"Vietnamese Brain" in Google's AI project and the race with ChatGPT

(Dan Tri) - 20 years ago, Luong Minh Thang was a math student of teacher Le Ba Khanh Trinh. Now he has returned to his old school to introduce to his teacher the artificial intelligence (AI) software that can solve difficult Olympic-level math problems.

One day in early December 2023, Dr. Luong Minh Thang visited the school where he spent his high school years in Ho Chi Minh City, introducing the artificial intelligence software AlphaGeometry to his former teacher, Dr. Le Ba Khanh Trinh. AlphaGeometry uses a neural language model and is trained with large-scale synthetic data, which can solve IMO (International Mathematics Olympiad) problems. In a test of 30 latest Olympic-level problems, AlphaGeometry demonstrated an amazing ability to solve 25 problems, which is equal to the achievement of an IMO gold medalist. The main author of AlphaGeometry is Dr. Trinh Hoang Trieu (30 years old, graduated from New York University) and a team of senior experts at Google DeepMind: Dr. Luong Minh Thang, Dr. Le Viet Quoc; and Dr. Yuhuai Wu (co-founder of xAI, formerly worked at Google). Solving IMO problems at the medal level is the dream of most excellent math students around the world . Dr. Le Ba Khanh Trinh became the first Vietnamese to win an IMO gold medal (in 1979), and is the only Vietnamese to date to have won the special prize of this competition thanks to his beautiful and concise solution to a geometry problem. When Dr. Luong Minh Thang introduced AlphaGeometry to his former teacher, he received a "very impressive!" share. However, "teacher Khanh Trinh was still not satisfied with the AI's solution because it lacked the soul and beauty of a solution that he expected," Dr. Thang said.

Dr. Luong Minh Thang (far right) and Dr. Le Ba Khanh Trinh (middle) are discussing a problem from the 2015 IMO competition that AlphaGeometry solved. (Photo: Wendy Uyen Nguyen)

For his part, Dr. Luong Minh Thang believes that AlphaGeometry marks an important milestone towards software with human-like intelligence and the ability to self-learn. This is the prerequisite for moving towards artificial superintelligence AGI. A former math student at the High School for the Gifted, Ho Chi Minh National University, Dr. Luong Minh Thang studied Computer Science at the National University of Singapore, received a PhD in Computer Science from Stanford University (USA) in 2016 and has spent 7 years working on Google's AI projects. At Google DeepMind, Dr. Luong Minh Thang built the most advanced models in both language (QANet, ELECTRA) and vision (UDA, NoisyStudent). He co-founded the Meena project, the world's best chatbot in 2020, which later became Google LaMDA, Bard and is now Gemini - Google's main AI product. Besides his relationship with AI during his studies and research, Dr. Thang also met Master Wendy Uyen Nguyen, Director of Global External Relations, founder of the Stanford Institute for Microbiology and Epidemiology (Stanford University) thanks to AI. Journalist Vo Thanh, Dan Tri newspaper, had a conversation with two young researchers Luong Minh Thang and Wendy Uyen Nguyen.

Vo Thanh: How did Dr. Luong Minh Thang become a senior researcher at Google? - Luong Minh Thang: I officially worked at Google in 2016, but the relationship started in 2014 when I interned at Google Brain. At that time, I participated in a project to improve translation quality, researching the application of artificial neural networks to help translation programs automatically translate complex sentences instead of translating single phrases as before. The mission of this project is to help machines understand the meaning of words in many different languages, and at the same time process long paragraphs. With the new approach, what we did in 2 years is evaluated as a total of 20 years ago. Also from 2014 to 2016, I completed my PhD thesis on translation, becoming one of the pioneering researchers in the field of deep learning, applying machine learning methods based on artificial neural networks to develop software that can train itself in machine translation. Our research supports the translations served at translate.google.com, allowing users to translate text, web pages, and even voices. Google Translate is a product deployed in many Google services, having a broad and profound impact in helping to bridge language barriers around the world. In 2018, I co-founded the Meena project, a chatbot that can chat with users about anything. This was a new direction because at that time, chatbots from Google or Microsoft were limited to simple tasks, tending to specialize in a certain field.

