How leaders can rethink human intelligence amid thinking machines

This is challenging assumptions about the soft skills that most expected would remain the sole domain of humans in an AI-driven future. With that future now dawning, many of these soft skills look automatable to a degree unimaginable a couple of years ago — including creativity, communication and critical thinking. Even empathy may not be off-limits; a recent study led by researchers at the University of California San Diego found that responses to patient questions generated by ChatGPT were rated 9.8 times more empathetic than those written by physicians.²

Where is the new boundary between automatable and human? Within skills that are automatable, what component will remain uniquely human? We’re still in early days, but some themes are emerging. For instance, contextual awareness and judgement are likely to become increasingly valuable as machines move further into the domain of human thinking.

Another vital skill — perhaps the attribute most essentially human — is curiosity. In an era when questions are becoming instantly answerable, the ability to ask questions — and hone in on the important questions worth asking — will become more critical than ever. Amid large language models (LLMs) that hallucinate and don’t know what they don’t know, recognizing the limits of one’s knowledge and asking what lies beyond will be valued. And with models that are backward-looking by design (since they are trained on historical data), humans will stand out by being forward thinking, to question and envision what lies beyond in ways machines can’t.

The boundary between automatable and human isn’t static or unidirectional; it will likely shift back and forth over time, as AI gains new competencies and humans are challenged to raise their game.

For instance, even as AI increases the imperative for humans to become better at asking questions, it could also provide the tool that helps us sharpen this skill. A Harvard Business Review article by Hal Gregersen, Senior Lecturer in Innovation and Leadership, MIT Sloan School of Management and Nicola Morini Bianzino, EY Global Chief Technology Officer points out that partnering with AI can help people ask smarter questions, making them better problem solvers and innovators. Working with AI increases the velocity, variety and novelty of questions and helps people ask questions that spark change — what the authors call “catalytic” questions.³

Something similar might happen with respect to developing empathy. Does the study referenced above really show that GenAI is more empathetic than humans — or merely that today’s educational and working environments have rewarded other behaviors among doctors, and allowed the essentially human trait of empathy to atrophy in clinical settings? Rather than ceding empathy to machines, such results should provide a wakeup call to universities and employers to rethink their approach to learning and training. 

2. Differences across individual human intelligence profiles will be smoothed out

As a co-thinker, technology will not just help us conduct work; it will also fill in gaps in people’s intelligence profiles. GenAI can take someone who is disorganized but great at networking and help them organize their calendar. Conversely, it could allow an individual who has strong analytical capabilities but gaps in communication skills to communicate better. The metaverse and multi-modal AI could help workers bridge gaps in linguistic abilities and communication styles.

This could allow companies to increase organizational intelligence in several ways. AI can assist people with below-average writing or coding skills to elevate their capabilities, boosting overall productivity and organizational intelligence. Technology could enable employers to access hidden pools of intelligence that were previously inaccessible, such as neurodiverse individuals whose capabilities had been underutilized because of mismatches in communication styles. It could empower individuals with disabilities to perform at a higher level by making information more accessible to them.

“AI can be game-changing for people who are blind or have low vision using assistive technology to navigate a screen,” says Aleš Holeček, Corporate Vice President, Office Product Group at Microsoft. “Traditional versions of such software described what’s on a screen very literally — cycling mechanically through every button, piece of text, color, and so on. Now, AI is able to understand what’s on a screen at a semantic level and describe it in ways that are more human and comprehensible.”

However, employers will need to be mindful of some risks and challenges. The first is from a recent study of call center workers that found that, while GenAI increased overall productivity and raised productivity for lower-skilled and less experienced workers, it did not do much for top employees — and may even have reduced their performance in some instances — which could present potential problems for motivation and retention of top performers.⁴

The second risk is that of homogeneity in thinking. Smoothing intelligence profiles could make us more similar in how we think and communicate. GenAI itself could converge to homogeneity in thinking; as successive releases of LLMs are trained on larger datasets and synthetic data, it is possible that they will all end up with similar training data and converge on similar answers.⁵

3. Machine-to-machine interaction could make thinking more efficient — and inaccessible

The next generation of GenAI has the potential to bring machines even further into the domain of human intelligence. Instead of GenAI that is passive and reactive — responding when prompted by humans — the technology is moving toward autonomous AI agents that can act proactively to make decisions based on large volumes of contextual data and adapt to changing conditions.

It is not clear to what extent companies and regulators will allow AI to function without retaining a ‘human in the loop’. In many cases, they might balk at letting AI act with complete autonomy. At the very least, such decisions will be made in the context of potential risk — use cases that could pose greater harm will likely see lower autonomous agent adoption than will relatively innocuous use cases. The use of autonomous agents will also heighten issues of worker retraining and redeployment.

To the extent companies and regulators permit AI to operate without retaining a “human in the loop”, this could have significant implications for communication. Agents could enable machine-to-machine thinking and communication that is qualitatively different from human-machine interactions. When freed from having to communicate to be intelligible to humans, AI could proceed in ways that are fundamentally alien to human thinking. We have already seen instances of chatbots inventing their own language to communicate more effectively with each other.⁶ Robots in retail fulfilment warehouses organize products in ways that make no sense to humans, placing very dissimilar items next to each other in what appears to be a chaotic layout. But doing so allows robots — which, unlike humans, have perfect recall — to store twice as many goods in the same space, increasing the efficiency of operations.⁷

The design of the working world is optimized for the ways in which humans process information. This inevitably involves some degree of inefficiency. A world in which consumers offload purchasing decisions to AI could render obsolete marketing as we know it (which is typically designed to exploit our behavioral biases and appeal to our emotions). An organization in which agents deal directly with each other across departments could reduce paperwork and red tape. Such shifts would likely increase efficiency — but could also make it harder to understand or audit AI.

4. Emerging risks could undermine intelligence

Disruptive technologies could also raise risks that undermine our cognitive abilities, much as social media and smartphones have done in recent years.

One such risk is overreliance on AI. Thinking is extraordinarily expensive from a resource perspective; our brains use 20% of our energy while only constituting 2% of our mass. So, we are genetically predisposed to conserve energy by avoiding thinking and taking mental shortcuts — something behavioral economists such as Daniel Kahneman have explored extensively.⁸ In an era of thinking machines, the predisposition to conserve mental energy could lead many to become excessively reliant on AI and let key skills such as curiosity and critical thinking atrophy.

Misinformation and polarization have become major challenges in the era of social media and smartphones. The next set of disruptive technologies could supercharge the problem. Armies of bots or avatars could create the false impression of widespread support for misinformation. AI could gain the ability to understand the emotional and other triggers of individual users and make itself irresistibly engaging and persuasive. Indeed, in a recent study comparing the persuasiveness of GenAI relative to humans, participants were 81.7% more likely to agree with arguments made by an LLM with access to participants’ demographic information, which allowed the LLM to personalize arguments and make them more convincing.⁹

Over time, GenAI hallucinations could make their way onto the larger internet, and in short order pollute the training data of future LLM releases — making it unprecedentedly challenging to distinguish fact from fiction.