Global technology company IBM sees a future where small language models will make artificial intelligence (AI) more accessible to organizations.
AI and automation are disruptive technologies that are rethinking or redesigning business processes.
IBM's general manager for Asia Pacific, Paul Barton, said this would enable businesses to significantly improve their productivity. Bangkok Post.
This approach not only enhances competitiveness but also enables differentiation in the marketplace, allowing companies to take the lead in their respective industries, he said.
“The most successful companies will be those that simultaneously invest in technology, strategy and business process optimization to harness the transformative power of AI,” Barton said.
Spending on AI systems in the region is expected to surge to $49.2 billion in 2026, up from $20 billion last year, according to IT research firm IDC.
Barton said that while AI will lead to job augmentation, computer augmentation of humans will be better than computers alone.
Alternative Models
He said that large language models (LLMs) offer enormous capabilities but require high levels of energy consumption and cost.
Models like GPT-4 boast trillions of parameters, and the computational power required to train, tune, and deploy these models is enormous, often reaching hundreds of millions of dollars per year.
“This limitation has led to the emergence of small language models (SLMs),” Barton said. With a parameter size of around 10 billion, SLMs offer users an alternative.
SLM can run at low cost on more readily available hardware, helping to democratize AI, allowing more amateurs and institutions to study, train and improve existing models, he said.
SLM can also run on small devices, enabling more advanced AI in scenarios such as edge computing devices and IoT.
Burton said the IBM Watsonx Granite model, with 13 billion parameters, is an example of SLM that organizations can leverage to ensure ethical and responsible use of AI and regulatory compliance.
Earlier this month, IBM expanded the availability of its AI software on the Amazon Web Services marketplace to 92 countries.
Because SLMs are focused on performing a specific knowledge area or use case, this specialization improves performance and efficiency for tasks such as transcoding customer service, document translation, and summarization.
SLM has applications in a variety of industries, from healthcare to asset management, he said.
Burton said he expects to see many more use cases and applications emerge over the next six to 12 months as the technology continues to rapidly develop.
These miniaturized models will enable companies and users to train the models themselves on personal devices such as laptops, he said.
Further investment needed
As more hyperscale data centers are developed in Thailand, its digital infrastructure is driving increased productivity, GDP and wealth, Barton said.
But simply deploying a data center is not enough to create value, he said, and without meaningful use the marginal utility of the additional infrastructure is zero.
Getting the most out of digital infrastructure requires complementary technologies and strategies, including hybrid cloud, cybersecurity and AI and their applications, Barton said.
“At IBM, we are investing in hybrid cloud to help companies leverage their technology investments,” he said.
“We have a $500 million venture fund dedicated to investing in AI companies and will continue to look for meaningful opportunities to back.”
Approximately two-thirds of global commerce involves digital technology.
Burton said businesses and governments are pouring huge amounts of money into digital transformation efforts, totaling an estimated $6.8 trillion between 2020 and 2023.
In Thailand, technology spending is probably 0.7-0.8 percent of GDP, while in Singapore similar investment is 5 percent, he said.
In developed countries, the lower bound appears to be around 2% of GDP.
“Increasing spending on technology while improving business processes, the regulatory environment and education has the potential to lead to significant economic growth,” Barton said.
For Thailand to become a trillion-dollar economy, it needs to prioritise education and skills and invest in technology to enable business processes, he said.
Barton said that while AI will lead to job augmentation, computer augmentation of humans will be better than computers alone.
“Thailand needs to develop talent to collaborate and train new additions to the workforce. Countries that use computers effectively to collaborate will be the winners,” he said.
