Is this the path to AI personhood? (Could you sentence an AI to “Life?”)
CRIMINAL LIABILITY FOR ACTIONS OF ARTIFICIAL INTELLIGENCE: APPROACH OF RUSSIA AND CHINA
In the Era of Artificial intelligence (AI) it is necessary not only to define precisely in the national legislation the extent of protection of personal information and limits of its rational use by other people, to improve data algorithms and to create ethics committee to control risks, but also to establish precise liability (including criminal liability) for violations, related to AI agents. According to existed criminal law of Russia and criminal law of the People’s Republic of China AI crimes can be divided into three types: crimes, which can be regulated with existed criminal laws; crimes, which are regulated inadequately with existed criminal laws; crimes, which cannot be regulated with existed criminal laws. Solution of the problem of criminal liability for AI crimes should depend on capacity of the AI agent to influence on ability of a human to understand public danger of committing action and to guide his activity or omission. If a machine integrates with an individual, but it doesn’t influence on his ability to recognize or to make decisions. In this case an individual is liable to be prosecuted. If a machine influences partially on human ability to recognize or to make decisions. In this case engineers, designers and units of combination should be prosecuted according to principle of relatively strict liability. In case, when AI machine integrates with an individual and controls his ability to recognize or to make decisions, an individual should be released from criminal prosecution.
Has the pendulum swung too far?
Trial of Former Uber Executive Has Security Officials Worried About Liability for Hacks
Joe Sullivan, a former federal prosecutor, is accused of helping to cover up a security breach, a charge he denies
The federal trial of a former Uber Technologies Inc. executive over a 2016 hack has raised concerns among cybersecurity professionals about the liability they might face as they confront attackers or seek to negotiate with them.
Joseph Sullivan, the former executive, is facing criminal obstruction charges in a trial that began Wednesday in San Francisco for his role in paying hackers who claimed to have discovered a security vulnerability within Uber’s systems.
Federal prosecutors have charged Mr. Sullivan with criminal obstruction, alleging that he helped orchestrate a coverup of the security breach and sought to conceal it to avoid required disclosures.
AI will not be as friendly (or as smart) as Judge Judy!
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4206664
AI, Can You Hear Me? Promoting Procedural Due Process in Government Use of Artificial Intelligence Technologies
This Article explores the constitutional implications of algorithms, machine learning, and Artificial Intelligence (AI) in legal processes and decision-making, particularly under the Due Process Clause. Regarding Judge Henry J. Friendly’s procedural due process principles of the U.S. Constitution, decisions produced using AI appear to violate all but one or two of them. For instance, AI systems may provide the right to present evidence and notice of the proposed action, but do not provide any opportunity for meaningful cross-examination, knowledge of opposing evidence, or the true reasoning behind a decision. Notice can also be inadequate or even incomprehensible. This Article analyzes the challenges of complying with procedural due process when employing AI systems, explains constraints on computer-assisted legal decision-making, and evaluates policies for fair AI processes in other jurisdictions, including the European Union (EU) and the United Kingdom (UK). Building on existing literature, it explores the various stages in the AI development process, noting the different points at which bias may occur, thereby undermining procedural due process principles. Furthermore, it discusses the key variables at the heart of AI machine learning models and proposes a framework for responsible AI designs. Finally, this Article concludes with recommendations to promote the interests of justice in the United States as the technology develops.
People for the Ethical Treatment of Fish? Automating bias: determining outcomes by how the salmon looks?.
https://www.wageningenacademic.com/doi/abs/10.3920/978-90-8686-939-8_73
Ethics through technology – individuation of farmed salmon by facial recognition
One fundamental element in our moral duties to sentient animals, according to some central ethical approaches, is to treat them as individuals that are morally significant for their own sake. This is impossible in large-scale industrial salmon aquaculture due to the number of animals and their inaccessibility under the surface. Reducing the numbers to ensure individual care would make salmon farming economically unfeasible. Technology may provide alternative solutions. FishNet is an emerging facial recognition technology which allows caretakers to monitor behaviour and health of individual fish. We argue that FishNet may be a solution for ensuring adequate animal welfare by overcoming current obstacles to monitoring and avoid stress caused by physical interaction with humans. This surveillance can also expand our knowledge of farmed fish behaviour, physical and social needs. Thus, we may learn to perceive them as fellow creatures deserving of individual care and respect, ultimately altering the industry practices. However, the technology may serve as a deflection, covering up how individual salmon are doomed to adverse and abnormal behaviour. This may strengthen a paradigm of salmon as biomass, preventing the compassion required for moral reform, where the understanding of fish welfare is restricted to the prevention of suffering as a means to ensure quality products. Whether FishNet will contribute to meet the moral duty to recognize and treat farmed fish as individuals or not, requires reflection upon the ethical dualities of this technology, simultaneously enabling and constraining our moral perceptions and freedom to act. We will discuss the conditions for realizing the ethical potential of this technology.
I wonder how easily this translates to humans listening to misinformation? (Should we test every AI?) Could we design ‘self testing’ into our AI?
https://www.the-sun.com/tech/6158380/psychopath-ai-scientists-content-dark-web/
‘Psychopath AI’ created by scientists who fed it content from ‘darkest corners of web’
PSYCHOPATHIC AI was created by scientists who fed it dark content from the web, a resurfaced study reveals.
In 2018, MIT scientists developed an AI dubbed 'Norman', after the character Norman Bates in Alfred Hitchcock’s cult classic Psycho, per BBC.
The aim of this experiment was to see how training AI on data from "the dark corners of the net" would alter its viewpoints.
'Norman' was pumped with continuous image captions from macabre Reddit groups that share death and gore content.
And this resulted in the AI meeting traditional 'psychopath’ criteria, per psychiatrists.
… This led to insight for the MIT scientists behind 'Norman', who said that if an AI displays bias, it's not the program that's at fault. [I think it is! Bob]
"The culprit is often not the algorithm itself but the biased data that was fed into it,” the team explained.
Some headlines just catch your eye.
https://www.washingtonexaminer.com/news/justice/how-trump-fbi-raid-may-have-exposed-lawyers
‘Make Attorneys Get Attorneys’: How Trump FBI raid may have exposed MAGA lawyers
No comments:
Post a Comment