With Meena, we wanted to build a chatbot that could converse about virtually anything a user wanted to, with specific and reasonable content. In other words, how can anyone converse with a machine as if it were a natural conversation with a wise person, without being limited by domain, knowledge, and without being stuck or confused. In 2020, Meena became the world's best chatbot. The Meena model had 2.6 billion parameters and was trained on 341 GB of text filtered from public domain social media conversations. Compared to the OpenAI GPT-2 model, Meena had a 1.7x larger model and was trained on 8.5x more data. However, Google did not release Meena at that time because it was afraid of the risk. This was the period when Microsoft had just released an AI chatbot and encountered some problems such as answering wrong information, arguing with users, racism... Just a short time after introducing this chatbot, Microsoft had to take it down to fix it. This development made Google cautious with Meena. By the end of 2022, ChatGPT was launched and quickly attracted attention thanks to its ability to chat fluently in many different fields of knowledge, then Google started to... startle and follow. Currently, I continue to lead a number of important projects at Google, building better AI models for reasoning, logical analysis, problem solving, image processing, etc.

Vo Thanh: Hello Wendy Uyen Nguyen. Mr. Luong Minh Thang just shared his connection with the field of artificial intelligence, so how did the connection between the two young Vietnamese researchers in the US come about? - Wendy Uyen Nguyen: I met and got to know Mr. Thang because of our common interest in AI. My initial major was in psychology, but later I decided to switch to hospital administration (Master's degree from UCSF School of Medicine - University Of California San Francisco), then got an additional Master's degree in senior leadership in business administration from Stanford University. Academically, Mr. Thang and I follow two completely different paths, one is a scientist, the other is a strategist/business administration expert, but perhaps because of this, we can see different perspectives and support each other. Working in the medical field, I realize that the application of technology in general and AI in particular can help reduce risks for patients, because as we know, errors in some other fields can be fixed at a high cost, but errors in medicine often have to be paid for with human health and life. Previously, I participated in a project to research a technology model to help doctors practice their profession, then coincidentally met Mr. Thang in a group of friends, I asked him to be an AI consultant for the project. At first, we probably didn't think about love. But besides our interest in AI, we also met in the desire to contribute to the community, to be a bridge for scientific and technological cooperation between the United States and Vietnam. Mr. Thang and I are also very passionate about training generations of AI talents in Vietnam. We were destined to be together on such a path. - Luong Minh Thang: I would like to share that both my husband and I are passionate about music . Wendy studied piano at the Conservatory since she was young and played very well, and I loved to sing. Many times we met in a group of friends and played and sang together, which was very fun, and from then on we became closer.

Mr. Luong Minh Thang and Ms. Wendy Uyen Nguyen when they were students at the Faculty of Business Administration, Stanford University (Photo: NVCC)

Vo Thanh: So, besides applying technology, in the story that Mr. Thang and Ms. Wendy Uyen Nguyen just told, AI also plays a role in connecting emotions. - Luong Minh Thang: Yes, in Vietnamese, if we add an accent to the word "AI", it becomes the word "Ai", which also means love. Vo Thanh: Do Mr. Thang and Ms. Wendy Uyen Nguyen always agree on AI or have they ever argued about a specific issue, such as the benefits and risks of AI? - Wendy Uyen Nguyen: I am not an engineer specializing in AI, but in the medical field, I see that everything has two sides, good and bad. Talking about risks, it is the fear that AI will become too smart, escape human control, and even destroy humans in the future. That fear is understandable. But I think AI researchers have also thought about this issue, and governments have also considered how to develop AI within control. So the more important issue now is how to apply AI in practice, to help improve human life. For example, applying AI in self-driving technology or in medical examination and treatment...

