AI is a powerful tool. There is no getting away from it, and in a world where every chancer calling themselves a venture capitalist is charging headlong to fund anything with ‘AI’ in a domain name, development is ricocheting around us all in what is essentially an uncontrolled way.
And earlier this month, the EU released its AI Act to regulate how the technology is to be used.
Legislation for AI was inevitable, and I for one am quietly glad that it is on the cards. Back in October 2023, Rishi Sunak made a speech on how we have a global responsibility to understand and address the risks it poses, and this reminded me of an unsettling TED Talk I watched all the way back in 2018.
The speaker was one Kai-Fu Lee, a Taiwanese computer scientist who works in Beijing, dresses like Dr. No, and delivered a TED Talk titled: “How AI can save our humanity,” (April 2018 - the link is below).
But at 4 minutes in, his segment describes how a Chinese work ethic of ‘996’ (09.00am to 21.00, 6 days a week, so 9-9-6!), and how Chinese companies are far more competitive than their western counterparts, claiming it’s a fight to the death. These attributes, he claims, will put China in the global lead for AI. Subsequent events have proved otherwise however (such as the Baidu debacle when its Ernie Bot failed to impress in 2023).
When I watched Kai-Fu’s speech however, I was depressed. Here was an industry leader claiming that AI can save our humanity, whilst promoting working conditions that would lead to burn out, and claiming that competition is so fierce that, logically, in such an environment, short cuts would almost certainly be used by stressed developers to get their product to market.
His whole argument seemed to wither, frankly, despite his attempt to claim that AI will do away with mundane jobs and let humans focus on the care sector and be more humane to one another, with compassion, love, and empathy being the new job growth sectors.
It’s a nice thought. The cynic in me scoffs slightly.
Nonetheless, my point is that AI is not a technology where we should be cutting corners, and with the EU’s new AI Act, we at least have a start.
What is in the EU’s AI Act?
The EU's new AI Act is designed to regulate the development and use of artificial intelligence within the Union, introducing a framework to ensure AI technologies are developed and deployed in a manner that's safe, ethical, and respects fundamental human rights.
Below, I review the main points from the act, and then discuss their impact on marketing.
Risk-Based Approach: The AI Act employs a risk-based classification for AI systems. Systems posing limited risk will need to meet transparency requirements, whereas high-risk AI systems, which could significantly impact health, safety, or fundamental rights, must adhere to strict regulations including risk management and data governance.
Prohibited AI Systems: The act bans AI applications that pose a clear threat to people's fundamental rights, including biometric categorisation systems and emotion recognition in certain contexts. It also outlaws social scoring systems and AI that could manipulate human behaviour.
Promoting Innovation: The act encourages innovation through regulatory sandboxes and real-world testing environments, allowing for the development and training of AI before market placement.
New Regulatory Bodies: It establishes new administrative structures, including an AI Office within the Commission for overseeing advanced AI models and fostering standards, and an AI Board for coordination and advice.
Penalties for Non-Compliance: Companies could face fines ranging from €7.5 million or 1.5% of global turnover to €35 million or 7% of global turnover for breaches of the act.
Implementation timeline: The final text of the AI Act is expected to be published in early 2024, becoming applicable two years after its entry into force, with some provisions kicking in earlier.
For businesses, particularly those using or planning to use AI systems, the act calls for early preparation to ensure compliance. This involves mapping processes, assessing AI systems against new regulations, and developing a robust AI governance strategy that aligns with business objectives and legislative requirements.
The AI Act represents a significant step by the EU to regulate a fast-evolving technology sector, aiming to balance innovation with ethical use and public safety, and, undeniably, it sets a precedent that could influence global standards for AI governance.
How will the EU’s AI Act change marketing?
Chatbots - the automation that increases your blood pressure!
The rise of chatbots in recent years is one key area where companies will very likely have to review their systems. Chat GPT has driven the adoption of Chatbots (very much to my personal chagrin frankly as I can’t stand them!). Alas, Gartner, with their usual sage-like foresight, have told us that nearly 50% of organisations will be using chatbots for customer care (is assisted suicide included in that I wonder?), and that 40% will deploy virtual assistants.
Nonetheless, perhaps my stone age prejudice will change as chatbots get injected with AI which will make them genuinely useful, but it is clear that any responding to a human’s emotional state, as might be likely as they are forecast to be used to give answers in an emergency, or to resolve a complaint (both areas where human customers will be under stress), are addressed in the EU AI Act. So their forecast most-use cases could now be under the spotlight.
And perhaps this is right. In an emergency, or making a complaint, isn’t it better to speak to a living human? I can’t help think that this might be an optimistic deployment, and we might see it rolled back after some years.
Social scoring - the invisible hand that can throttle us.
AIs have been proven to have biased applications in some cases. Infamously, predicative algorithms used by a US court for risk assessment was biased against black prisoners. It has also been the case that Amazon’s own hiring algorithm favoured men over women job applicants, as its training data was educated on resumes submitted over the previous decade, and the majority were male applicants.
Whilst social scoring is common in states like China, coercing obedience from the public, AI and Machine Learning will almost certainly end their way into credit ratings companies, affecting our ability to access loans, buy a home, or even jet off in a desperate hope to up our vitamin D levels! But more seriously, it will certainly be used on those earning less and who are credit dependent. It is this ‘invisible hand’ that we should be alert for, and the EU’s AI Act hopefully addresses this.
Likewise, advertising platforms that use AI, such as Google Ads Performance Max, or Meta / Facebook’s learning period, might restrict ads appearing to users it notes as undesirable. Is this not a form of consumer apartheid, where some members of society will be denied various opportunities due to training data? I’m all for targeted ads, but having it done out of sight and ‘under the bonnet’ is something that, in reality, means the advertiser is as much at the mercy of the algorithms as the recipient (or non-recipient).
Will the UK follow the EU’s AI Act?
Here under the leaden-skies of Blighty in a spring that has been as wet as a Baywatch babe, we are moving to legislate. I suspect it will be similar to the EU’s AI Act, to allow for product compatibility and ease of trade, although it does seem as though the UK will define AI differently to that of the EU Act, and they are trying to go for an innovation-friendly approach (Remember Kai-Fu Lee anyone?).
One of the great problems of course, is trying to legislate for technology that moves and develops at such a pace. AI is far from alone here, but perhaps uniquely over other technologies, AI might be able to help us in this area?
Now, what will Chat-GPT say when I ask it to regulate itself? At least it can’t swear back! (Yet).
References and Sources:
EU Shaping Europe’s Digital Future: The AI Act.
MIT Technology Review: Eu’s AI Act: What will and won’t change.
EU High Level Summary of AI Act:
Rishi Sunak’s Speech on AI: https://www.gov.uk/government/speeches/prime-ministers-speech-on-ai-26-october-2023
Kai-Fu Lee TED Talk: https://www.ted.com/talks/kai_fu_lee_how_ai_can_save_our_humanity?language=en (See 4 minutes in).
Baidu AI Bot fails to impress: https://www.theguardian.com/world/2023/mar/16/chinese-chatgpt-rival-search-engine-baidu-fails-impress-ernie-bot
US based justice algorithm biased against black prisoners: https://stories.surrey.ac.uk/ai-and-the-law/
Kommentare