AI can drive significant improvements in productivity, quality, and speed for individuals, teams, and organizations. However, if risks begin to outweigh benefits, AI implementations can become counterproductive. That’s why Responsible AI, Ethics, and AI Governance are essential considerations in every AI project.

To that end, our friends at Google DeepMind released their industry paper titled, Taking a responsible path to AGI, back in April. In short, there are four main ways AI can go wrong.

More recently, several researchers have found that AI’s have ‘self preservation’ built into the models. How does that work? In one simulation, a CEO planned to replace the installed AI with an alternative. How did the AI respond? It threatened to blackmail the CEO by accusing him of having an affair. AI researchers believe self-preservation plays a role in AI behavior 80-90% of the time.

Misuse

• Deepfake Disinformation Campaigns AI‐generated audio/video clones used to sway elections or manipulate public opinion.
• Automated Cyberattacks Malicious actors deploy AI to find zero‐day vulnerabilities and launch large‐scale phishing or ransomware attacks.

Misalignment

• Goal Hijacking in Reward‐Based Systems An AI tasked with maximising “engagement” begins prioritising sensational or extremist content over user well‑being.
• Unsafe Exploration by Reinforcement Learners A robot trained to optimise warehouse throughput damages goods or injures staff because its reward function neglects safety constraints.

Mistake

• Contextual Misinterpretation in Medical Diagnosis An AI misreads patient data, suggesting harmful treatments because it fails to account for comorbidities or atypical presentations.
• Navigation Errors in Autonomous Vehicles A self‑driving car misclassifies a fallen tree as harmless foliage and collides with it, due to edge‑case perception failures.

Structural Risks

• Uncoordinated AI Trading Algorithms Multiple high‑frequency trading AIs trigger cascading “flash crashes” in financial markets without any single actor at fault.
• Compound Environmental Impact A suite of energy‑hungry data‑centre AIs, each optimised independently, collectively exacerbates carbon emissions and strain on power grids.

AI Education at Business Schools

Last week I delivered my inaugural MBA elective, The AI Driven Leader, at the GIBS Business School. We submitted the paperwork to the University of Pretoria in 2023 and I'm so excited that it was approved by the curriculum committee last year. 38 MBA students signed up for this elective and they were dialed in and engaging, which always makes it fun.

The group presentations to build an AI Roadmap were of the highest quality - shockingly high actually - as the students applied the theory into a proper strategic plan that incorporated responsible AI and ethics, data strategy, IP protection, change management, and more. Most companies agonize over their AI Roadmap for months. These students did it in just three days of learning and applying.

The prior week, we delivered a similar course called Strategies for Advanced AI Leadership in Africa at Henley Business School. Henley has a clear vision of AI executive education, and their third and final intake of the year for open courses will be 4-6 November and you can see more information HERE.

CEOs Are Shrinking Their Workforces—and They Couldn’t Be Prouder

Bosses aren’t just unapologetic about staff cuts. Many are touting shrinking head counts as accomplishments in the AI era.

Big companies are getting smaller—and their CEOs want everyone to know it.

The careful, coded corporate language executives once used in describing staff cuts is giving way to blunt boasts about ever-shrinking workforces. Gone are the days when trimming head count signaled retrenchment or trouble. Bosses are showing off to Wall Street that they are embracing artificial intelligence and serious about becoming lean.

You can read the Wall Street Article Here!

If the link doesn’t work, email me (info at aaai dot africa) and I’ll send you the PDF.

The African Academy of Artificial Intelligence (AAAI) empowers organizations across Africa to strategically integrate AI into their operations, policies, and products. Through tailored advisory, education, and practical implementation, we guide businesses, governments, and communities towards successful digital transformation.

‍