Benjamin Harack

Gioele Barabucci, Victor Shia, Eugene Chu et al.

Background: Large language models (LLMs) such as OpenAI's GPT-4 or Google's PaLM 2 are proposed as viable diagnostic support tools or even spoken of as replacements for "curbside consults". However, even LLMs specifically trained on medical topics may lack sufficient diagnostic accuracy for real-life applications. Methods: Using collective intelligence methods and a dataset of 200 clinical vignettes of real-life cases, we assessed and compared the accuracy of differential diagnoses obtained by asking individual commercial LLMs (OpenAI GPT-4, Google PaLM 2, Cohere Command, Meta Llama 2) against the accuracy of differential diagnoses synthesized by aggregating responses from combinations of the same LLMs. Results: We find that aggregating responses from multiple, various LLMs leads to more accurate differential diagnoses (average accuracy for 3 LLMs: $75.3\%\pm 1.6pp$) compared to the differential diagnoses produced by single LLMs (average accuracy for single LLMs: $59.0\%\pm 6.1pp$). Discussion: The use of collective intelligence methods to synthesize differential diagnoses combining the responses of different LLMs achieves two of the necessary steps towards advancing acceptance of LLMs as a diagnostic support tool: (1) demonstrate high diagnostic accuracy and (2) eliminate dependence on a single commercial vendor.

N. Zöller, J. Berger, I. Lin et al.

Artificial intelligence systems, particularly large language models (LLMs), are increasingly being employed in high-stakes decisions that impact both individuals and society at large, often without adequate safeguards to ensure safety, quality, and equity. Yet LLMs hallucinate, lack common sense, and are biased - shortcomings that may reflect LLMs' inherent limitations and thus may not be remedied by more sophisticated architectures, more data, or more human feedback. Relying solely on LLMs for complex, high-stakes decisions is therefore problematic. Here we present a hybrid collective intelligence system that mitigates these risks by leveraging the complementary strengths of human experience and the vast information processed by LLMs. We apply our method to open-ended medical diagnostics, combining 40,762 differential diagnoses made by physicians with the diagnoses of five state-of-the art LLMs across 2,133 medical cases. We show that hybrid collectives of physicians and LLMs outperform both single physicians and physician collectives, as well as single LLMs and LLM ensembles. This result holds across a range of medical specialties and professional experience, and can be attributed to humans' and LLMs' complementary contributions that lead to different kinds of errors. Our approach highlights the potential for collective human and machine intelligence to improve accuracy in complex, open-ended domains like medical diagnostics.

Paolo Bova, Jonas Emanuel Müller, Benjamin Harack

Society could soon see transformative artificial intelligence (TAI). Models of competition for TAI show firms face strong competitive pressure to deploy TAI systems before they are safe. This paper explores a proposed solution to this problem, a Windfall Clause, where developers commit to donating a significant portion of any eventual extremely large profits to good causes. However, a key challenge for a Windfall Clause is that firms must have reason to join one. Firms must also believe these commitments are credible. We extend a model of TAI competition with a Windfall Clause to show how firms and policymakers can design a Windfall Clause which overcomes these challenges. Encouragingly, firms benefit from joining a Windfall Clause under a wide range of scenarios. We also find that firms join the Windfall Clause more often when the competition is more dangerous. Even when firms learn each other's capabilities, firms rarely wish to withdraw their support for the Windfall Clause. These three findings strengthen the case for using a Windfall Clause to promote the safe development of TAI.