AI engineers share their views at a recent seminar held by Nectec titled “Challenges, Potentials, and Opportunities for ChatGPT in Thailand”.
Artificial intelligence (AI) engineers from public and private organizations in Thailand collaborated to develop a generative AI-powered chatbot in Thai. This is expected to benefit several areas, from education to business.
The National Electronics and Computer Technology Center (Nectec) recently announced partnerships with three organizations to initiate the OpenThaiGPT project.
The three organizations are the Thai Artificial Intelligence Entrepreneurs Association (AIEAT), the Thai Artificial Intelligence Association (AIAT), and the NSTDA Supercomputer Center (ThaiSC).
Generative AI is a type of AI that can interact with users and create data in natural language, from articles to multimodal content such as images, videos, and audio.
The OpenThaiGPT project tracks the popularity of generative AI, which started when the ChatGPT chatbot took the world by storm last November. This AI-powered tool can answer questions in beautifully written sentences, help programmers code, and even compose music.
Created by OpenAI, an AI lab in San Francisco, ChatGPT made headlines by surpassing 1 million users within a week of its launch.
Now other global tech giants are jumping on the bandwagon to create their own versions.
Thepchai Supnithi, Director of AI Research Group at Nectec, said at the Center’s recent seminar “Challenges, Potentials and Opportunities of ChatGPT in Thailand”: “.
Thepchai, who also heads AIAT, said Thailand needs to have the ability to develop, modify and exploit the full potential of the Thai language using AI.
According to Nectec, the OpenThaiGPT project is a chat-based assistant project, and understanding Thai is essential to enable dynamic information retrieval and a user-friendly interface.
He said OpenThaiGPT is developed on an open source basis, is freely distributed, and has options for customization and extension.
timing is everything
“Thailand has a wealth of local language data from Pantip.com and LANTA at ThaiSC, the largest supercomputing facility in ASEAN, so it is an ideal time to start this project,” said Thepchai. says.
The latest release, OpenThaiGPT 0.1.0-alpha, is capable of various tasks such as question and answer, machine translation, step-by-step explanation, paraphrasing, and coding suggestions.
Future releases will increase the number of OpenThaiGPT parameters from 3.74 billion, improving language understanding and natural interaction, he said.
This project uses public data of which Pantip.com is one of the current data contributors.
Thepchai said the project has always paid attention to data privacy and copyright and encourages more data contributors to join the project.
Kobkrit Viriyayudhakorn, President of AIEAT, said OpenThaiGPT can be connected with external systems and can be easily extended, customized and further developed by anyone.
The project can also be integrated with other applications and serves as a building block for developing next generation AI in Thailand.
Use cases for OpenThaiGPT include chatbots, news summaries, automated translation, and document classification.
Nectec natural language processing researcher Prachya Bookwan says OpenThaiGPT 0.1.0-alpha is capable of answering common questions, translating languages, and coding.
Version 1.0.0, due out mid-year, contains up to 10 billion parameters, he said.
Additionally, the OpenThaiGPT team aims to develop multimodal options that allow physical references to be specified across languages.
Prachya said the technology could work in smartphones in the future.
Pantip.com founder Wanchat Padungrat said he was happy to help the project make a positive impact on society.
“If this project does not deliver good results, it could drag us down,” Wanchat said.
AI you can trust
Viwan Jarerattanachat, ThaiSC’s head of scientific support and domain research, said the project should improve Thailand’s AI capabilities.
However, she admits it’s still a case of trial and error and requires a safety review to ensure it doesn’t harm people or be misused.
The project requires transparency, accountability and responsibility, Viwan said.
Apivadee Piyatumrong, a senior researcher in Nectec’s AI research group, said many countries have guidelines and frameworks for creating trustworthy AI.
In Thailand, the Ministry of Digital Economy and Society produced “Digital Thailand – AI Ethics Guidelines” in 2018 to include ethical considerations in AI development principles, Apivadee said.
Regarding the development of OpenThaiGPT, she said there are three main areas to consider to ensure its reliability.
First, Apivadee said, safety measures need to be put in place to prevent harm to life and property that could result from the use of this technology. For example, when using OpenThaiGPT in the medical field, it is necessary to consider safety and efficacy.
Second, transparency is essential for OpenThaiGPT. It should be clear where the data and models used come from and what their quality and performance are.
Third, she said, accountability and governance are key factors in ensuring that OpenThaiGPT is trustworthy.
Nectec executive director Chai Wutitiwatchai said Thailand has a master plan for a national AI strategy that was approved by the cabinet last year.
The country’s AI Government Reserve Index score has improved from 60th in 2020 to 31st in 2022, he said.
According to Kobkrit, an AIEAT study titled ‘The State of AI Business in Thailand in 2022’ reported that total AI revenue increased by 24% from 35.2 billion baht in 2020 to 40.8 billion baht in 2021. It has been.
This amount is primarily from 105 AI companies, 50 system integrator companies, and 50 hardware automation companies.
Touchapon Kraisingkorn, co-founder of tech startup Amity, said he hopes OpenThaiGPT will become the de facto open source GPT model and gain wide acceptance in the Thai market.
He said GPT technology has immense ability to solve inefficiencies in both public and private sectors.
“This is a once-in-a-decade technology that will enable Thailand to make breakthroughs in productivity and technological capabilities,” said Mr Touchapon.
Enterprise use
Panutat Tejasen, chief executive of ThaiGPT Co, a new AI startup, said the company sees business opportunities by leveraging generative AI technology and the popularity of ChatGPT.
The company is open to using other generative AI models, he said.
Thai cryptocurrency exchange Bitkub is the main investor in ThaiGPT Co.
Panutat said ThaiGPT aims to develop, integrate and customize AI solutions using the Thai language to serve enterprise customers.
The company is already developing AI-based customer service solutions for its customers to answer their product and service inquiries and provide the ability to upsell and cross-sell.
He said the development of the solution should be completed within two months.
Augmented Intelligence
Ruangroj Poonpol, group chairman of Kasikorn Business-Technology Group (KBTG), said that in recent years, many companies see AI as playing a more important role in various business functions.
“The rise of generative AI like ChatGPT has created a buzz in the business world that humans will soon be replaced by AI,” he said.
Human-AI collaboration brings the highest efficiency and sustainability to work processes, says Ruangroj.
This collaboration brings with it the concept of “augmented intelligence,” in which AI is designed to help enhance human work and capabilities, not replace them.
“Through the creation of human-centric machine learning models, AI is trained to receive human feedback and convert that data into guidelines to help humans make decisions and specific actions more accurately and accurately. ” said Ruangroj.
KBTG and the MIT Media Lab developed AI through research called K-GPT (Knowledge-GPT). It leverages the power of ChatGPT to provide deep domain knowledge by using more natural language in conversations.
In this project, I improved my Thai language skills to be able to answer questions and offer advice from different angles.
KBTG Labs and the MIT Media Lab also collaborated on a proof-of-concept study called “Kookid.” This research uses two of his conversational agents, Kana and Kacha, which provide responses from different angles required for users’ decision making.
