On November 6, 2023, Zalo AI Challenge 2023 officially announced 3 topics corresponding to 3 different areas of AI. Specifically, in the Natural Language Processing (NLP) area, with the Elementary Maths Solving topic - the teams are required to build an AI model to solve elementary math problems according to Vietnamese education standards.
In the image section, with the Advertising Banner Generation topic, AI will automatically design advertising banners based on the given descriptive information.
With the Background Music Generation problem – in the audio field, the AI model needs to “generate” background music based on requirements for melody, instruments, music genre, etc. All input requirements for the above 3 problems are expressed in text form.
All three problems in this year's competition are part of the Generative AI trend, a technology wave that is gaining attention worldwide. With Generative AI, scientists expect artificial intelligence to move towards self-producing content the way humans do.
This is a breakthrough development of AI, marking the transition from traditional data analysis and classification AI models to content-generating AI to effectively support human life in the future.
Generative AI is a fairly new AI trend not only in Vietnam but also around the world, so this year's topic is considered more challenging than previous times.
According to Dr. Chau Thanh Duc, Head of Research and Development at Zalo AI, every year, contestants can start building their own solutions and have a high chance of getting good results. This year, the teams need to invest in both research and training equipment.
“Competition teams should survey and research suitable models and then build solutions from there. Generative AI also requires certain requirements for training machines, so they need to determine their own direction, suitable for their current conditions,” Dr. Chau Thanh Duc further explained.
In addition, each problem itself also has its own challenges for the competing teams. In the Elementary Maths Solving problem, to solve elementary math problems, AI models must not only understand and answer, but also have the ability to reason and solve problems.
In the Advertising Banner Generation problem, AI models compete to create images – a difficult task when creativity has no boundaries and evaluation is not easy even for humans.
Similarly, in the Background Music Generation topic, sound in general and music in particular are things that humans can only feel, but cannot see or touch. Therefore, how to optimize the model will be a "brain-weighing" problem for the teams choosing this topic.
By choosing Generative AI as the main topic for this year's competition, through a prize of up to 15,000 USD and a quality training dataset, Zalo AI hopes that students and small research groups can access the necessary resources and have more motivation to develop Generative AI models.
From there, promoting this trend in the AI community, not just limited to large companies with great potential, helps Vietnam have more opportunities to catch up with the world's wave of generative AI technology.
Explaining this, the competition's organizers said that the more individuals and organizations participate in AI research and development, the more reason there is to expect more complete AI models in the future.
“Winning a competition is not as important as overcoming yourself to persevere on the journey of AI – which is inherently full of difficulties and challenges. With Zalo AI Challenge, we hope to motivate and inspire those who are pursuing this journey. We believe that subsequent research will be more inspired to continue and do better in the future,” Mr. Duc added.
In addition, community engagement and learning are equally important to promote the development of the Vietnamese AI industry. Research groups can share results to create competitive motivation and improve models for other groups, while the parties also learn and inherit experiences from each other.
For that reason, Zalo AI Challenge chose the Kaggle format and kept it the same throughout the 6 times it was held. During the competition, the competing teams will continuously update their model results on the Leaderboard Ranking so that they can directly monitor each other's training process in real-time.
At the same time, after the competition ends, Zalo AI always encourages teams to share their solutions with the AI community within the framework of Zalo AI Summit or on Vietnamese technology forums.
Zalo AI will also contribute training data from the competition so that the competing teams, and the broader AI community, can continue their research. Preparing the dataset is also an elaborate process with the desire to help AI people have quality datasets.
This year, Zalo AI's dataset is more focused on academic standards, aiming to be used for scientific research.
In addition to senior staff at Zalo AI, contestants also received advice from leading AI experts such as Professor Nguyen Le Minh, Director of the Interpretable AI Research Center at the Japan Advanced Institute of Science and Technology (JAIST), Associate Professor Tran Thanh Long, Deputy Dean, Research Director of the Faculty of Computer Science, Warwick University (UK), Dr. Ngo Duc Thanh - Head of the Faculty of Computer Science, Ho Chi Minh City University of Information Technology.
The competition officially opens on November 6, 2023 and the submission deadline on the training dataset ends on December 1, 2023. During this time, the results of the teams will be publicly available on the Leaderboard. After that, the models of the competing teams will be tested on a different dataset than the original training data and the final results of the competition will be calculated based on the model results on this dataset. The winning teams will share their solutions at Zalo AI Summit 2023.
Source
Comment (0)