Category Archives: Data Mining

An AI Origin Story

Posted on by
Reply

Nowadays, no TV or movie franchise worth its salt is deemed complete unless it has some sort of origin story – from “Buzz Lightyear” to “Alien”, from “Mystery Road” to “Inspector Morse”. And as for “Star Wars”, I’ve lost count as to which prequel/sequel/chapter/postscript/spin-off we are up to. Origin stories can be helpful in explaining “what came before”, providing background and context, and describing how we got to where we are in a particular narrative. Reading Jeanette Winterson’s recent collection of essays, “12 Bytes”, it soon becomes apparent that what she has achieved is a tangible origin story for Artificial Intelligence.

Still from “Frankenstein” (1931) – Image sourced from IMDb

By Winterson’s own admission, this is not a science text book, nor a reference work on AI. It’s a lot more human than that, and all the more readable and enjoyable as a result. In any case, technology is moving so quickly these days, that some of her references (even those from barely a year ago) are either out of date, or have been superceded by subsequent events. For example, she makes a contemporaneous reference to a Financial Times article from May 2021, on Decentralized Finance (DeFi) and Non-Fungible Tokens (NFTs). She mentions a digital race horse that sold for $125,000. Fast-forward 12 months, and we have seen parts of the nascent DeFi industry blow-up, and an NFT of Jack Dorsey’s first Tweet (Twitter’s own origin story?) failing to achieve even $290 when it went up for auction, having initially been sold for $2.9m. Then there is the Google engineer who claimed that the Lamda AI program is sentient, and the chess robot which broke its opponent’s finger.

Across these stand-alone but interlinked essays, Winterson builds a consistent narrative arc across the historical development, current status and future implications of AI. In particular, she looks ahead to a time when we achieve Artificial General Intelligence, the Singularity, and the complete embodiment of AI, and not necessarily in a biological form that we would recognise today. Despite the dystopian tones, the author appears to be generally positive and optimistic about these developments, and welcomes the prospect of transhumanism, in large part because it is inevitable, and we should embrace it, and ultimately because it might the only way to save our planet and civilisation, just not in the form we expect.

The book’s themes range from: the first human origin stories (sky-gods and sacred texts) to ancient philosophy; from the Industrial Revolution to Frankenstein’s monster; from Lovelace and Babbage to Dracula; from Turing and transistors to the tech giants of today. There are sections on quantum physics, the nature of “binary” (in computing and in transgenderism), biases in algorithms and search engines, the erosion of privacy via data mining, the emergence of surveillance capitalism, and the pros and cons of cryogenics and sexbots.

We can observe that traditional attempts to imagine or create human-made intelligence were based on biology, religion, spirituality and the supernatural – and many of these concepts were designed to explain our own origins, to enforce societal norms, to exert control, and to sustain existing and inequitable power structures. Some of these efforts might have been designed to explain our purpose as humans, but in reality they simply raised more questions than they resolved. Why are we here? Why this planet? What is our destiny? Is death and extinction (the final “End-Time”) the only outcome for the human race? Winterson rigorously rejects this finality as either desirable or inevitable.

Her conclusion is that the human race is worth saving (from itself?), but we have to face up to the need to adapt and continue evolving (homo sapiens was never the end game). Consequently, embracing AI/AGI is going to be key to our survival. Of course, like any (flawed) technology, AI is just another tool, and it is what we do with it that matters. Winterson is rightly suspicious of the male-dominated tech industry, some of whose leaders see themselves as guardians of civil liberties and the saviours of humankind, yet fail to acknowledge that “hate speech is not free speech”. She acknowledges the benefits of an interconnected world, advanced prosthetics, open access to information, medical breakthroughs, industrial automation, and knowledge that can help anticipate danger and avert disaster. But AI and transhumanism won’t solve all our existential problems, and if we don’t have the capacity for empathy, compassion, love, humour, self-reflection, art, satire, creativity, imagination, music or critical thinking, then we will definitely cease to be “human” at all.

The Bibliography to this book is an invaluable resource in itself – and provides for a wealth of additional reading. One book that is not listed, but which might be of interest to her readers, is “Chimera”, a novel by Simon Gallagher, published in 1981 and subsequently adapted for radio and TV. Although this story is about genetic engineering (rather than AI), nevertheless it echoes some of Winterson’s themes and concerns around the morals and ethics of technology (e.g., eugenics, organ harvesting, private investment vs public control, playing god, and the over-emphasis on the preservation and prolongation of human lifeforms as they are currently constituted). Happy reading!

Next week: Digital Perfectionism?

 

An open letter to American Express

Posted on by
2

Dear American Express,

I have been a loyal customer of yours for around 20 years. (Likewise my significant other.)

I typically pay my monthly statements on time and in full.

