Veranstaltungen

This study examines whether access to generative AI (GenAI) technologies affects entrepreneurial entry and, if so, how. We propose two mechanisms for a potential positive effect: (1) an augmentation channel that pulls prospective entrepreneurs into opportunity-driven entrepreneurship as they automate various peripheral tasks, and (2) an automation channel that pushes displaced wage workers into necessity-driven entrepreneurship as firms automate their core tasks. Leveraging the sudden release of ChatGPT, which democratized public GenAI access, we exploit industry variation in GenAI exposure for the science, technology, engineering, and mathematics (STEM) workforce in a difference-in-differences design. We find that GenAI access leads to increased incorporated entrepreneurship for individuals with higher GenAI exposure. Mechanism tests support the augmentation channel and reveal important heterogeneities in who benefits more from GenAI.
 

Ansprechpartner: Daehyun Kim

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

This paper analyzes M&A patterns of R&D projects in the antidiabetics industry. For this purpose, we construct a database with all corporate individual antidiabetics R&D projects over the period 1997–2017 and add detailed information on firms’ technology dimension using patent information, next to their position in product markets. This allows us to identify the identity of targets and acquirers (who), the timing of acquisitions along the R&D process (when), and which type of R&D projects changes hands in terms of technology novelty (what). The main results can be summarized as follows. First, most of the action in M&As is in early R&D stages, still far from product markets. Second, most of the early-stage projects that change hands are high-risk/high-gain novel projects. Third, the industry leaders in the product markets are rather inactive in acquiring those novel early-stage projects. The likely acquirers of such projects are small or pipeline firms. Our results put into perspective the narrative that large incumbents acquire small targets with low-risk projects close to product launch. (joint work with J. Malek, J. Seldeslachts and R. Veugelers)

Ansprechpartnerin: Elisabeth Hofmeister

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Bei dieser Veranstaltung wird der erste umfassende Bewertungsrahmen für generative KI-Modelle (COMPL-AI) vorgestellt, der von Forschenden der ETH Zürich, INSAIT und LatticeFlow AI entwickelt wurde, um die Lücke zwischen den regulatorischen Anforderungen des EU AI Acts und den technischen Realitäten zu schließen. Mark Vero, Projektmitarbeiter und Co-Autor des Papiers, wird erläutern, wie das COMPL-AI-Rahmenwerk die Grundsätze und Anforderungen des KI-Gesetzes in konkrete, messbare technische Standards umsetzt, wobei der Schwerpunkt auf großen Sprachmodellen (LLMs) liegt. Mark wird auch die Ergebnisse der Bewertung von 12 hochmodernen LLMs vorstellen und dabei auf deren Mängel hinweisen, insbesondere in Bezug auf Robustheit, Sicherheit, Vielfalt und Fairness. Ziel der Veranstaltung ist, die Herausforderungen und Chancen einer Abstimmung der KI-Regulierung mit der technischen Umsetzung zu untersuchen – ein Beitrag zu den umfassenderen Bemühungen der EU, einschließlich der Ausarbeitung des Allgemeinen Verhaltenskodex für KI.

Zum Forschungspapier: COMPL-AI Framework: A Technical Interpretation and LLM Benchmarking Suite for the EU Artificial Intelligence Act

Über den Referenten
Co-Autor Mark Vero promoviert am Secure and Reliable Intelligent Systems Lab (SRI Lab) unter der Leitung von Prof. Martin Vechev an der ETH Zürich. Seine Forschung befasst sich mit dem Datenschutz und der Sicherheit von LLMs, wobei der Schwerpunkt auf der Aufdeckung von Sicherheitsrisiken in benutzerseitigen Anwendungen von LLMs liegt. Seine Arbeit wurde auf Top-Konferenzen und Workshops in Spotlight- und mündlichen Präsentationen vorgestellt, hat den Privacy Papers for Policymakers Award gewonnen und wurde in internationalen populären Medien vorgestellt. Vor seinem Promotionsstudium schloss er ein Masterstudium an der ETH Zürich in Elektrotechnik mit Auszeichnung ab.

Anmeldung

Initiative Max Planck Law Tech & Society

The crises of democracy in and across different societies can be articulated as crises of public spaces. In any theory of democracy, be it electoral, liberal, radical, or otherwise, the public spaces feature prominently as one of the core ingredients for democratic societies. In fact, most constitutions create, require, and/or protect public spaces in one way or another. Our expectations for public spaces are correspondingly high. We expect them to foster and form human relationships, offer everyone equal opportunities to participate, and structure and facilitate public debates, while being safe, activating, and inspiring. This begs the question: Do we expect too much?

