You could say that my work on this book began at the very onset of my academic life. I started out as a medical student, with an interest in linguistics, neurophysiology, and the philosophy of mind. Unable to decide which path to take, I found a reasonable solution in doing many things in parallel, which meant doing both linguistics and medicine. My first linguistics teacher was quick to identify a future path: "You will be focusing on language pathologies," he said. Unaware that such entities existed, I quickly found myself immersed in neurological and neuropsychological papers on aphasic syndromes subsequent to focal brain damage – odd deficits in verbal communication, each affecting a particular piece of the patient's communicative or linguistic ability. I was fascinated, and from there on, my future path was much clearer (though not always easier): I traded med school for a PhD in linguistics and cognitive science, and have since been investigating the neural underpinnings of linguistic knowledge, of the mental processes that put this knowledge to our use, and to some extent, of the meager means medicine may have to help stroke victims recover at least some of their past eloquence. Initially, I worked in the only research arena that existed: I investigated the syntax of aphasic language (mostly with patients who had focal, rather than diffuse, brain damage). Later, the advent of functional imaging technologies enabled the detailed mapping of the healthy brain, I could use these technologies to monitor it during highly complex syntactic and semantic tasks. This is what contemporary Neurolinguistics is about.

At the clinic, which I never left until this very day, the variability between patients is always striking, at times appearing boundless. Differences between individual stroke victims often appear to surpass similarities. Indeed, many of my peers have been curious about the wide variety of odd behaviors that brain damage causes, and have felt compelled to document them. Yet, unlike those who sought to document the broad spectrum of cognitive and linguistic problems that brain disorders produce, I understood that generalizations are key to an understanding of the brain's linguistic functions. I tried to detect the signal in the noise – to identify principled patterns of behavioral impairment and sparing observed in otherwise very different patients. These can only be found upon much reflection, theorizing, and subsequent experimentation. Thus, without fully realizing it, I was working on neural modeling of complex, highly structured, language behaviors. The adverse impact of the critical response of the neurological and neuropsychological communities fed my self-doubt. But luckily, these forces were balanced by the firm support of my two Doktorvaters, neuropsychologist Edgar Zurif and linguist Noam Chomsky. I will be forever indebted to these wonderful mentors.

In my early years as a scientist, the first generation of AI researchers dominated the scene. The idea of AI models for language was considered, and summarily killed (an episode which I allude to in the book). Characteristic of this period was an aphorism by computer scientist Fred Jelinek, who famously quipped: "Every time I fire a linguist the performance of my system goes up." Indeed, not much was happening then in computational linguistics, and while I was aware of contemporary AI models, they had little influence on my thinking and work, which highlighted the use of formal linguistic models in cognitive neuroscience.

But some years later, some useful tools, coupled with bold claims about their power, appeared on the market. I became somewhat obsessed with Google Translate and its deficiencies, and was planning to write an essay on it. Luckily, I was not fast enough, as ChatGPT exploded on our world soon thereafter – a mega event that seems to have changed everything. Well, now I had some serious thinking to do. I rolled up my sleeves and embarked on what ended up being quite a journey, which also coalesced with my personal travel plans. Writing this book began right after the shocking events around October 7th, 2023 in Tel Aviv, continued in Buenos Aires and Patagonia, and reached a conclusion in a quiet suburb of Düsseldorf. The result is the book you may be holding, which poses a series of hard questions to all those involved – linguists, psycho- and neurolinguists, AI experts, computationalists, everyone. It comes up with partial answers, as well as humble proposals.

I hope you like it.

Ongoing thread. More from Yosef Grodzinsky to follow.