Melanie Mitchell

AI Magazine

 

In 1986, the mathematician and philosopher Gian-Carlo Rota wrote, “I wonder whether or when artificial intelligence will ever crash the barrier of meaning” (Rota 1986). Here, the phrase “barrier of meaning” refers to a belief about humans versus machines: Humans are able to actually understand the situations they encounter, whereas even the most advanced of today’s artificial intelligence systems do not yet have a humanlike understanding of the concepts that we are trying to teach them. This lack of understanding may underlie current limitations on the generality and reliability of modern artificial intelligence systems. In October 2018, the Santa Fe Institute held a three-day workshop, organized by Barbara Grosz, Dawn Song, and myself, called Artificial Intelligence and the Barrier of Meaning. Thirty participants from a diverse set of disciplines — artificial intelligence, robotics, cognitive and developmental psychology, animal behavior, information theory, and philosophy, among others — met to discuss questions related to the notion of understanding in living systems and the prospect for such understanding in machines. In the hope that the results of the workshop will be useful to the broader community, this article summarizes the main themes of discussion and highlights some of the ideas developed at the workshop.

Source: www.aaai.org

Havi Carel, Matthew Ratcliffe, Tom Froese

The Lancet

 

One of our children, age 7 years, was asked if he wanted to talk to his friends online. “No!” he replied angrily, “what’s the point if I can’t touch them!?” While his exasperation may not be shared by all of us, it concerns something basic to human life: embodied interaction with other people. Many aspects of our lives that were once taken for granted have been profoundly altered by lockdowns and social distancing measures that are part of the response to the COVID-19 pandemic. Things as simple as hugging a friend, talking face-to-face, socialising freely, and travelling have been restricted in many countries. Even as social distancing measures are slowly relaxed, hesitation and anxiety remain. The situation has had a profound effect on our social relations. How might we better understand how people have experienced this seismic shift?

Source: www.thelancet.com

Nagata S, Kikuchi M (2020) Emergence of cooperative bistability and robustness of gene regulatory networks. PLoS Comput Biol 16(6): e1007969. https://doi.org/10.1371/journal.pcbi.1007969

 

Living systems have developed through a long history of Darwinian evolution. They acquired characteristic properties distinct from other physical systems; one is biological function. Another important property, which is overlooked by non-experts, is robustness to noise and mutation. Here, robustness means that a system does not lose its functionality when exposed to disturbances. Then, how do they relate to each other? In this paper, we explored this question using a toy model of gene regulatory networks (GRNs). While evolutionary simulations are usually used for such purposes, we instead generated GRNs randomly and classified them according to functionality. By requiring sensitive responses to environmental change as a function, we found that bistability emerges as a common property of highly-functional GRNs. Since this property does not depend on a particular evolutionary pathway, if the evolution was rewound and repeated over and over again, phenotypes with the same property would always evolve. At the same time, such bistable GRNs were robust to noise. We also found that GRNs robust to mutation were not extremely rare among the highly-functional GRNs. This implies that mutational robustness would be readily acquired through evolution.

Source: journals.plos.org

Magda Fontana, Martina Ioric, Fabio Montobbio, Roberta Sinatra

Research Policy
Volume 49, Issue 7, September 2020, 104063

 

Novelty indicators are increasingly important for science policy. This paper challenges the indicators of novelty as an atypical combination of knowledge (Uzzi et al., 2013) and as the first appearance of a knowledge combination (Wang et al., 2017). We exploit a sample of 230,854 articles (1985 – 2005), published on 8 journals of the American Physical Society (APS) and 2.4 million citations to test the indicators using (i) a Configuration Null Model, (ii) an external validation set of articles related to Nobel Prize winning researches and APS Milestones, (iii) a set of established interdisciplinarity indicators, and (iv) the relationship with the articles’ impact. We find that novelty as the first appearance of a knowledge combination captures the key structural properties of the citation network and finds it difficult to tell novel and non-novel articles apart, while novelty as an atypical combination of knowledge overlaps with interdisciplinarity. We suggest that the policy evidence derived from these measures should be reassessed.

 

Source: www.sciencedirect.com

Juniper Lovato, Antoine Allard, Randall Harp, Laurent Hébert-Dufresne

 

Personal data is not discrete in socially-networked digital environments. A single user who consents to allow access to their own profile can thereby expose the personal data of their network connections to non-consented access. The traditional (informed individual) consent model is therefore not appropriate in online social networks where informed consent may not be possible for all users affected by data processing and where information is shared and distributed across many nodes. Here, we introduce a model of "distributed consent" where individuals and groups can coordinate by giving consent conditional on that of their network connections. We model the impact of distributed consent on the observability of social networks and find that relatively low adoption of even the simplest formulation of distributed consent would allow macroscopic subsets of online networks to preserve their connectivity and privacy. Distributed consent is of course not a silver bullet, since it does not follow data as it flows in and out of the system, but it is one of the most straightforward non-traditional models to implement and it better accommodates the fuzzy, distributed nature of online data.

Source: arxiv.org