The Future of AI

We can examine the future of AI from two perspectives: the current state of Narrow AI, including its development, penetration, and adoption, and the potential technological advances leading to General AI and Super-Intelligent AI.

Current State of Narrow AI: Development, Penetration, and Adoption

It is frequently reported that over 50 per cent of current jobs will require reskilling and upskilling within the next five years, largely due to the impact of technological advancements and especially those driven by AI. But moments like this are not new in human history. We have repeatedly encountered new tools and technologies that challenge the way we do things. The Narrow AI we know today may be able to replace many manual processes but cannot replace human judgment or possess cognitive, empathic, or common-sense skills. Let’s look at this idea in some more detail.

Impact on Jobs

Certainly many jobs will be directly impacted by AI and roles driven by repetitive tasks, where patterns can be easily predicted, are most at risk. Other roles may not be significantly affected but will still require some level of upskilling.

For instance, consider a doctor who currently relies on a nurse for pre-diagnoses and an assistant for managing schedules, payments, and agendas. Some of these supporting roles could be replaced by a virtual assistant. As a result, the doctor would need to learn how to effectively “prompt” the virtual assistant, an example of necessary upskilling.

Jobs involving repetitive tasks with predictable outcomes may be entirely replaced by AI, requiring those in such roles to undergo reskilling. The good news is that new jobs will emerge, creating opportunities for those willing to adapt and learn.

As we’ve seen during other disruptive periods in human history, technological evolution brings both challenges and opportunities. While AI will undoubtedly reshape the workforce, it also opens the door to innovation, new roles, and advancements that benefit society as a whole.

There is no doubt that AI is impacting and will continue impacting many aspects of our society. It is happening every day in our smartphones, virtual assistants (Alexa, Siri), and recommendation systems (Netflix, YouTube). We see it in the following areas:

Healthcare, through apps in diagnosis, personalised medicine, research on drugs to be prescripted and operational healthcare robots.

Finance, with investment co-pilots, fraud detection, risk management, and in customer services.

Industry, with predictive maintenance, automated factories, and supply chain management.

Autonomous systems, with self-driving cars, drones, and robots.

Entertainment and media, where AI is used in content creation (e.g., music, art, writing), gaming, and advertising.

Technological Future of AI

Now let’s discuss the future of AI, focusing on technological advancements that could lead to the next stages: General AI and Super-Intelligent AI. At the beginning of this module, we discussed what these future systems are expected to achieve, and we can easily extrapolate how current AI capabilities might evolve. However, contrary to the claims of some tech gurus, the consensus is that it will likely take 20 to 30 years for General AI to become a reality.

Several fundamental challenges remain unsolved in current Narrow AI. For a model to reason and possess cognitive capacities, significant advancements are required. Current AI models are primarily driven by correlations, not causation. True autonomy in AI would require systems that can define their own goals, evaluate outputs and progress, and redefine objectives, all while being aware of their surroundings. We have discussed these concepts in this module, and it’s clear that we are still far from achieving such capabilities.

In addition to these challenges, there’s the issue of computing power. General AI would require vastly improved computational capacity to handle its advanced functionalities compared to Narrow AI. While quantum computers are under development, the question remains: When will they become functional beyond research centers, and at what cost?

Now, let’s assume that all these obstacles are overcome, and it’s clear that there’s currently massive investment in new data centres and more computing power. Safely deploying General AI and ensuring alignment with human values presents yet another significant challenge. The ethical and security implications will require extensive discussions, rules, and agreements to guide its implementation.

And finally: can you even imagine the implications of Super-Intelligent AI?