Table of Contents
Artificial Intelligence Translates the Language of Animals
Kim Kyung-jin, Attorney at Law
The Story of AI Learning to Listen to Whales, Dolphins, Birds, and Bees
Twelve chapters on how AI listens to dolphins, sperm whales, humpback whales, birds, and bees to find rules in their sounds, and what this technology means for its risks and for animal rights. Written in simple sentences a child can read, with verified sources in every section.
Table of Contents
AI Deciphers Ancient Scripts
Kim Kyung-jin, Attorney at Law
Ancient Records Revived by AI
In twelve chapters, this book explains how AI revives records once unreadable, from burned scrolls and wooden slips buried in mud to broken clay tablets. It covers virtual unrolling at Herculaneum, virtual collation of oracle-bone texts, reading Silla wooden tablets, computational analysis of undeciphered scripts, and multispectral archives, with verified references for each chapter.
Table of Contents
Artificial Intelligence for New Materials Design and Rocket Propulsion Engineering
Kim Kyung-jin, Attorney at Law
AI Potentials, Self-Driving Laboratories, and Physics-Informed Machine Learning (PIML)
Ten chapters on how artificial intelligence is changing new materials and rocket propulsion: atomic simulation, generative models, self-driving labs, high-temperature alloys, metal 3D printing, combustion, cooling design, and engine diagnosis and control. Written without equations, with verified sources in every chapter.
AI Library
The Age of Autonomous Scientific Discovery
Kim Kyung-jin, Attorney at Law
AI Scientists and Self-Driving Labs
This book follows how AI scientists and self-driving labs are changing the way science generates and verifies claims. It covers literature-based discovery, natural-language protocols translated into robot commands, multi-agent research systems, closed-loop laboratories, materials search, the verification gap, chains of evidence, research harnesses, journal ethics, and legal responsibility.
AI Library
A New Era of Life Sciences Opened by Artificial Intelligence
Structural Proteomics, Genomic Foundation Models, Autonomous Laboratories, and Global Governance
Kim Kyung-jin, Attorney at Law
This book is a research volume compiled with artificial intelligence. A human selected the materials and structured the work, while AI models drafted the sentences and cross-checked the facts.
AI Library
The Double Structure of Digital Sovereignty
Europe’s Departure from Palantir and the Chains of American Big Tech
Kim Kyung-jin, Attorney at Law
This is a record of 2026, when European intelligence agencies and defense ministries began removing analytics tools from America’s Palantir. It covers the replacement decisions made by France’s General Directorate for Internal Security (DGSI), Germany’s Federal Office for the Protection of the Constitution (BfV), and the Netherlands Ministry of Defense; the incident in which US export controls severed an ally’s ac…
New English Edition
Artificial Intelligence in Horticulture
Kim Kyung-jin, Attorney at Law
Across five chapters and ten sections, this book examines computer vision for crop diagnosis, harvesting robots and autonomous field systems, smart greenhouses and digital twins, precision irrigation and supply-chain quality control, high-throughput phenotyping, and predictive breeding.
New English Edition
Artificial Intelligence in Food Crop Agriculture
Kim Kyung-jin, Attorney at Law
Across six chapters and eighteen sections, the book examines digital agricultural infrastructure, remote sensing, crop diagnosis, yield forecasting, precision irrigation, genomics, molecular breeding, agricultural robotics, climate-smart agriculture, and global food security.
New English Edition
The Future of Forestry and Agroforestry
Kim Kyung-jin, Attorney at Law
Driven by Artificial Intelligence and Digital Innovation
Across five chapters and fifteen sections, the book follows satellites, drones, LiDAR, digital twins, forest-specific language models, wildfire and pest forecasting, forestry robotics, agroforestry, timber traceability, and forest carbon markets.
New English Edition
Smart Livestock Farming: AI Enters the Barn
Kim Kyung-jin, Attorney at Law
Sensors listen, cameras watch, and artificial intelligence helps farmers decide.