I’ve opted for paperless statements.

I pay my annual membership fee.

I even accept the fact that 7-8 times out of 10, I get charged merchant fees for paying by Amex – and in most cases I incur much higher fees than other credit or debit cards.

So, I am very surprised I have not been invited to attend your pop-up Open Air Cinema in Melbourne’s Yarra Park – especially as I live within walking distance.

It’s not like you don’t try to market other offers to me – mostly invitations to increase my credit limit, transfer outstanding balances from other credit cards, or “enjoy” lower interest rates on one-off purchases.

The lack of any offer in relation to the Open Air Cinema just confirms my suspicions that like most financial institutions, you do not really know your customers.

My point is, that you must have so much data on my spending patterns and preferences, from which you should be able to glean my interests such as film, the arts, and entertainment.

A perfect candidate for a pop-up cinema!

Next week: Life After the Royal Commission – Be Careful What You Wish For….

 

Big Data – Panacea or Pandemic?

Posted on by
1

You’ve probably heard that “data is the new oil” (but you just need to know where to drill?). Or alternatively, that the growing lakes of “Big Data” hold all the answers, but they don’t necessarily tell us which questions to ask. It feels like Big Data is the cure for everything, yet far from solving our problems, it is simply adding to our confusion.

Cartoon by Thierry Gregorious (Sourced from Flickr under Creative Commons – Some Rights Reserved)

There’s no doubt that customer, transaction, behavioral, geographic and demographic data points can be valuable for analysis and forecasting. When used appropriately, and in conjunction with relevant tools, this data can even throw up new insights. And when combined with contextual and psychometric analysis can give rise to whole new data-driven businesses.

Of course, we often use simple trend analysis to reveal underlying patterns and changes in behaviour. (“If you can’t measure it, you can’t manage it”). But the core issue is, what is this data actually telling us? For example, if the busiest time for online banking is during commuting hours, what opportunities does this present? (Rather than, “how much more data can we generate from even more frequent data capture….”)

I get that companies want to know more about their customers so they can “understand” them, and anticipate their needs. Companies are putting more and more effort into analysing the data they already have, as well as tapping into even more sources of data, to create even more granular data models, all with the goal of improving customer experience. It’s just a shame that few companies have a really good single view of their customers, because often, data still sits in siloed operations and legacy business information systems.

There is also a risk, that by trying to enhance and further personalise the user experience, companies are raising their customers’ expectations to a level that cannot be fulfilled. Full customisation would ultimately mean creating products with a customer base of one. Plus customers will expect companies to really “know” them, to treat them as unique individuals with their own specific needs and preferences. Totally unrealistic, of course, because such solutions are mostly impossible to scale, and are largely unsustainable.

Next week: Startup Governance

 

Assessing Counterparty Risk post-GFC – some lessons for #FinTech

Posted on by
Reply

At the height of the GFC, banks, governments, regulators, investors and corporations were all struggling to assess the amount of credit risk that Lehman Brothers represented to global capital markets and financial systems. One of the key lessons learnt from the Lehman collapse was the need to take a very different approach to identifying, understanding and managing counterparty risk – a lesson which fintech startups would be well-advised to heed, but one which should also present new opportunities.

In Lehman’s case, the credit risk was not confined to the investment bank’s ability to meet its immediate and direct financial obligations. It extended to transactions, deals and businesses where Lehman and its myriad of subsidiaries in multiple jurisdictions provided a range of financial services – from liquidity support to asset management; from brokerage to clearing and settlement; from commodities trading to securities lending. The contagion risk represented by Lehman was therefore not just the value of debt and other obligations it issued in its own name, but also the exposures represented by the extensive network of transactions where Lehman was a counterparty – such as acting as guarantor, underwriter, credit insurer, collateral provider or reference entity.

Before the GFC

Counterparty risk was seen purely as a form of bilateral risk. It related to single transactions or exposures. It was mainly limited to hedging and derivative positions. It was confined to banks, brokers and OTC market participants. In particular, the use of credit default swaps (CDS) to insure against the risk of an obiligor (borrower or bond issuer) failing to meet its obligations in full and on time.

The problem is that there is no limit to the amount of credit “protection” policies that can be written against a single default, much like the value of stock futures and options contracts being written in the derivatives markets can outstrip the value of the underlying equities. This results in what is euphemistically called market “overhang”, where the total face value of derivative instruments trading in the market far exceeds the value of the underlying securities.

As a consequence of the GFC, global markets and regulators undertook a delicate process of “compression”, to unwind the outstanding CDS positions back to their core underlying obligations, thereby averting a further credit squeeze as liquidity is released back into the market.

