Why does working in AI often feel ‘schizophrenic’?

The more I immerse myself in the world of AI, both personally and with the Diplo team, the more I realise that truly excelling in this field demands a unique, often conflicting blend of personal traits. I describe this ‘AI schizophrenia’ not as a clinical diagnosis, but as a metaphor for the internal tug-of-war that defines the best AI minds.

Think about it. On the one hand, you need the quintessential explorer. This is the curious, agile mind that thrives on experimentation and is unafraid to dive into the unknown. They are drawn to the bleeding edge, eager to tinker with new models, APIs, and frameworks, constantly asking, “What if we tried this?” This is the sprinter, launching into new ideas with bursts of creative energy.

But then, the very same individual, or at least the very same team, must possess the unwavering discipline of the engineer. This is the meticulous, rigorous professional who can tolerate endless iteration, refining datasets, tuning hyperparameters for often marginal gains, and painstakingly building robust validation frameworks.

They are the ones who ensure that AI outputs are not just plausible but verifiably correct and genuinely useful. They embody a certain approach akin to OCD-like traits of focusing on details, where every data point and every line of code matters. Increasing precision of AI is like a marathon, committed to seeing things through to a high-quality, reliable finish.

The tension is palpable. How can one person be both an explorer, comfortable with ambiguity, and a detail-obsessed perfectionist who demands high precision? It’s a fundamental paradox. The mind that enthusiastically embraces the chaotic exploration of novel AI applications is rarely the same one that finds joy in the methodical work of ensuring robustness and reliability.

This inherent conflict makes finding top AI talent a daunting exercise. We are searching for individuals who can tolerate the fuzziness and occasional hallucinations of AI systems, yet simultaneously possess the obsessive focus required to make those systems precise and useful. To use a medical metaphor, we need the restless curiosity of ADHD-like conditions coupled with the methodical rigour of OCD, all within a single professional.

This tension is also reflected in different professional backgrounds. For example, diplomats are conditioned to have a high tolerance for probability, as they operate in a highly uncertain world. In contrast, software developers have traditionally been trained in deterministic systems, where a specific line of code always generates the same response—a paradigm that large language models are upending.

While this blend of traits may sound counterintuitive, it’s not so far-fetched when we consider our own ‘multiple identities.’ The “me” in a family setting isn’t always the same “me” at work, meeting new business partners, or reconnecting with long-lost friends.

As finding talents with ‘AI ambidexterity’ is difficult, the most effective solution lies in strategic team formation. True power emerges when we build teams that consciously combine these seemingly contradictory personal traits and strengths. By harnessing and harvesting ‘schizophrenic’ demands in our teams, we can realise AI potentials in realistic and tangible ways.