Across five chapters and fifteen sections, the book follows precision livestock farming from animal health and reproduction to robotic milking, virtual fencing, digital twins, methane reduction, welfare, and data ownership.
Table of Contents
Han Dong-hoon, Busan Buk-gu Gap: A Record of the 100 Days Before and After the Election (Mar. 26-Jul. 3, 2026)
Kim Kyung-jin
Table of Contents and 13 sections
From March 26 to July 3, 2026, this record follows the spring after expulsion, the Busan Buk-gu Gap by-election, victory as an independent, and the first bill submitted in the National Assembly.
Table of Contents
Artificial Intelligence and Medicine
Kim Kyung-jin, Attorney at Law
AI in clinical care, hospitals, education, and research
AI in medical imaging, risk prediction, treatment planning, hospital operations, education, and research, with patient safety, privacy, and accountability.
[AI Library] 5.1 Where Models Are Heading
Sam Altman Biography: Pioneer of the AI Revolution
Part 5: The Future and Roadmap of AI Technology
5.1 Where Models Are Heading
Kim Kyung-jin, Kim Kyung-ran
Algorithmic Innovation: The Potential for 10x to 100x Improvement
There's a point Sam Altman raises more often than any other in OpenAI's conference rooms.
"Bigger computers, more GPUs. Yes, those matter. But what truly changes the game is the algorithm."
He began pressing this message harder from late 2024, because the entire AI industry was running into a single wall: the limits of scaling.
Up through the jump from GPT-3 to GPT-4, the formula was straightforward. Gather more data, chain together tens of thousands of more powerful GPUs, and performance climbed steadily, the way pouring more fuel into a car engine makes it run faster.
By 2024, things had changed. Research labs across the industry started reporting the same problem: doubling the size of a model no longer delivered the gains it once had. This was what people began calling the slowdown of scaling laws.
Altman saw this coming earlier than most. So he began searching in a different direction: algorithmic breakthroughs.
"The potential for a 10x to 100x improvement in performance lies in algorithmic innovation."
It sounded like hype. But a look back at AI history showed it had already been proven more than once.
The Transformer architecture, which appeared in 2017, is the clearest example. That single new algorithm made it possible to achieve dozens of times better performance from the same computing resources. The Transformer was also the foundation on which ChatGPT was built.
Another example is Reinforcement Learning from Human Feedback, or RLHF. That one technique transformed GPT models from simple text generators into capable assistants that could hold intelligent conversations with people.
Altman wrote in an internal document:
"We are already approaching the physical limits of computing. The approach of simply scaling up size will hit a wall sooner or later. What breaks through that wall is an entirely new class of algorithms."
In December 2024, OpenAI released a new reasoning model called o1. It worked in a fundamentally different way from earlier GPT models. Rather than producing an answer immediately, it broke problems into multiple steps and worked through them slowly, genuinely "thinking."
The results were striking.
On International Mathematical Olympiad problems, o1 performed at the level of a human gold medalist. It solved problems that GPT-4 had failed to crack, and it did so using far fewer computing resources than it took to train GPT-4.
That was the power of an algorithmic breakthrough.
In a January 2025 interview, Sam Altman said:
"When superintelligence arrives, the pace of scientific discovery will be ten times faster. Progress that once took ten years will happen in one. And the following year will bring just as much again. Let that compound and the world will look completely different."
He also moved quickly to temper expectations. "The hype on Twitter is way too intense. Scale your expectations back by a factor of a hundred. We have not built AGI yet."
But the core message was clear: the next stage of AI progress would come not from a hardware race but from algorithmic innovation.
Inside OpenAI, new algorithmic research is in full swing.
Researchers are developing new reasoning patterns such as Chain-of-Thought and Tree-of-Thought. These techniques push AI beyond merely predicting the next word, toward something closer to genuine reasoning.
Efficient architectures like Mixture of Experts are also under active study. The approach embeds a set of smaller specialist models inside a large one, and when a question comes in, only the most relevant specialist is activated to answer. It's like calling only the right department into a meeting rather than pulling in the entire company.