- Luong Minh Thang: If we look back 10 years ago when I was doing research under the guidance of Professor Christopher Manning - a leading professor in the field of applying deep learning to natural language processing, no one would have imagined that ChatGPT or Gemini could help us write a poem for our wife, write an email to our boss asking for a raise, etc. Building a natural language model that can perform tasks in all fields with high conversation quality was something no one could have imagined before, because each task has its own requirements, very complex. For example, translation requires different things than writing emails, and even more different things than writing poetry... Making a machine write rhyming poetry was an extremely difficult task, almost impossible 10 years ago. But today, this is already possible. With the current development of AI, I think next year, it will reach important milestones, such as there will be short films with Hollywood quality made from AI, or breakthroughs in the field of mathematics... So what about the next 10 years? I am sure that AI will have developments beyond our imagination. When AI develops too quickly, many people will be worried because they do not know what will come. That is why there is research on the circle of AI control, so that AI will develop within that circle, not "jump" out by itself and outsiders cannot attack and take control of AI. Personally, I am very happy and excited about the rapid development of AI and believe that AI will bring many benefits to humanity. For example, in the medical field, with generative AI, humanity can move towards a future where someone provides input information about gender, age, weight, medical records... then AI will create the most suitable medicine formula for them. So each person will have their own medicine formula. Or ineducation , when I went to school, I had to have a handheld calculator to solve advanced math problems, but now AI is a supercomputer. With AI, children can find new mathematical laws, new research on physics, space, time... That means AI can help them explore the universe, explore black holes in time and mysterious things that humanity does not fully understand. Demis Hassabis, founder and CEO of Google DeepMind, once said something I really like, that is, AI has the potential to become one of the most important and beneficial technologies ever invented, a tool to help us learn about science and the future of humanity.

Vo Thanh : Returning to the Meena project, Google once had a powerful chatbot but did not release it, while OpenAI went ahead with the launch of ChatGPT in November 2022. What challenges does this situation pose to Google's AI project team? - Luong Minh Thang: ChatGPT was born, attracting great attention, and was a shock that led to Google having to issue a "Code Red" order for the entire company. The "Code Red" order means a red alert, the company is in a dangerous state. Right after ChatGPT appeared, we entered the 100-day campaign of the AI ​​race. From the Meena chatbot, our team developed it into Bard (currently Gemini), Google's main AI product, released in February 2023 (about 100 days after ChatGPT was born). The reason we were able to deploy such a difficult project at such a fast pace was because Google had done research on AI before. For example, my research on translation is based on the transformer architecture (a deep learning model for natural language processing), which is the architecture that supports most major language models and has revolutionized the field of AI. ChatGPT and Gemini are both developed on the transformer architecture platform. Moreover, Google has very good engineers who possess the ability to build everything from nothing to something; then Google also masters GPU chips and huge data sources for AI training, many product distribution channels such as Gmail, Youtube... Of course, besides the strengths and competitive advantages, I think because Google is a large corporation, the implementation of projects often has to follow a careful process. This makes Google slow in releasing products. Other companies are more flexible, sometimes releasing products with errors.

Mr. Luong Minh Thang and Ms. Wendy Uyen Nguyen met with Mr. John Kerry - former US Secretary of State, to share about efforts in realizing the vision of training and developing a talented young generation in the field of Artificial Intelligence in Vietnam, September 2023 (Photo: VietAI)

Vo Thanh: I heard that when Google issued a red alert, engineers had to step into the battle "Put both hands on the keyboard! Even if we have to work overtime, skip meals and sleep, we have to solve the problem". So what happened during that 100-day campaign? - Luong Minh Thang: Actually, when developing the chatbot product Bard, no one said it had to be done in 100 days, but everyone understood that they needed to make every effort for Google's survival. That effort brought productivity three times, five times, and even 10 times higher than usual. Sometimes I jokingly told my team that I had just started working but by Tuesday it felt like the week was over. Because the large amount of work we set for a week was solved in just two days, and in fact, everyone had been thinking and discussing since the weekend before to get started right away at the beginning of the week. To me, those 100 days felt like a year. Perhaps "in every misfortune there is a blessing", thanks to the birth of ChatGPT, everyone at Google became more united, more dedicated, and together aimed at the goal of creating a really good AI product that is really useful for users. The race was full of pressure, but it was also a memorable time because it coincided with Christmas in the US. In the early days, Sundar Pichai - CEO of Google and Sergey Brin - co-founder of Google, came down to the company to visit our project team, chat and encourage us. Before that, Larry Page and Sergey Brin, the two founders of Google, were no longer in leadership roles at the company, meaning they "retired", no longer in the office. But when the AI ​​project started, Sergey Brin spent time sitting and talking with the engineers, and there were evenings when he texted the project team "Who wants to go out for dinner at 8pm?". This is something I find very interesting.

