18 Aug 2020 | 20:07 UTC London

INTERVIEW: Commodities producers join ranks of AI users to enhance decision-making: ChAI CEO

Commodities producers are increasingly making use of artificial intelligence and machine learning for short-term market forecasting, joining the ranks of speculators and hedge funds first to take up AI in the commodities space, while analysts and consumers continue to lag behind, according to the CEO of Commodities Artificial Intelligence.

ChAI is a UK-based company that uses AI and machine learning to make short-term price predictions, typically of up to three months, in the nonferrous metals, oil and petchems markets

"Speculators were the first to take up machine learning," said Tristan Fletcher in an Aug. 17 interview with S&P Global Platts.

Fletcher noted there used to be more speculators in the big investment banks, particularly in the US, before new laws were introduced after the financial crisis.

"More recently the producers have really got into this and some of them are very sophisticated," he said. "They have data that nobody else has. My view is that the banking analysts are more behind, perhaps because they typically make longer-term forecasts and do them less frequently. And the last people to adopt these tools are at the end of supply chains: the manufacturing companies and commodity end-users."

Nowadays, well over 50% of commodity traders, major hedge funds and investment banks active in the commodities space are using AI and machine learning to enhance decision-making in their everyday business, Fletcher estimated.

Machine learning is described as the study of computer algorithms that improve automatically through experience, using predictability, and is seen as a subset of artificial intelligence. AI and machine-learning tend to bring together data supplied and purchased from various sources on contracts: macroeconomic, econometric, production and inventories data, satellite data that could cost "millions of pounds a year" and freight data that could costs a fraction of this. Basic AI tools are already in use all around us, from search-engines on the internet, and in smart electronics: the trend is for the cost of these types of data to decrease as they come more widely available and can benefit from scale economies, which should help "democratize data," Fletcher said.

While AI and machine learning can be used to make short-term price predictions in any commodities markets, the "downside" with oil is that oil prices are very sensitive to geopolitical events that may be impossible to predict and could have major and sudden market impact, according to Fletcher. Metals prices meanwhile tend to be subject to forces including trade wars and sanctions that may lend themselves to machine learning models, he said.

Satellite images can also be useful for mapping of mines and mine inventories for major metals commodities such as copper produced in open-cast mines, while for a commodity such as lithium where tonnages produced and shipped are "a tiny amount" in comparison, they may not so useful. There is always the possibility that producers may "hide" their stocks by covering them over so that they are invisible on satellite images, he pointed out.

ChAI last week signed an agreement with the European Space Agency to proceed with a feasibility study entitled "Mitigate Commodity Price Volatility with Space Data," using satellite imagery to map and track key features of markets, such as changing activity levels at open cast mines and infra-red heat signatures omitted by smelters, in a bid to add value to market intelligence and improve price predictions. Oil tankers and storage may also be observed this way, in what is becoming "a very crowded market," Fletcher said.

Use in optimizing supply chains

Emerging from COVID-19, AI has been used to help optimize supply-chain information, according to Fletcher.

"One of the major impacts of COVID-19 has been the acceleration of a trend that was already in motion, making supply chains that were previously globalized, become more localized," he said. "AI is good at understanding exchangeability between different commodities and the complexities in the trade-offs between choices of places to source them."

While the use of AI and machine learning is becoming more widespread in short-term commodity price forecasting, the use by big players of trading algorithms which use AI is being more heavily regulated against, Fletcher said.

"Exchanges are pushing back on high frequency algorithmic trading – trying to reduce it," he said. "It can provide liquidity when times are good but withdraw it when times are bad: so doesn't necessarily work in the interests of all parties."

AI should be employed to explain markets, what moves them and in general help make them more transparent. It should not influence their movements directly, Fletcher said.

"The future of AI is explainability and the ability to leave an audit trail," he said. "This is crucial amidst the advent of new legislation in the financial services with regards to explainability, for example to prove an absence of bias and discrimination in loan provision or lack of manipulation in capital markets."

Register for free to continue reading

Gain access to exclusive research, events and more

Already have an account?Log in here