Techniques like these can cut the computing resources needed to reach the same level of performance down to a tenth.
Jakub Pachocki, Altman's chief scientist, put it this way:
"Current models can handle tasks of roughly five hours in length. But that window will expand quickly. For a major scientific breakthrough, it would be worth pouring an entire data center's worth of computing power into a single problem."
The key word is efficiency.
The human brain runs on about 20 watts, barely enough to light a single bulb. On that budget, we think, create, and solve problems. GPT-4, by contrast, consumes enormous amounts of power to produce a single response.
Closing that gap is what algorithmic innovation is ultimately about.
Sam Altman painted the future this way:
"One day we'll be able to run a model a hundred times smarter than today's at a hundred times lower cost. That moment is where the real AI revolution begins."
He believes this breakthrough goes far beyond a technical achievement. Discovering cures for cancer, solving climate change, developing new energy sources. The hardest problems humanity faces could yield to algorithmic innovation.
In October 2025, Altman offered a more concrete timeline.
"By September 2026, we will have an AI research assistant that works at intern level. And by 2028, we will have a fully automated AI researcher."
This AI researcher can conduct research projects on its own, without any human instruction. It reads papers, forms hypotheses, designs experiments, and analyzes results. And the most important thing: this AI can research how to build a better AI.
This is the self-accelerating effect of algorithmic innovation.
Once a good algorithm emerges, that algorithm finds an even better one. From there, the pace of progress accelerates exponentially.
Of course, there's no guarantee that any of this will go smoothly.
Algorithmic innovation still faces many hard problems. Long-horizon reasoning, consistent planning, grasping complex causal relationships , these are capabilities AI hasn't yet mastered.
But Sam Altman is optimistic.
"We now know that 90% of a model's limitations are algorithmic. The moment we break through that wall, AI becomes an entirely different kind of thing."
He believes that moment will come sooner than most expect. "Within a few thousand days" is a phrase he uses often. It points, roughly, to around 2030.
An algorithmic breakthrough is not a mere technical upgrade. It's a redesign of the future of human-built intelligence. And Sam Altman stands at the door of that future, quietly but firmly reaching for the next step.
The Trinity of Algorithms, Data, and Computing
Sam Altman has a comparison he returns to often.
"An AI model is like cooking. You need a good recipe (algorithms), fresh ingredients (data), and a well-equipped kitchen (computing). Leave out any one of them and the dish falls apart."
These three elements are the pillars of AI progress. Every strategic decision at OpenAI centers on how to strengthen all three and develop them in balance.
The First Pillar: Algorithms
Algorithms are the 'brain architecture' of AI.
Sam Altman calls algorithms "the blueprint of intelligence." Use the same data and the same hardware, but swap the algorithm, and the results are worlds apart.
He felt this truth deeply while building the GPT series.
When GPT-3 arrived, the public was amazed. But inside OpenAI, the prevailing view was "we're not there yet." The model frequently gave wrong answers or lost track of context.
That's when RLHF , Reinforcement Learning from Human Feedback , entered the picture. That single algorithm turned GPT-3 into something entirely different. Same model, same data, but a change in how it learned suddenly made it capable of conversation that felt genuinely human.
That's when Altman became certain. "The future will be built not by larger models, but by smarter algorithms."
OpenAI's research direction shifted decidedly toward algorithmic innovation. New reasoning approaches, more efficient attention mechanisms, architectures with long-term memory, tree-based reasoning that mirrors human thought , all of it came out of algorithm research.
Then, in late 2024, the o1 model arrived. It solved problems by extending its 'thinking time.' Where earlier models rushed to produce an answer, o1 reasoned through each step slowly.
The result was far superior performance from the same amount of training data. That was the power of algorithmic innovation.
The Second Pillar: Data
Data is AI's 'experience.'
Sam Altman describes data as "the textbook through which a model learns the world." No matter how good the algorithm, training on poor-quality data produces a useless AI.
"Garbage in, garbage out." That old adage from computer science has never felt more urgent than in the age of AI.
