Subscribe
About

South Africa must make moves on AI regulation

Simnikiwe Mzekandaba
By Simnikiwe Mzekandaba, IT in government editor
Johannesburg, 28 Feb 2024
Professor Johan Steyn, founder of AIforBusiness.net. (Photograph by Strike A Pose Studios)
Professor Johan Steyn, founder of AIforBusiness.net. (Photograph by Strike A Pose Studios)

The South African government needs to earnestly start looking at artificial intelligence (AI) regulation, as this year is going to be a landmark year for the technology.

So said professor Johan Steyn, founder of AIforBusiness.net, speaking at the recent ITWeb Governance, Risk and Compliance 2024 conference. He also unpacked some of the big AI themes he anticipates as the year unfolds.

According to Steyn, there’s progress with the European Union’s AI Act, which will be enacted into law in a year or two, with most of the world hopefully following its principles.

“One of its main principles that have been proposed is visibility. In other words, if I apply for a loan or a bond with my bank and I’m declined by an AI, the bank will need explain how its algorithm came to its conclusion.

“There’s bias in those systems – a person can be excluded based on their gender, ethnicity or because they come from a different part of town. There are a lot of reasons why these algorithms could say no to me that has nothing to do with my financial status.”

Unfortunately, SA is nowhere when it comes to AI regulation, he commented. “We [South Africa] rank 196th on the world AI-readiness index, compiled by the University of Oxford. We are 12th on the African continent, which includes Mauritius.

“Imagine the future our children will have – the lack of privacy and biases against them influenced by automated decision-making systems if AI cannot be regulated.

“The other question is whether regulation is the answer. If we do have AI regulation, how long will it take because it is complex and we know that innovation is always a few steps ahead of regulation. This is going to be a big theme in 2024.”

AI-embedded devices will take off in 2024.
AI-embedded devices will take off in 2024.

Steyn noted this year will increasingly see AI embedded in consumer-facing devices. On-device AI is an exciting area, he said. “Most likely all of us have smartphones but the AI doesn’t sit in your handset – you’re using cloud-based AI that’s being processed by your handset.”

Samsung’s new AI-embedded smartphone range marks the first for AI-on-device, he indicated. “We’re now going to see all the brands jump on this trend. The potential business applications can be incredible, especially from a security point of view.”

Last month, the South Korean smartphone maker unveiled Galaxy AI for the S24 smartphone series, which includes AI-powered features, like AI photo editing, text and call translations, and a new way to search online.

“We’ve spoken a lot about edge computing, where the computing happens in the security camera. Imagine bringing full-on generative AI onto the device that runs your security; full-on AI that’s not necessarily connected to any cloud on a device,” commented Steyn.

At MWC Barcelona, which is currently under way, Chinese smartphone brand Honor unveiled its AI strategy, introducing a suite of smart devices. It also announced MagicOS 8.0, the latest version of its Android-based operating system, with the company’s platform-level AI.

Last week, Honor rival Oppo unveiled its AI strategy after launching its China-based AI centre.

Everything AI was in the spotlight last year, with a wide range of tools, solutions and industries adopting the technology.

The launch of OpenAI’s ChatGPT in November 2022 brought to public attention the power of generative AI. Since then, Google’s Gemini, Microsoft’s Bing AI and Elon Musk's xAI have been launched.

Spotify is using OpenAI’s technology to translate podcast episodes into other languages, and LinkedIn’s AI-generated content helps create posts for users, according to research firm Forrester.

Steyn noted that another key AI theme this year is around large language models, with the next five years seeing these become 1 000 times more powerful than they are today.

“If you think content creators and writers have a challenge now…imagine where we will be in the near future.

“This isn’t only about language but it’s multi-modal. If you consider OpenAI’s latest text-to-video model and what it allows you to do with just a prompt. You can create beautiful things like training videos and so forth. However, the amount of deception we can have is also incredible. We’re going to see a lot in this domain this year and into the future.”

Share