The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation

Brundage, Miles; Avin, Shahar; Clark, Jack; Toner, Helen; Eckersley, Peter; Garfinkel, Ben; Dafoe, Allan; Scharre, Paul; Zeitzoff, Thomas; Filar, Bobby; Anderson, Hyrum S.; Roff, Heather M.; Allen, Gregory C.; Steinhardt, Jacob; Flynn, Carrick; hÉigeartaigh, Seán Ó; Beard, Simon; Belfield, Haydn; Farquhar, Sebastian; Lyle, Clare; Crootof, Rebecca; Evans, Owain; Page, Michael Le; Bryson, Joanna J.; Yampolskiy, Roman V.; Amodei, Dario

doi:10.17863/cam.22520

Cited by 38 publications

(8 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Designing algorithms to nudge collective behaviour may add to an already long list of ethical concerns in AI [47,49,66,67]. Our results further suggest that even algorithms that could objectively improve human performance might be limited.…”

Section: Discussionmentioning

confidence: 80%

Hybrid social learning in human-algorithm cultural transmission

Brinkmann

Gezerli

Kleist

et al. 2022

Phil. Trans. R. Soc. A.

View full text Add to dashboard Cite

Humans are impressive social learners. Researchers of cultural evolution have studied the many biases shaping cultural transmission by selecting who we copy from and what we copy. One hypothesis is that with the advent of superhuman algorithms a hybrid type of cultural transmission, namely from algorithms to humans, may have long-lasting effects on human culture. We suggest that algorithms might show (either by learning or by design) different behaviours, biases and problem-solving abilities than their human counterparts. In turn, algorithmic-human hybrid problem solving could foster better decisions in environments where diversity in problem-solving strategies is beneficial. This study asks whether algorithms with complementary biases to humans can boost performance in a carefully controlled planning task, and whether humans further transmit algorithmic behaviours to other humans. We conducted a large behavioural study and an agent-based simulation to test the performance of transmission chains with human and algorithmic players. We show that the algorithm boosts the performance of immediately following participants but this gain is quickly lost for participants further down the chain. Our findings suggest that algorithms can improve performance, but human bias may hinder algorithmic solutions from being preserved. This article is part of the theme issue ‘Emergent phenomena in complex physical and socio-technical systems: from cells to societies’.

show abstract

Section: Discussionmentioning

confidence: 80%

Hybrid social learning in human-algorithm cultural transmission

Brinkmann

Gezerli

Kleist

et al. 2022

Phil. Trans. R. Soc. A.

View full text Add to dashboard Cite

show abstract

“…The use of such automated tools and bots enables attacks to be carried out at scale and with greater speed and precision than possible using just human attackers. Artificial intelligence systems are also increasingly being leveraged by hackers to develop more sophisticated attacks that are harder to detect and defend against (Brundage et al, 2018).…”

Section: State-sponsored Cyber Economic Espionagementioning

confidence: 99%

“…Cloud computing allows users to access and store data over the Internet, rather than on local computers or servers. Hackers that target these cloud storage systems can conduct espionage operations at scale as they can target and gain access to large amounts of data with a single attack (Brundage et al, 2018).…”

Section: State-sponsored Cyber Economic Espionagementioning

confidence: 99%

Who spies on whom? Unravelling the puzzle of state-sponsored cyber economic espionage

Akoto

2024

Journal of Peace Research

View full text Add to dashboard Cite

Traditional conceptions of state-sponsored cyber economic espionage suggest that countries with different product profiles should experience high levels of espionage between them. However, this is not what we observe empirically. Incidence of economic espionage tends to be prevalent between countries with similar product and manufacturing profiles. This suggests that we may be missing critical parts of what drives state-sponsored cyber economic espionage. To help unravel this puzzle, I develop a novel theoretical framework that proposes that because attackers seek to maximize the expected utility of stolen information, they target countries that possess similar productive capabilities as themselves. Consequently, countries with dissimilar product profiles should avoid targeting each other for espionage. I test this argument using data on states’ product complexities and cyber economic espionage for a global sample of countries in a dyadic analytical framework. The results robustly show that for any pair of countries, as the complexities of their products diverge, they become significantly less likely to aim espionage attempts at each other. This study thus contributes new insights to explain why cyber economic espionage appears restricted to only a small number of advanced economies. It also illustrates the utility of large-N dyadic approaches in studying state-sponsored cyber espionage.

show abstract

“…(Jordan, Mitchell, 2015). Of course, the benefits for optimizing work processes and business productivity have been positively influenced, but on the other hand in the scientific environment there have been concerns that Artificial Intelligence would endanger jobs (Nature editorial, October 2017) or could be used as malicious tendencies (Brundage et al, 2018), or would exclude liability as there is a risk of inadvertently disseminating prejudices that may undermine the idea of fairness (Zou and Schiebinger, 2018). Other studies aimed to identify parameters to assess the ethics of Artificial Intelligence (Cath et.…”

Section: Ethical Perspectives On Artificial Intelligencementioning

confidence: 99%

Untitled

2021

Review of Economic and Business Studies

View full text Add to dashboard Cite

In a sample of 79 districts of the Slovak Republic for the period 1997 -2019, we examine the connection between the unemployment rate and the crude divorce rate. Our findings reveal a negative and statistically significant relationship, which suggests a procyclical nature in terms of the economic cycle. The estimated coefficients are relatively consistent even after the addition of control variables considering socio -economic factors or fixed effects. The results of the regression analysis show that an increase in the unemployment rate by one percentage point is accompanied by a decrease in the crude divorce rate by an average of 0.024 units. These findings support the assumptions of the theory of the high costs of divorce that spouses must bear. In a period of high unemployment, it is more acceptable for spouses to stay married than to bear the high costs associated with legal proceedings or the loss of part of the joint savings. The analysis of possible non-linear relationships between the unemployment rate and the crude divorce rate also showed a negative relationship. Quantile regression also points to the U-shaped relationship between variables and supports the opportunity theory. ii

show abstract

The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation

Cited by 38 publications

References 1 publication

Hybrid social learning in human-algorithm cultural transmission

Hybrid social learning in human-algorithm cultural transmission

Who spies on whom? Unravelling the puzzle of state-sponsored cyber economic espionage

Untitled

Contact Info

Product

Resources

About