"In recent discussions with CEOs, several misconceptions about AI have surfaced, leading to strategic missteps"
According to Founder of MKAI Richard Foster-Fletcher, understanding the fundamentals of data is crucial before any meaningful AI implementation can occur.
I recently interviewed Richard Foster-Fletcher, an AI advisor, author, speaker, and LinkedIn Top Voice. He is the founder of MKAI.org (Morality and Knowledge in Artificial Intelligence), an initiative dedicated to fostering AI’s responsible development and application.
Through his stewardship of the “Boundless Podcast”, Richard regularly delves into discussions about AI inclusivity and digital ethics, contributing to a more equitable technological future. His insights have illuminated lecture halls at globally renowned institutions, including the London School of Economics, University College London, Oxford University, and Imperial College London, guiding the next generation of tech leaders.
Richard publishes quite a popular newsletter on LinkedIn - AI Ethics, Work & Leadership.
Excerpts from this email interview:
Q1) It's been almost two years now since ChatGPT hit the market. Down the line, in your opinion, has gen-AI affected global employment the way it was anticipated to? If yes, why? If not, why? If it's the latter, do you see it affecting jobs on a big scale in the long run, say about 10 years from now?
Richard: It's true that it's been nearly two years since the release of ChatGPT 3.5. However, in business and enterprise terms, that’s really no time at all. We know how long it takes enterprises to adopt new technologies, considering the complexities of change management, procurement processes, and de-risking of different and new types of applications.
There were some early adopters, and some of them have made notable announcements. For instance, Klarna, a major player in the fintech industry, recently announced job reductions as they integrate AI into their operations. Anecdotally, many CEOs are inquiring about the opportunities for job displacement through AI. However, a better model to consider is the SAMR model: Substitute, Augment, Modify, and Replace.
CEOs should think specifically about each of these rather than focusing solely on replacement:
Substitute: Evaluate tasks that AI can perform as a direct substitute for human labour, freeing up employees to focus on more complex activities.
Augment: Identify areas where AI can enhance the capabilities of human workers, providing tools that improve productivity and decision-making without displacing staff.
Modify: Consider how AI can transform existing processes, making them more efficient or enabling new capabilities that were previously impossible.
Replace: Finally, determine where AI can fully take over tasks, but approach this step cautiously and ethically, ensuring that employees are transitioned smoothly to new roles or supported through reskilling initiatives.
In practical terms, SMEs, which can move much faster, have seen a significant impact on jobs. We've observed substantial changes in the freelance market, with platforms like Fiverr, Freelancer.com, and Upwork witnessing freelancers experiencing a reduced workload. Impacted jobs include translation, copywriting, research, design, and other creative services.
However, disruption doesn't have to lead to lower profits. Duolingo’s Experience with AI Integration provides a compelling example. Duolingo's share price initially dropped when ChatGPT was released, as investors were concerned about the potential competition from AI-driven language learning tools. However, Duolingo's subsequent strategic actions and innovations have contributed to its share price recovery.
One key factor in Duolingo's rebound has been the broader rollout and enhancement of its "Max" subscription tier. The company has also optimised their subscription offerings and focused on improving the Family Plan, which has boosted user retention. Analysts have noted that these improvements have made Duolingo's subscription model more attractive and sustainable, providing a solid foundation for revenue growth.
Moreover, Duolingo has started integrating AI-powered conversational experiences into its platform. This allows users to engage with the platform in a more interactive and practical way, enhancing the language learning experience. These AI-driven features, which include speaking with Duolingo characters, are seen as complementary rather than competitive to ChatGPT, offering a structured and playful learning environment that aligns with Duolingo's educational goals.
Additionally, Duolingo's access to advanced AI technologies through its partnership with OpenAI has given it an edge in developing these innovative features. This has helped boost investor confidence, as the company is perceived to be effectively leveraging AI to enhance its platform.
These strategic moves, combined with a favourable valuation after the initial drop, have led to a positive reassessment by analysts and a subsequent rise in Duolingo's share price.
In my own business, we’ve seen a reduced need for roles in marketing, HR, research, design, and legal. In some cases, we no longer require as many team members, and in others, we rely far less on external consultants. Recently, we automated significant portions of our website, saving thousands of dollars we would have otherwise spent on external consultants, all for a mere $20 a month per ChatGPT licence.
This scenario illustrates how AI is pulling wealth out of the traditional system. By scaling up, one can imagine the level of displacement possible, especially in larger companies. However, larger companies can't adopt AI as rapidly as smaller organisations.
In recent discussions with CEOs, several misconceptions about AI have surfaced, leading to strategic missteps. One common misconception is the belief that AI can be implemented without first addressing their data issues. Witnessing the potential of generative AI platforms like ChatGPT at home, many CEOs often rush to integrate similar technologies at work. However, leveraging AI effectively requires sorting out first-party and third-party data. Understanding the fundamentals of data is crucial before any meaningful AI implementation can occur.
Another misconception is that employees can be prevented from using ChatGPT. With multiple devices and remote working arrangements, it’s nearly impossible to stop them. ChatGPT is too valuable to ignore. Businesses should develop strategies that acknowledge and incorporate the inevitable use of these tools, ensuring that data shared into these models is managed securely and responsibly.
There’s a compelling argument for CEOs that employees will be more productive using generative AI through large language models. Hence, it’s vital to ensure that even basic LLMs are available to employees, where the data can be managed and protected in-house. This approach helps level up the workforce’s capabilities, even if it doesn’t necessarily compete with major players like Microsoft, Google, and Amazon.
