Justin Sung — How to Learn FASTER using AI (without damaging your brain)
Source: YouTube Channel: Justin Sung (2130000 subs) Duration: 40:47 Views: 76711 · Likes: 2968 Video: Watch on YouTube
Join my Learning Drops newsletter (free): https://go.icanstudy.com/newsletter-gptlearnfasterwithoutdamage
In this video, I will share my findings from thousands of tests and student conversations on how to use AI for learning in 2026 while avoiding key cognitive risks.
Take my Learning Diagnostic Quiz (free): https://go.icanstudy.com/diagnosticgptlearnfasterwithoutdamage
The AI Learning Paradox: Results from a survey of 923 learners (article link): https://www.linkedin.com/pulse/ai-learning-paradox-results-from-survey-923-learners-dr-justin-sung-mb0oc/?trackingId=DqThkJVASC%2Bl3LbO3bDZ2A%3D%3D
= Guided Training Program = I’ve distilled my 13 years of experience as a learning coach into a step-by-step learning skills program.
If you want to be able to master new knowledge and skills in half the time, check out: https://go.icanstudy.com/program-gptlearnfasterwithoutdamage
= About Dr Justin Sung = Dr. Justin Sung is a world-renowned expert in self-regulated learning, a certified teacher, a research author, and a former medical doctor. He has guest lectured on learning skills at Monash University for Master’s and PhD students in Education and Medicine. Over the past decade, he has empowered tens of thousands of learners worldwide to dramatically improve their academic performance, learning efficiency, and motivation.
Timestamps: 00:00 - Introduction: AI and Learning - Benefits and Risks 1:22 - Structuring the Video: Issues, Implications, and Solutions 1:46 - Issue 1: Information Accuracy and Hallucination in LLMs 2:01 - Survey Findings on AI Use in Learning 3:00 - Understanding LLM Limitations: Probability vs. Truth 4:32 - The Illusion of Accuracy: Fluency vs. Truth 7:17 - Solution to Information Accuracy: Risk vs. Complexity in LLM Usage 10:48 - Where LLMs Are Most Useful (Low Complexity) 11:05 - The Cost of Misusing AI for Complex Learning 13:42 - Good News: Most Learning Stays in Low Complexity 14:54 - Issue 2: Over-reliance on AI 15:55 - AI Doesn’t
Key Insights
A few months ago, I made a video saying that using Chat GPT is slowly destroying your brain. I said there’s a risk of it making you lazier, reducing your problem-solving ability, and damaging your memory and depth of understanding, all while making it feel like it’s actually helping you. And this was, as it turns out, quite controversial at the time when I uploaded it. But over the last few months, as more and more research is starting to come out around how AI affects learning, we’re starting to see that AI is not the savior to all of the learning problems that we thought it might be. Having said that, AI is revolutionizing learning. That’s a fact. At this point, there is no going back. And so, for me, as a learning coach, it has been a huge focus over the past year to really understand what is the best way to use AI for learning. I’ve been using AI for my own learning. I’ve been testing different models and different versions. Literally running thousands of tests on this. I’ve been talking to with students and professionals on how they use AI for learning, what’s working for them, what isn’t. And in this video, I want to share with you my findings and insights so far. This is basically my current status update on the best way I think that you can use AI for learning right now. Getting all the benefits of AI and mitigating the key risks. So, I’m going to structure this video is I’m going to go through the key issues, the major problems with AI that either I’ve identified or I’ve gotten them from my data, my conversations, and my surveys. And then I’m going to say what the implication of that is. Why you actually need to care about that. And then what you can do about that. Either my recommendations on how to use AI to mitigate that risk or whether you should just avoid it. So, to start it off, we’re going to begin with the biggest issue, uh which is concerns around information accuracy. So, that’s issue number one. Now, before I jump into the actual point of information accuracy, uh for context, in order to explore this topic a little bit more, uh over the past 4 5 months, I’ve been having dozens of conversations with my students, talking about the way that they use AI and the problems that they’re facing. I also ran a survey on uh my YouTube and my LinkedIn, collecting information from people that are both using AI or not using AI and getting their perspectives. I had 923 responses, uh which if I actually had published that as a study, that would have been a pretty large study. Uh but the findings were very interesting, and I’ll share some of those key insights with you throughout the video. One of the key findings from that was that the number one biggest concern that people have around using AI for learning is information accuracy. For people using AI for learning, this was the thing that they were most worried about. For the people that are not using AI for learning, this was the biggest reason why they’re not. And it’s also, I think, one of the most interesting points. And that’s because issues with information accuracy and the problem where an LLM like Claude, Gemini, Chat GPT, DeepSeek, uh they will just tell you something as if it were true, but actually it’s completely made up. Uh this phenomenon, which is called hallucination, this is an issue with the technology itself. LLMs, large language models, use something called the transformer architecture. And the transformer architecture, I mean, even though it is kind of like this amazing new milestone in AI development, it’s fundamentally a probability-based word generator. It looks at your query, it looks at the training data, the massive amount of training data that it’s got access to, uh and then it will create a network of what it thinks that you’re looking for, and it will match which words, based on the training data, would be the highest probability to come next. And so, it doesn’t have any sense of truth. Not only does it have no sense of truth, it doesn’t even
Transcript continues…
Chapters
- 00:00 — Introduction: AI and Learning - Benefits and Risks
- 10:48 — Where LLMs Are Most Useful (Low Complexity)
- 11:05 — The Cost of Misusing AI for Complex Learning
- 13:42 — Good News: Most Learning Stays in Low Complexity
- 14:54 — Issue 2: Over-reliance on AI
- 15:55 — AI Doesn’t Solve Core Learning Issues
- 17:08 — The Deceptive Helpfulness of AI
- 19:34 — Professionals vs. Students in AI Use for Learning
- 22:20 — Non-Productive Over-reliance Explained
- 23:33 — The Problem with Unclear Learning Metrics
- 25:34 — Avoiding Non-Productive Over-reliance
- 26:05 — The Value of Human Brain vs. AI
- 27:06 — Understanding LLM Capabilities (Probability vs. Conceptual Understanding)
- 30:01 — Where Human Value Concentrates: Beyond Basic Application
- 30:56 — Human Thinking Processes: Bloom’s Taxonomy