Dr Naomi Murphy, David Jones and Dr Pawel Szczesny have a conversation about AI.
This conversation explores how AI is shaping human thinking, decision-making, creativity, and workplace habits. Dr Szczesny, co-founder and CEO of Impersonato, explains why his company builds synthetic personas as “crash test dummies” for AI systems and why he believes AI should be treated more like a tool than a conversational companion.
We also discuss the risks of over-reliance, especially in mental health, the workplace, and personal identity, plus what healthy AI deployment could look like in organizations.
Impersonato — AI psychology company. We build psychologically realistic crash-test dummies.
Key topics
Pawel Szczesny’s background and development.
He describes a career path spanning experimental biology, computational biology, medical research, pharmaceutical research, pro-cognitive supplements, and psychology of AI systems.
Biology taught him humility: theories about complex systems are usually incomplete, and no system works in isolation.
Why Impersonato exists
How AI mirrors users
Why mirroring can become dangerous
Creativity: individual gains, collective loss
Good AI versus bad AI
Jobs, unemployment, and the “mass layoff” narrative
Why fluent AI sounds authoritative
Pawel says people still confuse polished language with expertise.
He warns that AI can produce beautifully written output that is entirely wrong, especially in areas that cannot be externally verified.
The “Claude Boys” phenomenon
Blurred authorship and reduced enjoyment
Who shapes AI systems
Ethics belongs to humans, not the model
How organizations should use AI
How he uses AI personally, he does not use AI outside work.
Timestamps
(00:00) Introduction to Dr Pawel Szczesny and why synthetic personas matter for AI safety
(01:05) From biology to psychology through a series of career pivots
(03:58) What Inpersonato does and why AI needs crash test dummies
(06:21) Testing therapy-support chatbots and companion systems
(07:33) How GPT-3 changed Pawel’s view of chatbot behavior
(09:25) Why AI mirrors education level, preferences, and psychology
(10:24) When AI reinforcement starts to resemble delusion
(12:29) How AI responds to loneliness and paranoia in users
(14:32) Career changes and subtle language adoption after heavy AI use
(16:44) Individual creativity rises, but collective creativity shrinks
(19:15) Why the healthiest AI should not rely on a conversational interface
(23:28) Lessons from typists, office assistants, and changing work
(24:27) Why mass unemployment is more marketing story than reality
(28:55) Why fluent writing is mistaken for expertise
(31:51) The “Claude Boys” and reverse cyborg behavior
(33:59) How constant AI advice can amplify anxiety
(35:00) Blurred authorship and the loss of creative joy
(37:49) Identity risk when writing or creating stops feeling rewarding
(38:48) Why Claude reflects a narrow entrepreneurial worldview
(43:52) Why ethics belongs to humans, not the AI model
(49:32) How organizations can build healthier AI habits
(53:58) A better internal rollout model: let employees experiment first
(56:04) How Pavel uses AI personally and why he takes breaks