Since Chat GPT was launched earlier this year, the world has seen a tidal wave of clickbait headlines hyping up artificial intelligence-powered (AI) technology. Readers will have undoubtedly been irritated by people on LinkedIn posting reels with titles like “how you can LIFE HACK AI with these SHOCKING Chat GPT plugins”
I don’t know whether these people have actually used Chat GPT before but my experience of it is that it’s not as amazing as they make it out to be. As some examples of this…
- I got Chat GPT to do a grammar / spell check on a long piece of writing I did for work. It came back with 10 errors. All of them were wrong – ie. they weren’t actually errors (one of them was to make the word ‘may’ plural). In one instance it changed the text I wrote and then said I’d made a spelling mistake. Note that spell check functions in Word have existed for two decades.
- Got it to try and find a solution to a very simple excel problem. It came up with five that were all based on formulae that didn’t work or make sense. It got it right the sixth time when it realised all you had to do was…cut and paste.
- Got it to check some historical facts. It just made up a bunch of stuff that never happened, including a battle that never happened in a place that doesn’t exist.
- Used it to plan a day of tourist activities on holiday. Again, it made up a bunch of places that didn’t exist (N.B. these places *never* existed, so this is not a function of it only having access to the internet up until 2021)
Having said all of this, it has been useful for some things…
- Making up funny stories (this is the primary use case)
- Some coding. Eg. I recently needed to get a passport urgently. There were no appointments. A close associate of CFDs Weekly used Chat GPT to create a web scraper that would check the passport appointments page and send a text to my phone when a new one appeared. The whole thing took around 20 minutes to set up and I got an appointment on the same day.
So all of this can be summed up as ‘Chat GPT is pretty amazing but not as amazing as everyone is making it out to be’.
But the last point does point to one potential use case for CFD brokers trying to attract traders. If you play around with Chat GPT, you can get it to write MQL code which can then be input into the MetaQuotes editor/compiler to make indicators or place trades.
The trouble is you are likely to still have to do debugging afterwards. If you are good at coding then this might not be so hard. If you have no experience then it’s going to be a more irksome, iterative process that will probably involve you telling Chat GPT ‘THIS DOESN’T WORK’ about 50 times before it actually comes up with something functional.
Leaving those teething problems aside, what that could mean is that you end up with a situation where the average person can create an automated trading strategy by taking simple text and converting it into code. This is not entirely novel as there is at least one company out there offering this as a solution, but presumably brokers have to pay them to integrate, whereas Chat GPT is free.
Now you may be thinking ‘hey but you just said that it doesn’t work well’. That’s correct but that is less important, from the broker’s perspective, than boosting engagement and trade volume. And given ‘how to’ YouTube videos on this subject are extremely popular, there is a marketing angle you could play into here as well.
Aside from attracting clients, the use of AI in other parts of a broker’s business doesn’t seem particularly novel to me. AI is a nebulous term and could apply to many pieces of tech. Today what most people seem to be referring to is a programme that is trained on a large data set and then able to perform a specific task. I don’t want to negate how impressive some of these technologies are but they aren’t that new.
For example, I remember speaking to a friend working at a crypto exchange in 2018 about how they used a tool to flag compliance breaches when people were trying to sign up for the platform. I also remember interviewing the founder of Simplex at around the same time and they were doing similar stuff to reduce fraud / chargebacks.
So what is actually new? Well, Chat GPT created a lot of hype, people realised that and now they think having AI is good for branding and attracting clients. Maybe that’s a cynical take and I’ll be proved wrong. On the other hand, how obvious was it that the picture in this article was made by AI?
CFD restrictions in South Korea
Some readers may be aware that there has been a recent crackdown on CFDs in South Korea. In the wake of a stock manipulation scandal, local brokers were prohibited from opening new accounts and existing clients prevented from opening positions.
It has been a bit tricky to pick apart what happened here, particularly as the investigations are still ongoing. However, from what I can tell there was a pump and dump scheme that took place over a period of several years. A few people involved in this have now been arrested.
These people had bid up a number of shares since 2019. When they ‘dumped’ their stock in April, people with long positions in CFDs on the relevant stocks got margin calls and then had their positions liquidated – which meant that the stock price got pushed down further. Korea Exchange, the country’s stock exchange, has a limit on how much a share price can fall in a day (30%) and at least 6 companies hit this limit for three consecutive days.
Looking through some local reports on the fallout, it seems as though regulators want to ensure brokers are holding enough capital to mitigate the risk that client losses exceed the amount of margin they put down, something that could have easily happened when those stocks crashed in April. So it’s Korea’s own SNB moment!
Another factor is client protection. CFDs in Korea were already only accessible to professional investors. Now restrictions on what counts as ‘professional’ are going to be made stricter, to the point that you will only be able to open an account via an in-person meeting. This is presumably because a lot of people got wiped out in the price drops.
Regulators also seem annoyed about how the transactions were conducted as they obfuscated what their origin was. In other words, it wasn’t clear that people were positioned in the market with CFDs or what the level of risk they were going to be liquidated was. Given that the sell offs appear to have been triggered by the hedging positions CFD providers made, I don’t get how it would be viable to see this (but I could also be wrong as to what they’re trying to do).
All of this was interesting to me, partly to figure out what actually happened but also because it’s clear none of it is actually relevant to most CFD providers that read this newsletter. For instance, I had no idea that Korea limited leverage on CFDs to 2.5x and to pro investors only.
As we noted last week, the most popular strategy for lots of brokers in markets like Korea is “stay offshore and keep doing stuff until the regulator says something”. So even though there may be more regulations, it’s not likely to make an impact on the people doing business there.