Vo Thanh: We can imagine the 100-day campaign as the peak period. After this period, the AI ​​race between Google and other technology companies will not stop but will continue every day, every hour? Luong Minh Thang: Announcing an AI product is an important milestone. After the product starts to have users, the work will enter the stage of receiving feedback and improvement... The story of the current stage is that we have to know how users receive the product, are they satisfied, what do they want? User feedback is extremely important. ChatGPT can quickly improve the product thanks to having many users and receiving a lot of feedback. Google's AI products are released later but also receive a large number of users. Google's advantage is that it has many product distribution channels such as Gmail, Youtube... In most of Google's products today, there is a "shadow" of Gemini, which means integrating AI. For example, when you make presentation slides, you can use AI to make the presentation more beautiful and intuitive. I myself now start using commands for AI to write when writing programming, I just check later, saving a lot of time. Our AI project team initially had only 40 people, after releasing the product, Google mobilized a powerful force of thousands of engineers to participate. You can imagine our team as the vanguard in a battle, after going into battle and solving the most difficult problems, we retreat to see the whole picture and have time to delve into new discoveries . For a long time, people have talked about AI, talking about imitation, how to make machines imitate humans. This is AI in the 2020 period, with the biggest goal of building chatbot software that can talk like humans. Now we have passed that stage, entering the next stage of developing AI systems that can find new knowledge about mathematics, medicine, invent new medicines, etc. This is a new journey, there will be more difficulties and challenges than before, but it is definitely a journey worth exploring.

Vo Thanh: In the vision of scientists at Google, how will AI develop in the next few years? - Luong Minh Thang: I am leading a team of more than 50 scientists and engineers to carry out a large project for Google. Our mission is to build an AI system that is capable of deep thinking and high-level reasoning. We imagine that to solve a simple problem, AI will reason through many steps, but to solve a difficult problem in the Olympic exam, AI will have to think deeply, not only reason through many steps but also know how to link everything together. This is a prerequisite for moving towards artificial super intelligence AGI. The story of AGI in the future will revolve around the issue of discovering new knowledge. For example, AGI can solve the Millennium Problems. Of the 7 Millennium Problems, only one has been solved, the remaining 6 have not been solved yet. Or AGI itself can become a PhD student, a Nobel Prize winner for an invention in a certain field, etc. In my mind, perhaps society will have to change to adapt to the development of AI. I have not yet imagined the full change, but I hope that there will be interaction between humans and machines in discovering new knowledge. One reality we will have to face is that at some point AI will be smarter than humans. So what will humans do to adapt to that reality? Maybe people will think about building a legal corridor, but I think a little differently. I believe that in the future there will be a biological combination of hardware and software in the direction that the neurotechnology company Neuralink is doing. That is, developing implantable brain-computer interfaces, or in other words, expanding and enhancing the functions of the human brain, so that humans become stronger with AI.

Vo Thanh: What projects are the group of 50 Google scientists and engineers that Dr. Thang mentioned above working on? - Luong Minh Thang : I am leading a special project of Google Deepmind with tasks spread across two different offices. One office is in Mountain View, the Silicon Valley of the US, and one office is in London. Our project is part of Google Deepmind's overall mission, which is to build the next generation of AI systems to solve today's most difficult scientific and technical challenges. Joining us are not only AI scientists and engineers but also leading experts in mathematics. Perhaps this is also a chance for me, after 20 years since I studied with Mr. Le Ba Khanh Trinh and participated in the national math competition, I am now able to return to math and use AI to solve math problems. Vo Thanh: From the perspective of a leading AI scientist, how does Dr. Luong Minh Thang view Vietnam's potential and opportunities in the field of artificial intelligence?