Post-GFC

Counterparty risk is now multi-dimensional. Exposures are complex and inter-related. It can apply to any credit-related obligation (loans, stored value cards, trade finance, supply chains etc.). It is not just a problem for banks, brokers and intermediaries. Corporate treasurers and CFOs are having to develop counterparty risk policies and procedures (e.g., managing individual bank lines of credit or reconciling supplier/customer trading terms).

It has also drawn attention to other factors for determining counterparty credit risk, beyond the nature and amount of the financial exposure, including:

  • Bank counterparty risk – borrowers and depositors both need to be reassured that their banks can continue to operate if there is any sort of credit event or market disruption. (During the GFC, some customers distributed their deposits among several banks – to diversify their bank risk, and to bring individual deposits within the scope of government-backed deposit guarantees)
  • Shareholder risk – companies like to diversify their share registry, by having a broad investor base; but, if stock markets are volatile, some shareholders are more likely to sell off their shares (e.g., overseas investors and retail investors) which impacts the market cap value when share prices fall
  • Concentration risk – in the past, concentration risk was mostly viewed from a portfolio perspective, and with reference to single name or sector exposures. Now, concentration risk has to be managed across a combination of attributes (geographic, industry, supply chain etc.)

Implications for Counterparty Risk Management

Since the GFC, market participants need to have better access to more appropriate data, and the ability to interrogate and interpret the data, for “hidden” or indirect exposures. For example, if your company is exporting to, say Greece, and you are relying on your customers’ local banks to provide credit guarantees, how confidant are you that the overseas bank will be able to step in if your client defaults on the payment?

Counterparty data is not always configured to easily uncover potential or actual risks, because the data is held in silos (by transactions, products, clients etc.) and not organized holistically (e.g., a single view of a customer by accounts, products and transactions, and their related parties such as subsidiaries, parent companies or even their banks).

Business transformation projects designed to improve processes and reduce risk tend to be led by IT or Change Management teams, where data is often an afterthought. Even where there is a focus on data management, the data governance is not rigorous and lacks structure, standards, stewardship and QA.

Typical vendor solutions for managing counterparty risk tend to be disproportionately expensive or take an “all or nothing” approach (i.e., enterprise solutions that favour a one-size-fits-all solution). Opportunities to secure incremental improvements are overlooked in favour of “big bang” outcomes.

Finally, solutions may already exist in-house, but it requires better deployment of available data and systems to realize the benefits (e.g., by getting the CRM to “talk to” the loan portfolio).

Opportunities for Fintech

The key lesson for fintech in managing counterparty risk is that more data, and more transparent data, should make it easier to identify potential problems. Since many fintech startups are taking advantage of better access to, and improved availability of, customer and transactional data to develop their risk-calculation algorithms, this should help them flag issues such as possible credit events before they arise.

Fintech startups are less hamstrung by legacy systems (e.g., some banks still run COBOL on their core systems), and can develop more flexible solutions that are better suited to the way customers interact with their banks. As an example, the proportion of customers who only transact via mobile banking is rapidly growing, which places different demands on banking infrastructure. More customers are expected to conduct all their other financial business (insurance, investing, financial planning, wealth management, superannuation) via mobile solutions that give them a consolidated view of their finances within a single point of access.

However, while all the additional “big data” coming from e-commerce, mobile banking, payment apps and digital wallets represents a valuable resource, if not used wisely, it’s just another data lake that is hard to fathom. The transactional and customer data still needs to be structured, tagged and identified so that it can be interpreted and analysed effectively.

The role of Legal Entity Identifiers in Counterparty Risk

In the case of Lehman Brothers, the challenge in working out which subsidiary was responsible for a specific debt in a particular jurisdiction was mainly due to the lack of formal identification of each legal entity that was party to a transaction. Simply knowing the counterparty was “Lehman” was not precise or accurate enough.

As a result of the GFC, financial markets and regulators agreed on the need for a standard system of unique identifiers for each and every market participant, regardless of their market roles. Hence the assignment of Legal Entity Identifiers (LEI) to all entities that engage in financial transactions, especially cross-border.

To date, nearly 400,000 LEIs have been issued globally by the national and regional Local Operating Units (LOU – for Australia, this is APIR). There is still a long way to go to assign LEIs to every legal entity that conducts any sort of financial transaction, because the use of LEIs has not yet been universally mandated, and is only a requirement for certain financial reporting purposes (for example, in Australia, in theory the identifier would be extended to all self-managed superannuation funds because they buy and sell securities, and they are subject to regulation and reporting requirements by the ATO).

The irony is that while LEIs are not yet universal, financial institutions are having to conduct more intensive and more frequent KYC, AML and CTF checks – something that would no doubt be a lot easier and a lot cheaper by reference to a standard counterparty identifier such as the LEI. Hopefully, an enterprising fintech startup is on the case.

Next week: Sharing the love – tips from #startup founders