John O’Neill is the SVP of Regional Sales, Western US and Canada for Silent Eight. Although he is now in sales, John began as an engineer, for example working on some of the early browser technology in Champaign Illinois and later at Motorola.
We are publishing a 5 part series on AI in Banking by Amber Sutherland, also of Silent Eight.
I have recently also been diving down some other AI rabbit holes, such as what Open AI released recently (GPT-3).
John is a practitioner at the front lines of deploying AI in financial services. So, he seemed like a good person to talk to.
Open AI and GPT-3 is general purpose bottom of the stack technology. I was keen to dig below the surface hype about AI to also look at the limitations and, within that, the practical applications today. In our Artificial Intelligence (AI) week on Daily Fintech in 2016 we wrote:
“This exemplifies the AI mantra that hard is easy and easy is hard.
- Hard is easy – Chess, Go and Jeopardy.
- Easy is hard – understanding what the expression on your mother’s face means.”
4 years later this still seems to be true. This human’s reaction – “phew we are still needed a bit longer”
I started by asking John to tell us something about Silent Eight, starting with the name (a silent electric version of the V8 engine?). No. It is more to do with Silent Eight being founded in Singapore where 8 is a lucky number. I knew that from my years in Singapore and it is why our annual subscription price is $143 (add those 3 numbers to get 8).
John is based in Chicago and there are employees all over the world. As John put it – “we are where the where banks are”.
I then asked him three questions that concern people who work in finance:
Q1 how can Banks use AI to more efficiently comply with anti money laundering regulations?
I knew from my career selling to banks that you have to sell to an existential risk issue. Failing to comply with anti money laundering regulations is an existential risk to banks and can land executives in jail. So if you have a solution to that, the bankers will pay attention. That is why Silent Eight sells AI compliance solutions to banks.
I also knew from working with Fintech startups that there is a tradeoff between time/friction and security/compliance. You can design the most perfectly secure/compliant system that takes so long and is such a pain for honest customers that you lose your customers. So Silent Eight works hard to deliver solutions that are both compliant and fast.
Money Laundering is a massive business. A 2009 study estimates money laundering at about 2.7% of the global economy annually and If it were a country it would be in the Top 10 economies in the world, between France and Brazil (#7 and 8).
So we can expect a lot of creative intelligence applied to coming up with patterns that fool the AI compliance machines. John confirmed that AI is evolving fast and while today’s AI is not good at spotting new patterns we should expect next gen AI to do much better. Even today AI is very good at spotting bad actors in existing patterns. Replaying what we learned 4 years ago:
This exemplifies the AI mantra that hard is easy and easy is hard.
- Hard is easy – reading millions of transactions or applications or financial statements in seconds.
- Easy is hard – spotting that strange anomaly from a scheme created to fool those AI machines.
Fortunately humans are good at the latter. Freed from the grunt work by AI machines, they will have the time to spot those unknown new patterns. John confirmed that once AI is taught a new pattern, it will track it it consistently, accurately and quickly. It won’t get fooled again by that pattern. The bad actors will have to apply a lot of creative intelligence to invent a new scheme (which they can only use one time).
Q2 How can investors use AI to find accounting fraud before the companies implode?
Silent Eight is focused on using a mixture of bank data and public data to pinpoint money laundering. Although accounting fraud isn’t strictly their problem domain, it’s not a big leap to see how the banks could use a similar approach to pinpoint accounting fraud.
The Investment Banking and Wealth Management part of the banks could profit from such a tool as it would benefit the Hedge Funds that they work with.
Market downturns surface a lot more frauds. As Warren Buffet puts it “only when the tide goes out can you see who was swimming naked”. So there is a lot of money to be made/saved today in finding accounting fraud before the companies implode.
Think Enron, Madoff in past cycles and Wirecard (so far) in this cycle.
The data is out there in the public domain. That is what XBRL is all about and why we track that so closely.
Of course if the owners control the data, they also control the XBRL data. Most accounting frauds are perpetrated by people right at the top of the company.
Which brings us back to my hopeful vision of AI machines and humans working together. The AI machines look through millions of financial statements looking for patterns similar to Enron, Madoff, Wirecard and all the accounting frauds from the past. Meanwhile a few creatively intelligent humans, freed from the grunt work by AI machines, slowly look for the new pattern where “something looks fishy” from the creative new fraudster.
Q3 How can financial education sites use AI to give consumers better early warning about scams?
This is fundamentally a business model issue. There is no obvious way to make money by saving millions of poor people from getting scammed. Technically it is the same solution of AI machines working with creatively intelligent humans.
Fortunately, as one learns by looking at Open AI, there are tech billionaires who will donate money to help AI to help humanity.
My big takeaway is that crooks will have to become so creative to beat the AI machine that they might as well apply that creative intelligence to something honest. They can invent a new scheme, but when they try replicating it they find the AI machine singing the Who song – won’t get fooled again. Once AI gets better the crooks won’t even be able to play those tricks once.
Daily Fintech’s original insight is made available to you for US$143 a year (which equates to $2.75 per week). $2.75 buys you a coffee (maybe), or the cost of a week’s subscription to the global Fintech blog – caffeine for the mind that could be worth $ millions.