- Luong Minh Thang : I think Vietnam's greatest strength is its human resources, its young talents. How to promote this resource is something I have been concerned about for a long time. Since 2018, I have established VietAI, a non-profit organization that contributes to training the next generation of AI talents in Vietnam, bringing them to the international environment and putting Vietnam on the world AI map. Vietnamese young people are very intelligent, hard-working in studying and researching, but they lack an environment to develop their talents. To date, VietAI has trained more than 4,000 students, equipping them with the latest knowledge about artificial intelligence. Many students have achieved initial successes. Currently, Vietnam has 4 young people who are Google Developer Experts, all of whom came from VietAI. Google Developer Experts is a talent search program by Google, with the desire to find experts who are not only excellent in skills, but also have the desire to contribute to the programming community of that country. There are some interesting examples, such as Nguyen Ba Ngoc, who started out knowing nothing about AI, went from Hanoi to Ho Chi Minh City to study the first course of VietAI, then returned to open VietAI in Hanoi, helping to train AI engineers with Google certification. Or Nguyen Hoang Bao Dai, who was also an early member of VietAI and is now a familiar face in the Vietnamese technology community. Bao Dai is known as a musician who created an AI model that can compose 10 songs per second. We are promoting training in generative AI, with the goal of training about 1,000 young people this year and by 2030 training 100,000 high-quality AI experts for Vietnam. Besides VietAI, my colleagues and I have continued to launch the New Turing Institute, a social enterprise that trains and inspires the next generation of AI talent in Southeast Asia.

Vo Thanh: Besides Mr. Thang, Ms. Wendy Uyen Nguyen is also an active person to promote AI bridge between the United States and Vietnam. Can you share some information about this? - Wendy Uyen Nguyen: Expected on August 18, Mr. Thang with me and some units will hold AI conference in Ho Chi Minh City. We have recently seen a number of leaders of large technology corporations have come to work in Southeast Asia, consider markets in this area and discuss the issue of cooperation in development as well as semiconductor and chip industry. The question is whether Vietnam is ready for this "war" yet? We want to contribute ourselves to the spread of knowledge about AI, not only directly training but also invite the leading professors and scientists about AI in the US to come to Vietnam to present. We also want to invite Google's senior leaders to Vietnam in the coming occasion, to see the potential of the Vietnamese market and to make AI bridge between the US and Vietnam become more bustling. In the field of medicine, I will continue to cooperate with professors and startups to create useful bridges for young people in the country.

The co -founding team and advisers of New Turing Institute, Vietai in the meeting and welcomed President and CEO of Nvidia Jensen Huang (Photo: Vietai).

Vo Thanh: Mr. Thang and Wendy Uyen Nguyen have any advice for young people who want to develop their career in the field of science and technology as well as who? - Luong Minh Thang: You need to pay attention not only to study well theory but to practice. On the internet there are open libraries, you can take the models there to improve and share with others what you do. When we hire talented engineers, we often pay attention to what the candidate has written, how your project looks. This means that we look at practical information, evaluate through the product, not just based on the theoretical memorization. - Wendy Uyen Nguyen: The focus point that I want to share is the leadership and soft skills thinking. Along with developing professional skills, young people also need to be prepared on these two sides, because through contact with many of you, I see this is what you are weak. This is very important, because for example, if you do research, you must also understand the output market of the product, grasp how to build a team ..., from which you can build your startup company. Sincerely thank Dr. Luong Minh Thang and Master Wendy Uyen Nguyen. Source: https://dantri.com.vn/xa-hoi/bo-nao-viet-trong-du-an-ai-cua-google-va-chien-dich-chay-dua-voi-chatgpt-20240713175306476.htm