Using generative AI is a moving feast, evolving almost daily. It requires significant skill to get anything other than mediocre, biassed, basic content from these applications. While a lot of textual content produced is basic at best, those skilled in using these tools will thrive, widening the gap between them and those who can't or won’t harness AI’s full potential.
Finally, while it is clear that predicting job impacts over a span of ten years is speculative, assuming that AI will simply create jobs is misguided.
AI is more likely to create the potential for opportunities, opportunities that will be transformed into more companies and therefore jobs with careful intervention from governments. Such intervention is necessary to ensure that more people can benefit from, leverage, and harness the power of AI to develop their ideas. Without this purposeful and strategic intervention, the spoils will only go to the existing tech companies and a few well-timed entrepreneurs. This is not enough to counter the potential job displacement over the next ten years. To me, this is urgent.
Q2) Do you think gen-AI is overhyped? And is it a bubble, as some are now proclaiming?
Richard: There’s a strong connection between this question and my previous response. Understanding the hype around artificial intelligence requires an appreciation of its practical applications and the proficiency of its users. While it’s normal for new technologies to be hyped, generative AI is different from trends like the metaverse. These AI platforms and tools are immediately usable and deliver tangible benefits.
Despite widespread scepticism from journalists and media who claim there are no real use cases for AI, I see plenty within our organisation. We’ve embraced these technologies through a willingness and robust change management process. Without such a framework, AI adoption can struggle. For instance, an article about Target supposedly implementing a large language model for their in-store employees mentioned that it wasn’t utilised. Whether this is factual or not, it highlights a significant point: expecting hourly-paid retail employees to effectively use generative AI in their daily tasks is unrealistic, leading to misconceptions that AI lacks practical applications. Reflecting on my experience at Oracle, I observed many large enterprises struggling to leverage advanced technologies despite their robust capabilities and user-friendly designs. Adopting new tools successfully is inherently challenging, and AI is no different. This makes it difficult to sift through the hype.
From a sustainability standpoint, providing a free AI service that consumes vast amounts of energy is problematic. Recently, I encountered teachers who were concerned about the £20 monthly cost of an AI service. While I understand that educators are not the highest-paid professionals, the time savings AI offers make this cost worthwhile. Personally, I would pay ten or even a hundred times more for the value AI brings to my business.
Although the long-term economic and environmental sustainability of AI services remains uncertain, the media’s current focus on hype and perceived lack of use cases misses the point. The true challenge lies in people’s ability to effectively harness AI like ChatGPT, a tool valued at $13 billion, and adapt their processes to integrate it. I firmly believe that those who don't adapt will be replaced by those who do.
Q3) How do you see the world of education pivoting because of AI? Will formal classrooms disappear after some years and turn into a one-on-one kind of learning experience, i.e. AI and 1 student?
Richard: When discussing this topic with educational leaders, principals, and CEOs of educational institutions, the common reaction is a bemused smile. They've heard predictions about the death of the classroom for decades, whether it was the advent of online learning, e-books, MOOCs (Massive Open Online Courses), or virtual reality being touted as the revolutionary change. The reality is that traditional classrooms are not going to disappear anytime soon.
Some leaders I speak to believe that, paradoxically, classrooms might become less technologically visible, as technology runs in the background, enhancing the learning environment without dominating it. The focus is on creating more time for direct interaction between teachers and students, as AI can never replicate the passion and experience a teacher brings to the classroom.
In further education institutions across the UK, the impact of a dedicated teacher is cited as the primary reason students continue their studies. This underscores that AI-generated personalised education is not on the immediate horizon. However, there are concerns about the amount of data teachers are feeding into free or low-cost AI models like ChatGPT. This data is effectively training these models on teaching methods, which raises questions about future implications in a sector that is often underfunded and inefficient.
While the notion of AI fundamentally transforming education seems distant, we must remain vigilant. History shows that disruptive forces like Amazon, Uber, or Airbnb can emerge and radically alter various sectors, including education. Teachers need to be aware that if they train AI on their teaching methods, it might eventually replicate their work to a high standard.
Despite these potential threats, the immediate opportunity lies in alleviating the workload of overburdened educators. AI can significantly reduce the time spent on creating lesson plans, curriculums, quizzes, and presentations. The multimodal capabilities of AI, such as generating music or visual content, can aid in creating engaging, varied learning materials that cater to different learning styles.
Furthermore, custom GPTs can be developed for individual homework assignments, allowing students to interact with the AI to understand tasks better, access additional learning resources, and receive feedback at multiple stages of their assignments. This iterative feedback process is far more effective than waiting for teacher assistance after each step.
In practical applications, we've seen AI being used to create simulations where students can interact with AI impersonating industry professionals, helping them understand career paths and techniques in a conversational manner.
Additionally, AI has proven especially beneficial for students whose first language is not English, aiding their communication with teachers and administrators. This use case helps English for Speakers of Other Languages (ESOL) students better engage with their educational environment, overcoming language barriers.
However, access to AI tools does not automatically translate to proficiency. Socioeconomic factors, such as the support system surrounding students, play a crucial role in their ability to utilise these tools effectively. In one college, when I suggested that students might use ChatGPT to pre-mark their homework to improve their grades, educators pointed out that many students struggle with basic literacy, highlighting the gap in AI tool proficiency.
While AI offers vast potential to enhance education, the transition will be gradual. Traditional classrooms will remain, enriched by AI technologies running in the background, supporting rather than replacing teachers. For the UK to stay competitive, it is essential to upskill our population to harness these tools effectively. Without this, we risk falling behind in the global educational landscape.
Want to feature in this “Minds And Machines” segment? Contact me.