By Daniel Tammas-Hastings on 25th September 2018
The use of artificial intelligence in alternative finance also presents new challenges, says RiskSave’s Daniel Tammas-Hastings.
The scalability of new technologies make the emergence of new dominant platforms using AI frameworks a distinct possibility, and could in turn lead to the existence of systemically important players which may not currently be on regulators’ radar. Whilst we believe that the use of AI can and should be good for markets there are risks to consider alongside the multiple potential benefits.
One area of excitement is that the cost savings created by streamlining operational processes should aid financial inclusion. In the UK, the high cost of providing financial advice has led to a ‘Financial Advice Gap’, new systems that replace Human advisors and can interact at low cost with the mass market could create an incredible amount of value in the near future.
In recent years users and providers of financial services have been talking more and more about artificial intelligence and machine learning. The largest institutions have spent billions investigating and researching the possibilities that these technologies can give us.
Their applications are limited not just too operational processes and digital infrastructure, but can be expanded to include the investment process itself. Sections of the industry are concerned about the commercial implications of the increased use of artificial intelligence and the very real possibility of these new technologies disrupting existing business models. Increased use of AI is becoming an important part of the fintech revolution.
Financial services providers large and small are embracing innovation and are looking at how AI driven processes can enable new products as well as improving existing business lines and streamlining operational processes.
Artificial Intelligence usually refers to technologies capable of performing tasks that normally require human intelligence and need fuzzy rather than binary logic. The ability to learn and improve upon failure is the key to the development of this intelligence, hence why we are seeing a great deal of reference to ‘Machine Learning’ in the marketing literature of firms embracing AI. As processes that can improve themselves can reduce the need for expensive (and sometimes error-prone) human intervention. But whilst the focus on AI is new and exciting, the process and application is not, the first AI technologies became commercially available in the 1980s, and I’m sure that everybody is familiar with IBM’s (AI) Watson. Not just Jeopardy fans!
Here are some of the terms that are being used in the industry.
Machine learning: This refers to the process of learning by discovering patterns in large data sets. Systems that use Machine Learning can improve their performance through iteration without the need for further development.
Deep learning: This is like Machine Learning but deeper, hence the name. It involves more complicated Machine learning algorithms that involve neural networks. Neural networks are structures that mimic the function of the brain.
Probabilistic inference: AI capabilities that use Data Science and Bayesian networks to identify
conditional dependencies of random variables from large data sets.
Whilst the data on the usage of AI in the Alternative Finance space is sparse. Interaction with various players in the ecosystem shows that many segments of the industry are embracing AI and machine learning. The use cases include non-investment functions such as fraud detection and KYC (Know Your Client) advertising and digital marketing, alongside investment tasks such as portfolio / risk management and optimised trade execution.
It is likely that AI will bring benefits to the capital markets in the long run. It is hoped by regulators that better data systems and management led by machine will lead to cost savings, more efficient markets, improved risk systems and ultimately more resilient and stable markets. However elderly and more cynical practitioners such as myself worry that the converse may be true, if regulators and the industry do not ensure that ‘crowded trades’ and ‘groupthink’ caused by dominant strategies don’t start to dominate the markets. As an example of hidden risks consider that if many players began using AI strategies to allocate capital but were unable to explain in detail the mechanisms of the model due to increasing complexity, then the firm’s management and regulator would struggle to adequately monitor and supervise their actions.
A further worry is, that if one form of AI were to become the dominant allocator of capital, the markets could become very brittle indeed. As at this stage the markets would become capable of mis-pricing in a consistent way, leading to higher volatility and adverse social outcomes. Indeed we have seen an over reliance on Data and Technology causing one form of capital allocation to become overly dominant before.
A decade ago, over-reliance on (flawed) data and advanced analytics (remember the Gaussian Copula) led to the development of the CDO and an over-investment in housing globally, think Subprime.
As the rating agencies spurred on by the banks worked out new ways to earn high yields with the safety of a Triple AAA asset using their hi-tech state-of-the-art quantitative systems. Unfortunately, as we now know it was an illusion and led to the Financial Crisis and colossal destruction of wealth.
The positive use cases for AI are numerous, but practitioners and regulators should be aware that that new technologies create new challenges.
Now in its sixth year, the AltFi London Summit returns on 18th March 2019 to 155 Bishopsgate. Last year proved to be a crucial turning point for the key players building the future of finance. Leading platforms launched oversubscribed IPOs, digital banks proliferated and mainstream financial institutions started their own disruptive propositions. With 2019 certain to be another landmark year, more questions will be asked by regulators with investor interest in disruption also poised for more rapid growth.