Latest / The Best of the Money Show / AI and gambling: Is South Africa’s regulation falling behind?
Transcript
- 0:00Speaker 1: Well, some interesting thoughts emerging in the last little while
- 0:02Speaker 1: about how online gambling is using AI, how AI helps
- 0:06Speaker 1: to simplify the choices facing gamblers. Some questions about whether
- 0:10Speaker 1: AI could in fact help prevent problematic gambling, could help
- 0:14Speaker 1: protect people gambling. But the real bigger issue is around
- 0:18Speaker 1: the regulation of all of this. Wendy Rosenberg is the
- 0:21Speaker 1: head of digital media and electronic advertising. At Worksmans. Wendy,
- 0:26Speaker 1: good evening. I mean, I can see the scope for AI.
- 0:29Speaker 1: Some of the work is slightly repetitive. Be very good
- 0:33Speaker 1: at pattern recognition, that kind of thing. I presume that
- 0:36Speaker 1: AI is finding all sorts of uses in online gambling.
- 0:41Speaker 2: Hi, Stephen. Yes, absolutely, I think. AI is increasingly embedded
- 0:46Speaker 2: across the board and gambling is no exception. So we're
- 0:50Speaker 2: seeing it in a number of AI enhancements in the
- 0:54Speaker 2: betting space. I think particularly in the online betting space
- 0:59Speaker 2: where AI has an ability to crunch numbers at scale.
- 1:03Speaker 2: So what we see, for example, is an ability to
- 1:06Speaker 2: process live information in addition to historical data. So an
- 1:11Speaker 2: example in a sports betting context would be live updates
- 1:15Speaker 2: regarding player injuries or weather conditions. And it really enables
- 1:19Speaker 2: betting companies to offer more bets to more opportunities to
- 1:23Speaker 2: place bets. And, for example, it wouldn't be just a
- 1:25Speaker 2: case of win or lose anymore. It could be, is
- 1:29Speaker 2: Harland going to score? Is McKenna going to score? Who's
- 1:32Speaker 2: going to win the MTN8? So, you know, it allows algorithms,
- 1:37Speaker 2: helps companies to understand users' nuanced behaviors and preferences and
- 1:42Speaker 2: to provide much more tailored offerings and also improve the
- 1:45Speaker 2: client's experience and reduce choice overload. So we're seeing it
- 1:50Speaker 2: being applied in a number of ways in the gambling context.
- 1:54Speaker 1: Can it protect people? Can it spot problematic gambling as well?
- 1:59Speaker 2: Yes, absolutely. So there have actually been quite a few
- 2:03Speaker 2: quite promising studies on how AI can be used to
- 2:08Speaker 2: detect harm. And again, AI can crunch the numbers and
- 2:13Speaker 2: we have predictive AI that can look at historic data
- 2:17Speaker 2: and then use that to predict problem gambling. For example,
- 2:22Speaker 2: looking historically at behavior like the number of bets made
- 2:26Speaker 2: or the amount of the bet placed. So they're quite
- 2:30Speaker 2: promising results in predicting problem gambling.
- 2:34Speaker 1: Obviously, I mean, this is moving all very quickly. Is
- 2:38Speaker 1: there a need for more sort of regulation or legislation
- 2:41Speaker 1: around the use of AI in gambling? I mean, one
- 2:44Speaker 1: of the things with AI, and I mean, there's a
- 2:46Speaker 1: story today about Jobeck City Power using AI for billing,
- 2:49Speaker 1: for example. is I kind of always feel you're one
- 2:52Speaker 1: hallucination away from utter disaster.
- 2:56Speaker 2: Yes, and I think that's absolutely true. In fact, we
- 2:58Speaker 2: asked an AI developer, we said to him, does your
- 3:02Speaker 2: AI hallucinate? And his response was, anyone who tells you
- 3:06Speaker 2: that their AI system doesn't hallucinate is hallucinating. So I
- 3:10Speaker 2: think people really need to be aware that AI systems
- 3:13Speaker 2: do hallucinate. Sometimes they just make things up. They're not
- 3:18Speaker 2: optimized for factual accuracy. Large language models are based on
- 3:22Speaker 2: filling gaps in sentences and making words that sound like
- 3:26Speaker 2: they make sense. So they do hallucinate. They do get
- 3:29Speaker 2: it wrong. And in fact, what we've seen as well
- 3:32Speaker 2: is that some AI, in fact, perpetuates and exaggerates human
- 3:39Speaker 2: biases and can operate quite unfairly. We've also seen that
- 3:45Speaker 2: some AI bots, for example, mimic or simulate problem gambling behavior.
- 3:51Speaker 2: So they pick up the same patterns and reinforce those
- 3:57Speaker 2: that they learn from humans. Sure.