More information 

The humanet3 project has been established as a joint research group by the Max Planck Institute for Comparative Public Law and International Law, the Max Planck Institute for Innovation and Competition, and the Max Planck Institute for Human Development. Its Principal Investigators are Prof. Armin von Bogdandy, Prof. Josef Drexl and Prof. Iyad Rahwan. The research group is led by Erik Tuchtfeld. It receives central funding from the Max Planck Society for the period from 2023 to 2026.

Following the Covid-19 pandemic, member states of the World Health Organization (WHO) are currently negotiating a Pandemic Preparedness Agreement to prepare the world for future pandemics. The draft agreement includes a mechanism for pathogen access and benefit-sharing under Article 12, which requires users of pathogens with pandemic potential, such as vaccine manufacturers, to provide fixed percentages of vaccines, therapeutics, and diagnostics to the WHO in case of a pandemic as well as annual payments.

One key challenge in the negotiations is whether access to sequence data should be limited as an enforcement mechanism for this mechanism. However, there is no research on how de-coupled approaches to benefit-sharing, that is, approaches that do not limit access to pathogen material and data, could be used to monitor compliance. Our research investigates the sequence data disclosed in patent applications as an indicator for benefit-sharing and compliance monitoring.

Using bioinformatics tools, we identify the users of nucleotide sequences from pathogens with pandemic potential. Our preliminary data suggests that 98% of patents disclosing nucleotide sequence data from pathogens with pandemic potential relate to vaccines, therapeutics, or diagnostics.  However, these patents often include sequences from various organisms, not exclusively pathogens with pandemic potential under the WHO agreement. This indicates that this potential monitoring mechanism would also need to be harmonized with other international instruments governing access and benefit-sharing such as the Convention on Biological Diversity.

The project is part of a work stream within the NIH-funded consortium Pathogen Data Network led by the Swiss Institute of Bioinformatics and part of a work package of the European Viral Outbreak Response Alliance funded by the European Union's HORIZON program.

Ansprechpartner: Peter Slowinski

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

We analyze the relationship between firms’ ability to leverage regulatory slack and their market entry strategies. Using newly constructed genomic measures of disease market similarity, we systematically document evidence of forum shopping, whereby pharmaceutical firms seek the most lenient regulatory environment for approval when drugs have multiple potential therapeutic uses. Firms seek regulatory approval in smaller disease markets to lower the costs of regulation and rely on complementary, non-regulatory pathways - in the form of unapproved, “off-label” drug use - to expand demand. Our data allow us to characterize the degree to which new technologies can exploit such opportunities, shedding light on how firms navigate regulatory environments to speed the entry of new products to market.

Ansprechpartnerin: Elisabeth Hofmeister

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

This research discusses the relationship between employee perceptions of sustainable human resource management and job satisfaction in 54 countries. The authors propose that sustainable HRM is positively associated with job satisfaction but that this relationship is moderated by employees’ identification with the organization and country-level individualism–collectivism. Thus, the authors suggest national culture functions as a second level moderator of the relationship of sustainable HRM with organizational identification on job satisfaction. Findings from the multi-level analyses using data from 14,502 employees nested within 54 countries provided support for our hypotheses, namely that employee perceptions of sustainable HRM were positively associated with job satisfaction and that this relationship was more pronounced for employees with lower levels compared to higher levels of organizational identification in individualistic rather than collectivistic countries. These findings bear important implications for both theory and practice.

This paper develops new procedures to measure environmental performance in cross-country firm-level data, explore the heterogeneity in environmental performance and its relationship with economic performance based on data from Croatia, France, Indonesia, and Lithuania. It documents extensive firm heterogeneity in carbon productivity (defined as value added per tonne of CO2 emitted) within narrowly defined industries, which significantly exceeds the extent of heterogeneity in labour productivity. On average, the 90th percentile firm is 22 times more carbon productive than the 10th percentile firm in the same industry (compared with seven times for labour productivity). This heterogeneity has important implications for aggregate emissions: raising the carbon productivity of the least productive firms to the carbon productivity of the median firm in their industry would reduce carbon emissions by 72% for the same level of output in total across the four countries. Furthermore, a growing carbon productivity dispersion is associated with a lower carbon productivity growth. Industries that are more dispersed in carbon productivity are also more dispersed in labour productivity and firms that are more carbon productive are also more labour productive, even when controlling for other factors. These correlations also hold in changes over time and suggest that structural characteristics of both firms and industries may jointly explain both economic and environmental outcomes. Firm-level regressions show that a plausibly exogenous increase in energy prices – as would be induced by a carbon tax – cause a significant fall in CO2 and an improvement in carbon productivity, without detrimental economic effects. This evidence suggests that there is a significant untapped potential of improved environmental performance and reduction in industrial emissions, and that these improvements could be achieved without affecting economic performance.