When building GPT-3, OpenAI scraped vast amounts of text from the internet. Web pages, blogs, forums, books, news articles , everything went in.
There was a problem. The internet doesn't contain only good information. Fake news, bias-laden writing, profanity and hate speech , the model learned all of that too.
Starting with GPT-4, the focus shifted to data quality. Rather than collecting as much data as possible, OpenAI began selecting verified, high-quality sources.
Scientific papers, specialized books, reliable news sources, vetted educational materials , this kind of data lifted the model's performance significantly.
Altman told his team: "AI should resemble the world. But it should resemble the best of it, not the worst."
Another important shift was data diversity.
Text alone was not enough. From GPT-4 onward, the model trained on a wide range of data types: code, mathematical formulas, scientific data, medical protocols, legal documents.
The model's way of thinking changed. Where it once produced plausible-sounding text, it could now actually solve problems.
From late 2024, 'synthetic data' also grew in importance. This is data generated by AI itself. The reasoning process that the o1 model produces while solving math problems, for instance, becomes the training data for the next model.
This makes it possible to break through the ceiling of data that can be scraped from the internet. A virtuous cycle begins, where AI generates progressively better data on its own.
The third pillar: computing
Computing is AI's muscle.
No matter how good the algorithms and data are, none of it matters without computers capable of running them.
Sam Altman is deeply pragmatic about computing. He knows that AI progress demands enormous amounts of computing power. But he also knows that power is not infinite.
Training GPT-4 is estimated to have cost over $100 million, most of it spent on GPUs and electricity. That is a scale most startups simply cannot sustain.
So Altman moved on two fronts.
One was securing more computing infrastructure. He partnered with Microsoft to build massive data centers. In 2025, he announced the Stargate Project, a plan to invest $500 billion in AI infrastructure.
The other was using computing more efficiently. This is where algorithmic innovation re-enters the picture. A smarter algorithm can accomplish the same task with one-tenth the compute.
Altman often puts it this way: 'The price of intelligence will eventually converge with the price of energy.'
What does that mean?
Today, building an AI model involves many costs: algorithm development, data collection, engineer salaries, and more. But in the future, once algorithms are mature and data is abundant, the only remaining cost will be electricity.
That is why he pays intense attention to energy. Nuclear power, nuclear fusion, renewables. He believes that securing cheap, clean energy is the central challenge of AI progress.
The harmony of three pillars
What matters is that these three elements shape each other.
A better algorithm reduces the amount of data needed and the computing resources required. Conversely, more computing power allows more complex algorithms to be tested and larger volumes of data to be processed.
Sam Altman believes that balancing these three elements is his most important responsibility as CEO.
'Focus too much on algorithms alone and you won't get practical products. Focus too much on computing alone and efficiency drops. Focus too much on data alone and you run into legal and ethical problems.'
'Advancing all three simultaneously without losing balance. That is exactly what OpenAI does.'
And on the future, he offers this outlook:
'Within the next ten years, algorithms will become 100 times more efficient than they are today. Data will approach something close to infinite, thanks to synthetic data. Computing will grow far more powerful, driven by new chip technologies and cheaper energy.'
'When these three wheels start turning together, that is when the real magic happens.'
A Trillion-Token Context and the Personalized Lifelong AI
What is the ultimate form of AI that Sam Altman envisions?
It is not simply the smartest AI in the world. If anything, it is closer to the opposite: an AI that is built for you alone and knows you better than anyone else.
He captures this vision in a single sentence.
'A very small reasoning model with a trillion-token context.'
That sentence contains everything essential about the AI of the future.
What is a trillion tokens?
First, you need to know what a token is.
A token is the basic unit by which AI understands language. Think of it as roughly one word, or sometimes a smaller fragment. The phrase 'Hello there,' for example, consists of roughly two to three tokens.
So how much is a trillion tokens?
The complete Harry Potter series runs to about one million tokens. A trillion tokens is enough to read Harry Potter one million times over. Put another way, it is more than enough to hold everything a single person reads, writes, and says across an entire lifetime.