- 3:59Speaker 1: Okay. Would the regulation of AI in gambling need to
- 4:03Speaker 1: be sort of on top of the regulation we have
- 4:06Speaker 1: for online gambling now? Would there need to be specific
- 4:09Speaker 1: provision about how it's used and and I suppose crucially,
- 4:13Speaker 1: the purpose it's used for within the industry?
- 4:16Speaker 2: I think we need AI principles in general. So what
- 4:20Speaker 2: we're seeing in countries like the UK is that there
- 4:24Speaker 2: are principles that apply to AI in general, and then
- 4:28Speaker 2: those can be applied to various sectors, such as gambling
- 4:33Speaker 2: would be one example. We've just seen that coming through
- 4:35Speaker 2: in Malta, for example, where the Malta Gaming Authority worked
- 4:39Speaker 2: with the Digital Innovation Authority in that country, and they've
- 4:42Speaker 2: just published an online, sorry, not online, an AI gambling policy. So, yeah,
- 4:49Speaker 2: I think that that will come with time, but I
- 4:52Speaker 2: suppose baby steps, we've got to start with generally applicable
- 4:56Speaker 2: AI principles first.
- 4:58Speaker 1: I mean, how intrusive can government be in this? I mean,
- 5:03Speaker 1: could you say to an online gambling company, you can
- 5:06Speaker 1: use the same eye model, but not that. You can
- 5:09Speaker 1: use it for this purpose, but not that. I mean,
- 5:12Speaker 1: if I were one of the online gambling companies, I
- 5:14Speaker 1: would sort of say, hold on, you can't kind of
- 5:16Speaker 1: come into my model like that. You can't come into
- 5:17Speaker 1: my business model in that particular way.
- 5:21Speaker 2: Yeah, I think as a general principle, you want to
- 5:24Speaker 2: regulate no more than necessary to protect people and to
- 5:28Speaker 2: achieve specifically tailored objectives. So you don't want to be
- 5:32Speaker 2: telling people how to run their business But what we
- 5:35Speaker 2: have seen, if you look at the EU, for example,
- 5:38Speaker 2: is that they've got a risk-based approach. And so the greater,
- 5:41Speaker 2: the higher the risk, the stricter the rules. The less
- 5:46Speaker 2: the risk to people, the more flexible the rules can be.
- 5:50Speaker 2: But I don't think you could have rules that dictate
- 5:52Speaker 2: to companies how to run their businesses. But I can
- 5:55Speaker 2: say you should be transparent, for example, when you're using AI.
- 6:00Speaker 2: you know, that would be an example. You should be
- 6:02Speaker 2: disclosing it. You should be fair in how you treat people.
- 6:05Speaker 2: And then it would be up to the operator to
- 6:07Speaker 2: ensure that it was fair in its dealings and, you know,
- 6:10Speaker 2: make sure it was doing that fairly, it's accountable, et cetera.
- 6:15Speaker 1: Wendy, I find it fascinating that online gambling is managed
- 6:19Speaker 1: by the provinces. So you can be gambling online, you know,
- 6:22Speaker 1: in Seapoint in Cape Town through a company licensed in Limpopo.
- 6:27Speaker 1: And I can't help but wonder, how that's sort of
- 6:30Speaker 1: going to end. Do you think that model might need
- 6:32Speaker 1: to change at some point? There's been promises of reviews
- 6:36Speaker 1: of legislation on this. I haven't seen any yet.
- 6:40Speaker 2: Yeah, you know, it's an age-old problem. We've got gambling
- 6:43Speaker 2: legislation in nine provinces, plus the National Gambling Act. A
- 6:47Speaker 2: lot of that legislation was drafted in the late 90s,
- 6:51Speaker 2: early 2000s, and we really do have a fragmented landscape,
- 6:57Speaker 2: and that's really not ideal. But of course, in terms
- 7:00Speaker 2: of the constitution, gambling is an area of what we
- 7:04Speaker 2: call concurrent national and provincial legislative competence. In terms of
- 7:08Speaker 2: the constitution itself, the provinces have concurrent jurisdiction. So I'm
- 7:15Speaker 2: not optimistic that that's going to be easy to change,
- 7:19Speaker 2: but that doesn't mean that the National Gambling Board can't
- 7:23Speaker 2: determine certain norms and standards, which I think could create
- 7:28Speaker 2: some kind of synergy and consistency. And, of course, the
- 7:32Speaker 2: provincial gambling authorities can also work together to try and
- 7:37Speaker 2: find consistent principles and synergy in their approach, I think
- 7:40Speaker 2: particularly on an area such as AI, which is so
- 7:44Speaker 2: nascent and rapidly developing.
- 7:47Speaker 1: Wendy, thanks so much. Wendy Rosenberg is the head of
- 7:49Speaker 1: digital media and electronic at Worksman's Attorneys.