-hay-dung-ai-thong-minh-196250816202226394.htm