57 MARTINDALE CENTER FOR THE STUDY OF PRIVATE ENTERPRISE bot Scientist Adam’s prediction of three yeast genes responsible for production of a specific enzyme, a hypothesis validated by protein purification of the gene products (King et al., 2009). Another example is AI Hilbert, recently developed by mathematicians and scientists, that can form new polynomial laws and equations (Cory-Wright et al., 2024). AI Hilbert could in some cases obtain the desired symbolic expression purely from a complete and consistent background theory without the use of numerical data, making possible scientific discoveries in areas of study where there are limited data and background theory and where gathering data is expensive (Cory-Wright et al., 2024). Discovery-based AI business model Discovery-based AI’s potential as an intangible product to transform Taiwan’s economy may appear theoretical rather than practical for real-world trade. However, the rise of data analytic services and big data applications has already demonstrated the viability of intangible products in global markets and has done so rather successfully. The ability to collect and rapidly analyze massive amounts of data—that can be sold as a product through analysis of historical patterns to predict future trends—has created multiple ways for data to be monetized and to transform various sectors (Wu et al., 2019). Ripple effects from that transformation are evident throughout business and society, from the education of college graduates with degrees in data science and related fields to the formation of new workplace roles such as data scientists (Walker, 2015). For instance, India was able to use data analytics to fuel its economy. Recognizing the potential of the relatively untapped market, India’s government nurtured a talent pool through educational programs such as the National Skill Development Mission, which partners with institutions in India to provide certifications and programs relating to data analytics (Technology sector in India…, 2023). India additionally stimulated data analytics through initiatives like Digital India, which encourages digital empowerment of its society. The plan is intended to boost India’s GDP via increased productivity, new jobs, and digitally centered businesses (Sharma, 2016). With such policies, the country has been able to export its digital services, positioning itself as a global leader in the data analytics industry. India has been able to take a significant share of the profits of the industry, which grew at a compound annual growth rate of ~26% between 2018 and 2023, equivalent to over US$2B in annual revenue (Technology sector in India…, 2023). As global markets increasingly embrace intangible products and cloud-based operations, Taiwan can capitalize on this shift by forming novel digital services, particularly with the introduction of discovery-based AI. As previously highlighted in the discussion of STEM education and tech infrastructure, Taiwan’s workforce is adept in fields ranging from the basic sciences and mathematics to AI and biomedical engineering (Charting Taiwan’s scientific…, 2020). Given that discovery-based AI is an emerging industry, and Taiwan’s workforce strengths are broad across STEM, it is not practicable to pinpoint a specific product to focus on but rather to suggest emphasis on the broader segment of discovery-based AI oriented to scientific applications. Some promising investment areas include services for molecular modeling, materials discovery, hypothesis generation, and experimental optimization. Molecular modeling comprises platforms that simulate computational biophysics to model biological molecules. Examples include drug discovery services, such as the company Atomwise, which predicts drug candidates based on simulations of molecular interactions, and DeepMind’s AlphaFold 3, which predicts biomolecule structure. Materials discovery is used to find novel optimized materials by modeling quantitative structure-activity relationships, which are quantitative measures of chemical structure in relation to a physical property or a biological effect (Liu et al., 2017). For instance, the Lawrence Berkeley National Laboratory open access database Materials Project computes properties of generated mediums to screen for applications such as optimizing battery electrodes. One former hypothesis generation service, Eureqa, developed mathematical equations based on experimental data, with a famous case the rediscovery of Newtonian laws using raw physics data (Schmidt & Lipson, 2009). Experimental optimization is demonstrated by services such as IBM RXN for Chemistry, which predicts chemical reactions and recommends optimal pathways to synthesize molecules. As the discovery-based AI industry has emerged only recently and seeks to establish itself, general industry valuations are not yet available. However, individual company valuations within the sector show encouraging trends. Recursion Pharmaceuticals, a drug discovery company, had a US$2.6B market capitalization as of December 2024 (Recursion Pharmaceuticals…, 2025), and Citrine Informatics, a materials discovery company, had a valuation of ~US$140M as of November 2022 (Citrine Informatics, 2025). Such valuations indicate a favorable outlook for the profitability of the growing discovery-based AI industry.
RkJQdWJsaXNoZXIy MTA0OTQ5OA==