Like millions worldwide, Southeast Asians have been trying out large language models such as Meta’s Llama 2 and Mistral AI — but in their native Bahasa Indonesia or Thai. The result has usually been gibberish in English.
This leaves them at a disadvantage, tech experts warn, as generative artificial intelligence (AI) transforms education, work and governance worldwide.
A Singapore government-led initiative aims to correct the imbalance with a Southeast Asian large language model (LLM), the first in a family of models named SEA-LION — Southeast Asian Languages in One Network — trained in the region’s languages and cultural norms.
Trained on data in 11 Southeast Asian languages including Vietnamese, Thai and Bahasa Indonesia, the open-sourced model is a cheaper and more efficient option for the region’s businesses, governments and academia, AI Singapore’s senior director for AI products Leslie Teo said.
“Do we want to force every person in Southeast Asia to adapt to the machine, or do we want to make it more accessible so people in the region can make full use of the technology without having to be an English speaker?” Teo said.
“We are not trying to compete with the big LLMs. We are trying to complement them, so there can be better representation of us,” he said.
There are more than 7,000 languages spoken worldwide. Yet LLMs including Open AI’s GPT-4 and Meta’s Llama 2, used to build AI systems such as chatbots and other tools, have largely been developed for and trained on the English language.
Governments and tech firms are trying to bridge this gap, with India creating datasets in local languages, an LLM in the United Arab Emirates powering generative AI tools in Arabic, and AI models in China, Japan and Vietnam in local languages.
These models could help local populations participate more equitably in the global AI economy that is largely dominated by big tech firms, said Nuurrianti Jalli, an assistant professor at Oklahoma State University’s school of communications.
“Regional LLMs are also needed because they support technology self-reliance,” she said. “Less reliance on Western LLMs could provide better privacy for local populations, and also align better with national or regional interest,” Jalli said.
Multilingual language models that are trained on text from several languages at once, can infer semantic and grammatical connections between high resource languages that have more data, and low resource languages, researchers say.
These models can be used in a variety of applications from translation to customer-service chatbots, to content moderation on social media platforms that have struggled to identify hate speech in low resource languages such as Burmese or Amharic.
About 13 percent of SEA-LION’s data is sourced from Southeast Asian languages — more than any other major LLM, Teo said, adding that more than 9 percent of its data is from Chinese text and about 63 per from English.
Multilingual language models, often train on translated text and other poor quality data that might have errors, so AI Singapore is “careful” about the data used in training SEA-LION, Teo said in his office at the National University of Singapore.
“The age of pristine data has passed — a lot of the stuff on the Internet now is LLM-generated material, so we need to verify and filter,” he said.
“We cannot be perfect, but we also cannot take out everything we consider to be bad,” he added.
More governments are contributing data, and businesses are testing SEA-LION, which can be deployed faster and is cheaper to fine-tune and adopt due to its smaller size, he said.
At Indonesian e-commerce company Tokopedia, a majority of customer interactions is in Bahasa Indonesia, so models “with that local fluency will enhance our ability to connect with customers and improve their experiences,” Tokopedia associate vice president of data science Paul Condylis said.
As more countries and regions build their own LLMs, digital and human rights experts fret that they would only reproduce dominant views expressed online, which could be particularly problematic in nations with authoritarian governments or strict media censorship, or those lacking a strong civil society.
Chinese social media platforms, for example, censor references to the Tiananmen Square uprising and criticism of the government, while several Southeast Asian nations have enacted laws to curb content that authorities deem misleading.
“Training models on such data risks perpetuating biased, prejudiced, incomplete and even misleading narratives,” Jalli said.
“The models may fail to surface important socio-political issues like human rights abuse, corruption, or valid criticism of political powers,” she said.
For example, in response to a query on Indonesian former president Suharto, Llama 2 and GPT-4 mentioned his spotty human rights record, while SEA-LION’s response focused largely on his achievements.
If a model is only trained on favorable articles about a government, then the model is “likely to adopt a worldview where the government is wholly positive and leave out dissenting viewpoints,” said Aliya Bhatia, a policy analyst at the Center for Democracy & Technology, a US non-profit.
“Regional LLMs might better reflect the linguistic and cultural nuances of local language speakers, but they might also have less information about the world in general,” she added.
“There is a real risk of government-backed models instilling a revisionist view of history and undermining democratic values.”
However, the alternative — relying entirely on Western LLMs with “disproportionately large influences” from wealthy, liberal, western democracies — means perpetuating different biases related to cultural values, political beliefs and social norms, AI Singapore said.
“These LLMs have a very particular West Coast American bias — they are very woke. They do not represent us,” said Teo.
“We are not saying ours is the only perspective — we are just trying to rebalance it.”
X-37B COMPARISON: China’s spaceplane is most likely testing technology, much like US’ vehicle, said Victoria Samson, an official at the Secure World Foundation China’s shadowy, uncrewed reusable spacecraft, which launches atop a rocket booster and lands at a secretive military airfield, is most likely testing technology, but could also be used for manipulating or retrieving satellites, experts said. The spacecraft, on its third mission, was last month observed releasing an object, moving several kilometers away and then maneuvering back to within a few hundred meters of it. “It’s obvious that it has a military application, including, for example, closely inspecting objects of the enemy or disabling them, but it also has non-military applications,” said Marco Langbroek, a lecturer in optical space situational awareness at Delft
AIRLINES RECOVERING: Two-thirds of the flights canceled on Saturday due to the faulty CrowdStrike update that hit 8.5 million devices worldwide occurred in the US As the world continues to recover from massive business and travel disruptions caused by a faulty software update from cybersecurity firm CrowdStrike, malicious actors are trying to exploit the situation for their own gain. Government cybersecurity agencies across the globe and CrowdStrike CEO George Kurtz are warning businesses and individuals around the world about new phishing schemes that involve malicious actors posing as CrowdStrike employees or other tech specialists offering to assist those recovering from the outage. “We know that adversaries and bad actors will try to exploit events like this,” Kurtz said in a statement. “I encourage everyone to remain vigilant
‘TERRORISM’: Israel slammed Palestinian President Mahmoud Abbas, saying that he has revealed his ‘true face’ by embracing the ‘rapists and murderers of Hamas’ Hamas yesterday announced that it had signed an agreement in Beijing with other Palestinian organizations, including Fatah, to work together for “national unity,” with China describing it as a deal to rule Gaza together once the war ends. Chinese Minister of Foreign Affairs Wang Yi (王毅), who hosted senior Hamas official Musa Abu Marzuk, Fatah envoy Mahmud al-Aloul and emissaries from 12 other Palestinian groups, said they had agreed to set up an “interim national reconciliation government” to govern post-war Gaza. “Today we sign an agreement for national unity and we say that the path to completing this journey is national
Soaring high across a gorge in the rugged Himalayas, a newly finished bridge would soon help India entrench control of disputed Kashmir and meet a rising strategic threat from China. The Chenab Rail Bridge, the highest of its kind in the world, has been hailed as a feat of engineering linking the restive Kashmir valley to the vast Indian plains by train for the first time. However, its completion has sparked concern among some in a territory with a long history of opposing Indian rule, already home to a permanent garrison of more than 500,000 soldiers. India’s military brass say the strategic benefits