Ansprechpartner: Albert Roger

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

This paper tests for employer learning about worker ability and quantifies the role of learning in improving the allocation of talent in the labor market for computer scientists. We match the job history of over 40,000 Ph.D. computer scientists (CS) with publications and patent applications that signal their research ability. Workers who publish at CS conferences are twice as likely to move to a top tech firm in the next year as similar coworkers without a publication. Higher-quality papers are often filed as patent applications, but the fact of filing remains private information at the incumbent employer for 18 months. Authors of such papers experience a delayed increase in inter-firm and upward mobility. Without employer learning from public research records, the innovation output of early career computer scientists would drop by 16%. Disclosing patent applications one year faster would increase innovation by 1%, driven by faster positive assortative matching. 

Ansprechpartnerin: Marina Chugunova

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Monetary Incentives for Repeat Interactions: Evidence from an Online Labor Market 
Presenter: Frederike Eulitz (ISTO)
Discussant: Cheng Li (MPI)

Can platform-level changes to monetary incentives increase repeat interactions in online labor markets (OLMs)? We study this question by exploiting a change in the fee structure of the OLM Upwork, which decreased fees if the lifetime billings of a freelancer-buyer relationship surpassed thresholds. We explore the effects of incentive design on repeat interactions, which is theoretically ambiguous in traditional organizational settings, in an OLM. Analyzing a panel dataset of 24,873 freelancers, we reveal that the effects of monetary incentives are not uniform but depend on sub-group characteristics and the unique contextual factors associated with OLMs. Low-earning freelancers respond by increasing repeat interactions, whereas high-earners exhibit a surprising decline. We argue that the impact of monetary incentives depends on platform type. In knowledge-sharing platforms, such incentives may reduce engagement by crowding out intrinsic motivation. However, in platforms with marketplace features, such as OLMs, we argue that monetary incentives induce rational behavior because participants are extrinsically motivated to interact. Our findings contribute to the literatures on platform governance and incentives, offering insights into how platform-level strategies shape participant behavior, and the unintended consequences of platform incentive design.

The Role of Team Composition in Explorative Invention: Analyzing the Impact of Tenure Disparity and Team Size  
Presenter: Jisoo Hur (TUM)
Discussant: Sophia Wetzler (ISTO)

Employee career development is a dynamic process in which individuals adopt different strategies at various stages, leading to variations in their innovative behavior based on organizational tenure. Junior employees, often in the early stages of their careers, bring fresh perspectives and external knowledge, making them valuable contributors to explorative innovation. However, fostering this type of innovation can be challenging when junior inventors collaborate with senior inventors, whose extensive experience in the firm’s technological domain and greater decision-making authority may unintentionally limit juniors’ autonomy and capacity to explore new ideas. This study investigates how tenure differences between junior and senior inventors influence their likelihood of engaging in exploratory inventions. Using patent data to analyze co-inventor relationships, the findings reveal that junior inventors are more likely to engage in exploratory invention when collaborating with peers of similar tenure rather than with senior inventors. Additionally, greater tenure disparity between junior and senior inventors is associated with a decline in exploratory invention. However, larger team sizes help mitigate this negative effect by fostering a more balanced and collaborative environment. These findings provide valuable insights into the dynamics of inventor collaboration and offer practical strategies for organizations to optimize team composition and enhance exploratory innovation.

Scientific Paradigms, Graphics Processing Units and the Evolution of Artificial Intelligence
Presenter: Anna-Sophie Liebender-Luc (MPI)
Discussant: Nicole Wenger-Wong (TUM)

A sudden shift in scientific and technological paradigms lies at the heart of recent advancements in artificial intelligence (AI). Around 2012, traditional symbolic AI gave way to neural networks (NN) as the dominant approach for AI research. This coincided with the sudden successful application of graphics processing units (GPUs) as computational technology. GPUs had been invented for a different application, i.e. accelerating complex graphics displays, mostly in video games. We claim that these developments reflect the nature of breakthrough innovations and have implications for regions competing to become AI leaders. We investigate the role of expertise in GPUs for the uptake of AI innovation across regions globally. To this end, we construct a global database covering 2,088 urban areas for the period from 2000 to 2020. The data encompass a broad set of measures describing AI research and innovation activities, based on publications, patents and startups. We document the ascendancy of neural AI and its association with GPU expertise. Panel OLS and IV regressions demonstrate that after 2012 GPU- and NN-related human capital had a strong effect on the growth of AI-related patents and startups. We discuss implications for innovation policy.