Timnit Gebru Believes There Is No'Existential Threat' From AI One of AI's fiercest critics believes the doom talk is about founders making money, not saving humanity. As the industry booms, an entirely new vernacular has emerged in the debates over how consequential artificial intelligence really is. Within that new parlance, two phrases have become focal points: stochastic parrots and existential risk . And at the center of is one prominent AI researcher who has never stood down from a fight, Timnit Gebru . Gebru came into the public eye several years ago after sparring with Google over a research paper she coauthored that called out biases in the company's AI, saying that LLMs basically parroted their training data and risked perpetuating biased viewpoints. The contested paper resulted in Gebru's departure from the company, and led her to found an institute that investigates harms perpetuated by technology and supports the creation of unbiased tech tools. She has also authored a new book,, expected to ship early next year. More recently, Gebru has spoken out against a faction of the industry that believes AI is so powerful it could destroy humanity. In turn, some members of that community, including an Anthropic cofounder, have alleged that Gebru's earlier research about stochastic parrots is no longer relevant, and that AI does have the ability to "think" or reason. The AI debate is no longer just about the technology itself, but about ideological groups and strategic narratives, and Gebru believes these narratives are a "harmful distraction" from the real issues with AI. I recently spoke with Gebru about what she believes these arguments are distracting from, and also got her response to critiques that her earlier research underestimates the AI of today.

Why Is Everyone in Silicon Valley Talking Like That? Did you mix something up? Silicon Valley has long had its own dialect. Venture capitalists are always talking about the value of being "high agency" and making "orthogonal bets." But lately, tech vernacular has taken a peculiar turn--people have started describing themselves as if they were chatbots.

Large language models (LLMs) are excellent in-context learners. However, the sensitivity of data contained in prompts raises privacy concerns. Our work first shows that these concerns are valid: we instantiate a simple but highly effective membership inference attack against the data used to prompt LLMs. To address this vulnerability, one could forego prompting and resort to fine-tuning LLMs with known algorithms for private gradient descent. However, this comes at the expense of the practicality and efficiency offered by prompting. Therefore, we propose to privately learn to prompt.

Large Language Models (LLMs) can be understood as Collective Knowledge (CK): a condensation of human cultural and technical output, whose apparent intelligence emerges in dialogue. This perspective article, drawing on extended interaction with ChatGPT-4, postulates differential response modes that plausibly trace their origin to distinct model subnetworks. It argues that CK has no persistent internal state or ``spine'': it drifts, it complies, and its behaviour is shaped by the user and by fine-tuning. It develops the notion of co-augmentation, in which human judgement and CK's representational reach jointly produce forms of analysis that neither could generate alone. Finally, it suggests that CK offers a tractable object for neuroscience: unlike biological brains, these systems expose their architecture, training history, and activation dynamics, making the human--CK loop itself an experimental target.

In this position paper we raise critical awareness of a realistic view of LLM capabilities that eschews extreme alternative views that LLMs are either 'stochastic parrots' or in possession of 'emergent' advanced reasoning capabilities, which, due to their unpredictable emergence, constitute an existential threat. Our middle-ground view is that LLMs extrapolate from priors from their training data while using context to guide the model to the appropriate priors; we call this "context-directed extrapolation." Specifically, this context direction is achieved through examples in base models, leading to in-context learning, while instruction tuning allows LLMs to perform similarly based on prompts rather than explicit examples. Under this view, substantiated though existing literature, while reasoning capabilities go well beyond stochastic parroting, such capabilities are predictable, controllable, not indicative of advanced reasoning akin to high-level cognitive capabilities in humans, and not infinitely scalable with additional training. As a result, fears of uncontrollable emergence of agency are allayed, while research advances are appropriately refocused on the processes of context-directed extrapolation and how this interacts with training data to produce valuable capabilities in LLMs. Future work can therefore explore alternative augmenting techniques that do not rely on inherent advanced reasoning in LLMs.

Large language models (LLMs) are excellent in-context learners. However, the sensitivity of data contained in prompts raises privacy concerns. Our work first shows that these concerns are valid: we instantiate a simple but highly effective membership inference attack against the data used to prompt LLMs. To address this vulnerability, one could forego prompting and resort to fine-tuning LLMs with known algorithms for private gradient descent. However, this comes at the expense of the practicality and efficiency offered by prompting. Therefore, we propose to privately learn to prompt.

This week, OpenAI launched what its chief executive, Sam Altman, called "the smartest model in the world"--a generative-AI program whose capabilities are supposedly far greater, and more closely approximate how humans think, than those of any such software preceding it. The start-up has been building toward this moment since September 12, a day that, in OpenAI's telling, set the world on a new path toward superintelligence. That was when the company previewed early versions of a series of AI models, known as o1, constructed with novel methods that the start-up believes will propel its programs to unseen heights. Mark Chen, then OpenAI's vice president of research, told me a few days later that o1 is fundamentally different from the standard ChatGPT because it can "reason," a hallmark of human intelligence. Shortly thereafter, Altman pronounced "the dawn of the Intelligence Age," in which AI helps humankind fix the climate and colonize space. As of yesterday afternoon, the start-up has released the first complete version of o1, with fully fledged reasoning powers, to the public.

In 2021, linguist Emily Bender and computer scientist Timnit Gebru published a paper that described the then-nascent field of language models as one of "stochastic parrots". A language model, they wrote, "is a system for haphazardly stitching together sequences of linguistic forms it has observed in its vast training data, according to probabilistic information about how they combine, but without any reference to meaning." AI can still get better, even if it is a stochastic parrot, because the more training data it has, the better it will seem. But does something like ChatGPT actually display anything like intelligence, reasoning, or thought? Or is it simply, at ever-increasing scales, "haphazardly stitching together sequences of linguistic forms"?

With the rapid development of artificial intelligence (AI), large language models (LLMs) have shown strong capabilities in natural language understanding, reasoning, and generation, attracting amounts of research interest in applying LLMs to health and medicine. Critical care medicine (CCM) provides diagnosis and treatment for critically ill patients who often require intensive monitoring and interventions in intensive care units (ICUs). Can LLMs be applied to CCM? Are LLMs just like stochastic parrots or ICU experts in assisting clinical decision-making? This scoping review aims to provide a panoramic portrait of the application of LLMs in CCM. Literature in seven databases, including PubMed, Embase, Scopus, Web of Science, CINAHL, IEEE Xplore, and ACM Digital Library, were searched from January 1, 2019, to June 10, 2024. Peer-reviewed journal and conference articles that discussed the application of LLMs in critical care settings were included. From an initial 619 articles, 24 were selected for final review. This review grouped applications of LLMs in CCM into three categories: clinical decision support, medical documentation and reporting, and medical education and doctor-patient communication. LLMs have advantages in handling unstructured data and do not require manual feature engineering. Meanwhile, applying LLMs to CCM faces challenges, including hallucinations, poor interpretability, bias and alignment challenges, and privacy and ethics issues. Future research should enhance model reliability and interpretability, integrate up-to-date medical knowledge, and strengthen privacy and ethical guidelines. As LLMs evolve, they could become key tools in CCM to help improve patient outcomes and optimize healthcare delivery. This study is the first review of LLMs in CCM, aiding researchers, clinicians, and policymakers to understand the current status and future potentials of LLMs in CCM.

Angie Wang illustrates her toddler’s language-acquisition process, and how it compares with the learning process of large language models.