'Context' refers to AI's working memory. The ChatGPT we use today tends to forget what was said earlier in a conversation as the exchange grows longer, because its context is limited.
As of 2025, even the models with the largest context windows hold around one million tokens. That's an enormous amount, yet it's only one-thousandth of one trillion tokens.
An AI that remembers an entire lifetime
What could an AI with a one-trillion-token context actually do?
It could remember every record of your life from the moment you were born until right now.
Diaries you wrote as a child, homework assignments from school, conversations with friends, things you said in job interviews, emails sent at work, memories with family, health records, favorite music and films, hobbies, worries, and dreams. All of it, held in the AI's memory.
Sam Altman calls this "a second brain for the individual."
What becomes possible with an AI like that?
In the morning, the AI greets you: "Today is the day of your big presentation. Do you remember that similar presentation five years ago, when nerves got the better of you and things went sideways? Here's how you might prepare differently this time."
At a health check-up, the AI steps in: "Looking at your ten years of health records, your blood pressure tends to creep up a little around this time every year. The exercise routine you tried last year showed real results , it might be worth picking that up again."
When you're about to launch a new project, the AI offers a thought: "You jotted down a similar idea three years ago. You didn't act on it then, but the circumstances look different now. Combining those old notes with what you're thinking today could produce something solid."
This is what a truly personalized AI looks like.
Why does it have to be a 'small' model?
There's an important twist here.
Sam Altman says this AI needs to be "a very small model." What does that mean?
Today's models, like GPT-4, are enormous. They run only inside massive data centers. But the personalized AI Altman envisions has to run inside your smartphone.
Why?
First, privacy.
Sending your life's records over the internet to be processed on a distant server is risky. The data could be hacked, or the company could misuse your information.
But if all processing happens inside your own device? No one can touch your information. Privacy becomes absolute.
Second, speed.
Sending data back and forth over the internet takes time. When everything is processed right on your phone, you get answers instantly , as fast as a thought forming in your own head.
Third, cost.
Keeping massive servers running costs a fortune. If a small model runs on each person's own device, the price drops dramatically.
"But can a small model actually be smart?"
That's the right question. Here's how Altman answers it.
"A personalized model doesn't need to know everything about the world. It only needs to understand one person deeply. That's why it can be small and still be powerful."
Think about it: a small AI that knows your tastes, habits, goals, strengths, and weaknesses perfectly, versus a giant AI that knows a little about everything. Which one is more useful in your daily life?
How is this possible?
None of this is easy, of course.
With the technology available in 2025, a full implementation is not yet within reach. But the pieces are falling into place one by one.
Advances in algorithmic design have made it possible to shrink models while preserving performance. Model compression techniques have matured to the point where cutting a large model down to one-tenth its original size barely dents its capabilities.
Long-term memory technology is advancing too. Rather than keeping everything in active memory at once, new approaches store only what matters most and retrieve it quickly when needed.
On-device AI chips are also moving fast. Apple's Neural Engine, Qualcomm's AI processors , these chips are growing more powerful by the year, and smartphones can now run AI models of real complexity.
A picture of the future
Sam Altman imagines the world ten years from now like this.
"By around 2035, most people will have their own personal AI. It will grow alongside you from birth, remember every moment of your life, and be there to help you for as long as you live."
"It won't be a mere assistant. It will be your external brain, your closest friend, your wisest advisor."
"And the most important thing: that AI will be entirely yours. No one else can access it. You alone control it."
This is the future of AI that Sam Altman envisions.
Not a future where some vast superintelligence rules the world, but one where every person has their own perfect AI partner. He believes that future is more beautiful, more human, and safer.
"AI isn't here to replace human beings. It's here to make each and every one of us more capable."
This is the true meaning of a small reasoning model with a one-trillion-token context.
Kim Kyung-jin
Lawyer · Former Member of the National Assembly · AI Policy Researcher
© 2026 Kim Kyung-jin. All rights reserved.