[ad_1]
South Korean conglomerate LG and Qraft Technologies partnered to launch an exchange traded fund that uses artificial intelligence to take investors’ emotions out of the stock-picking process. The LG Qraft AI-Powered U.S. Large Cap Core ETF (LQAI) , which debuted in early November, marks LG’s first step into tapping its AI technology for financial markets. Previously, it had used these capabilities for supply chain optimization, demand forecasting and purchasing raw materials for its chemicals business. The AI research division started testing its models against financial markets starting in 2022 “with great results,” according to Young Choi, director at LG AI Research. “We’re always looking for state-of-the-art technology and finding new ways to sort of create alpha in a unique approach to forecasting that’s a bit different than the traditional quants” said Choi. “We’re quite excited that this will also pay dividends within the financial markets.” Meanwhile, Seoul-based Qraft Technologies, which is backed by SoftBank , already has four other actively managed AI-powered ETFs. The partnership differs from the other offerings, however, owing to LG AI’s large language model capabilities and time series forecasting portfolio, which are continually being fine-tuned. LQAI uses LG’s AI tools to analyze financial data from large cap stocks to determine its 100 holdings and portfolio weight, forecasting individual stock prices four weeks out. The portfolio rebalances its holdings monthly, a move that helps to avoid “noises” that may be triggered if rebalancing took place more frequently, according to Qraft Technologies chief operating officer and Asia-Pacific CEO Francis Geeseok Oh. “The four-week frequency is fairly welcomed by advisors. If we rebalance too frequently, that could cause transaction cost issues [and] trigger taxable events,” he said. The LQAI focuses on large cap stocks, which Oh says are better suited toward the AI model. Small cap stocks, which feature higher idiosyncratic risks and more noise as a result, which make them a more difficult option for the model, which uses data as the primary decision-making source. The portfolio underwent its first rebalancing on Nov. 29. In its latest rebalance, the model raised its exposure to the information technology and communications services sectors, according to Weldon Rice, head of ETFs at Qraft. He added that one “unique decision” from the model was its increased allocation to the energy sector. Compared to Qraft’s other AI-powered funds, LQAI is currently more diversified in terms of securities and sectors, according to Rice. The 10 largest holdings in LQAI include UnitedHealth Group and energy companies Chevron and Exxon Mobil , in addition to Palo Alto Networks and JPMorgan . The fund currently has approximately $3.7 million in assets under management, with an expense ratio of 0.75%. An alternative to emotional bias The biggest advantage of having an AI-run portfolio is the lack of emotional bias in the decision-making process, Oh said. He has prior experience as an executive director at Vanguard and portfolio manager at Mirae Asset Global Investments. “As a human investor, it is really hard to not love the stock that I’m investing in. That attachment in the investment decision-making can trigger unnecessary risks,” said Oh. He noted that even during his time at Vanguard, retail investors were urged to be less emotional, regardless of market direction, so that they could make better decisions for the longer term. AI models don’t exhibit emotions when making investment decisions, and they are “much more ruthless than humans,” he added. “AI models are not shy about profit taking [and] taking an investment opportunity,” Oh said. When portfolio managers and human investment committees make decisions, conflicting opinions within a group might mean that they reach a compromise. Good decision-making in a group is better “for avoiding risk, but at the same time, it’s not necessarily an optimal decision for the investment,” Oh said. On the other hand, using AI models means that “the entire process is systematic, data-driven, and has some sort of transparency, instead of relying on one or two key people making decisions from just their guts or instinct.” Another way to put it is that “AI models are much more objective, or also cold-blooded [and] emotionless,” said Oh. Model weaknesses The work isn’t done for LG AI Research, said Choi. The AI model — specifically the company’s homegrown large language model, which he likened to OpenAI’s ChatGPT — needs to be further fine-tuned. “One known issue for language models is hallucination, which is one key homework assignment that we need to better optimize,” said Choi. Hallucinations in the context of large language models refers to when they generate incorrect or nonsensical information that appears accurate. Because of this issue, the large language model is not currently highly leveraged, Choi said. “Once we feel more competent, we will be slowly rolling this out a bit more and more,” he continued. According to Choi, the implementation of the large language model would help improve overall accuracy. There are also certain cases when an AI model can’t react as quickly as humans, such as in the case of an unexpected overnight event. “When a truly unexpected or unprecedented event happens, the AI model is able to ‘learn’ the event, but it takes a little time to adjust itself,” Oh said. Nonetheless, the rise in prominence of ChatGPT has also helped convince more investors on the merits of utilizing AI in their investment and development process. “But when we speak with portfolio managers, there is some natural resistance. I can understand the reason why,” Oh said. Oh remains optimistic for the possibilities ahead for AI applications in the financial sector. “AI can really transform asset management. So I knew I wanted to be part of it as soon as possible,” he said.
[ad_2]
