INNOVATION WITHOUT BORDERS

AI strengthens institutional investing at HESTA by revealing deeper market insights and supporting human judgement.

Trust and diverse perspectives help teams turn experimentation into better investment decisions. Combining quantitative modelling with qualitative economic analysis requires people to challenge established methods and work across different approaches. At HESTA, questioning the limitations of traditional forecasting models helped initiate the adoption of machine learning in 2021.

Machine learning can reveal market relationships that conventional models overlook. By examining interactions between variables across different market conditions, models can challenge prevailing assumptions. In 2023, HESTA’s models signalled opportunities to buy equities despite widespread recession concerns, prompting further analysis of the historical relationship between the yield curve and equity returns.

AI creates more room for critical thinking when its outputs remain open to scrutiny. HESTA checks model findings against fundamental analysis and uses inspectable code and spreadsheets to review calculations. Some work previously requiring three to six months now takes two weeks, freeing time to assess portfolio exposures, test scenarios, and interpret results.

Anygraph makes professional AI more reliable by combining AI adaptability with controlled workflows.

AI can add to auditors’ workloads when polished outputs conceal incomplete checks. Financial statement reviews require evidence to be connected across large volumes of varied documents. Inconsistent findings from repeated AI runs force humans to recheck and repair the work, weakening the productivity gains that automation promises.

Anygraph combines deterministic software with targeted AI to make workflows consistent and controllable. Anygraph structures tasks into predefined “graphs”, specifying which steps use conventional code and which require AI. This gives professionals greater control over the checking process while retaining AI’s ability to handle varied inputs, while reducing reliance on repeated model calls.

Anygraph makes AI findings traceable, easier to review, and more cost-efficient. In a financial statement demonstration, a RM1,000 discrepancy was traced to extracted figures, recalculated formulas, and its exact source location. Anygraph’s graph-based review cost US$0.10, compared with an estimated US$30 through a general AI workflow, illustrating how workflow design can improve reviewability and cost efficiency.

Cortical Labs demonstrates how biological computing could open a new frontier beyond conventional AI.

AI’s progress in digital intelligence has not translated easily into physical intelligence. Moravec’s paradox describes how computers can perform sophisticated intellectual tasks yet struggle with perception, movement, and interaction with the physical world. Progress in physical AI also faces constraints from the energy demands of conventional computing and the scarcity of large-scale physical-world training data.

Cortical Labs uses living neurons and silicon to create a form of biological computing. Adult cells can be converted into neurons and grown on electrode-equipped chips, allowing electrical signals to provide sensory information and feedback while neuronal activity controls an external system. In a Pong demonstration, predictable rewards and unpredictable penalties enabled the neuronal network to learn a simple task.

Biological computing is entering practical use, with hardware manufactured in Malaysia. Cortical Labs’ CL1 integrates living neurons grown on an electrode-equipped chip with a neural chamber and life-support system that keep the cells alive. Manufactured in Malaysia, it has been deployed at the National University of Singapore for biomedical research into neurological diseases. This provides an early application of the technology and demonstrates Malaysia’s capacity to manufacture emerging computing systems.

Quotes

“I definitely believe in the human-AI hybrid combination where AI is enabling the team to think a lot more as opposed to the AI doing the thinking for all of them.”
– Dr Alvin Tan

“If anyone were to ask me what technology is going to come next after AI, the answer is brain-computer interfaces.”
– Dr Hon Weng Chong

“We call this “provable finding”. We want people to find proof behind each and every finding that AI gets.”
– Zad Chin