Ecosystem Graphs: The Social Footprint of Foundation Models

Abstract

Foundation models (e.g. ChatGPT, StableDiffusion) pervasively influence society, warranting immediate social attention. While the models themselves garner much attention, to accurately characterize their impact, we must consider the broader sociotechnical ecosystem. We propose Ecosystem Graphs as a documentation framework to transparently centralize knowledge of this ecosystem. Ecosystem Graphs is composed of assets (datasets, models, applications) linked together by dependencies that indicate technical (e.g. how Bing relies on GPT-4) and social (e.g. how Microsoft relies on OpenAI) relationships. To supplement the graph structure, each asset is further enriched with fine-grained metadata (e.g. the license or training emissions). We document the ecosystem extensively at this https URL. As of March 16, 2023, we annotate 262 assets (64 datasets, 128 models, 70 applications) from 63 organizations linked by 356 dependencies. We show Ecosystem Graphs functions as a powerful abstraction and interface for achieving the minimum transparency required to address myriad use cases. Therefore, we envision Ecosystem Graphs will be a community-maintained resource that provides value to stakeholders spanning AI researchers, industry professionals, social scientists, auditors and policymakers.

Industrial‐generative pre‐trained transformer for intelligent manufacturing systems

“…Foundation model (FM) [11] has brought a huge wave in academia and industry. ChatGPT, Newbing, PaLM, PanGu‐Σ [12], ERNIE Bot [13], and Alibaba Cloud's Tongyi Qianwen are leading the way to the next‐generation AGI and empowering them to more business scenarios.…”

The Impact of Foundational AI on International Trade, Services and Supply Chains in Asia

“…Foundational models such as LLMs have a number of key features. One is the capacity for transfer learning, where knowledge gained from training on one task, such as object recognition, can be applied to another task (Bommasani et al ., 2022). This means that foundational models are increasingly generalizable in that they can be used across a wide range of applications (Bommasani et al ., 2022).…”

“…One is the capacity for transfer learning, where knowledge gained from training on one task, such as object recognition, can be applied to another task (Bommasani et al ., 2022). This means that foundational models are increasingly generalizable in that they can be used across a wide range of applications (Bommasani et al ., 2022). Another key element of foundational AI models such as LLMs is that scaling the AI compute (graphics processing unit [GPU] and memory) and training data results in significant improvements in performance (Devlin et al ., 2019).…”

“…Third, ever larger data sets and exponential increases in AI compute and in the number of parameters of foundational AI models has led to new capabilities emerging as the system scales (Wei et al ., 2022). In other words, foundational AI models can develop new capabilities to perform tasks for which the AI system was not originally programmed (Bommasani et al ., 2022).…”

From rules to examples: Machine learning's type of authority

“…“Emergence,” they explain, “means that the behavior of a system is implicitly induced rather than explicitly constructed ; it is both the source of scientific excitement and anxiety about unanticipated consequences. Homogenization indicates the consolidation of methodologies for building machine learning systems across a wide range of applications ” [emphasis ours] (Bommasani et al, 2022: 3). In their implicitness and generality, these models crystallize machine learning's exemplary type of authority.…”