Thoughts on Asimov’s '3 Laws of Robotics'

"The saddest aspect of life right now is that science gathers knowledge faster than society gathers wisdom." - Isaac Asimov

Isaac Asimov was an American biochemistry professor at Boston University and a science fiction writer, whose writings explored morality and interaction between AI robotics with humans. His collection of short stories were complied into one book in 1950 entitled ‘I, Robot’, which in 2004 was followed by a film entitled by the same name starring well known Hollywood actor Will Smith.

​In his writings, Asimov designed a fictional ‘3 Laws of Robotics’ for humanoids:

1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.

2. A robot must obey the order given it by human beings, except where such orders would conflict with the First Law.

3. A robot must protect its own existence, except where such protection would conflict with the First or Second Law. ​

Asimov’s laws reflect the shift of responsibility of human rights protection from the state onto the AI machinery itself. Humanoids and AI alike in real life are incapable of understanding ethics for themselves; they are a product of their creator. Thus the machines are unable to fully understand concepts such as harming a human being or rendering a human person vulnerable to harm from an AI entity through accident or instruction from another human being. The only way that the AI could understand such concepts is if they are programmed with the knowledge of what it is to be harmed - calling for a nuanced and multilateral approach to designing an agreed upon ethical code for the AI to align with. However this would also mean enabling the AI to understand the ethical code. Are we then to give it feelings, consciousness?*

What qualifies as causing harm to a human being is also a nuanced concept and can often be subjective. In the 2012 Hollywood movie ‘Robot and Frank’, the humanoid robot - (spoiler alert) - is trained to and successfully engages in breaking and entering people’s homes 

and stealing from human beings; obeying orders given to him from his human owner and unbeknownst to the robot that he is doing anything wrong. To avoid such scenarios in real life, does that mean that we should programme AI entities with international human rights law and the laws of the state in which they operate? There may be always risks, and to the best of our ability without excuses, risks must be managed; "appropriate forms of risk avoidance and diminution need to be applied, and regimes for fallback, recovery, and retribution must be established." (Clarke: 1993, 1994).

Asimov saw robotics and AI as just advanced tools, and thus believed that human beings would input safeguards into any potentially dangerous tool, (McCauley:2007). Yet currently humanoids and AI alike are being designed and implemented into society without even Asimov’s laws, and without a secure legal foundation that has experience in handling the risks that appear in AI and human interaction and how to mitigate them. Particularly when we still do not have clear accountability measures put in place.

 "(First)Controls are needed to ensure that intrinsic laws, if any, are operational at all times and that guidelines for design, development, testing, use, and maintenance are applied. Second- order control mechanisms are needed to audit first- order control mechanisms. Furthermore, those bearing legal responsibility for harm arising from the use of robotics must be clearly identified. Courtroom litigation may determine the actual amount of liability, but assigning legal responsibilities in advance will ensure that participants take due care." (Clarke, 1993, 1994)

The AI which we currently have cannot understand abstract concepts such as ‘harm’, it only understands its code and what it was programmed with the ability to do. For an AI entity “to have any real possibility of destroying humanity, even by accident, the AI had to ultimately have some physical way of causing harm.” (McCauley:2007).

The expanding ability of robots to take unsupervised decisions renders it imperative that mechanisms are in place to guarantee the safety of their behaviour, (Vanderelst, Winfield:2018). Yet to ensure the fulfilment of human rights obligations such as ensuring peoples safety and security, is it not just legislation which is required but also the capacity to realize it, (Barry, Woods:2013). This is where safeguards such as corporate social responsibility and clearly defined accountability measures would need to be put in place – with or without the 3 Laws of Robotics so to ensure peoples safety.

"Each jurisdiction, consistent with its own product liability philosophy, needs to determine who should bear the various risks. The law must be sufficiently clear so that debilitating legal battles do not leave injured parties without recourse or sap the industry of its energy. Information technologists need to communicate to legislators the importance of revising and extending the laws that assign liability for harm arising from the use of information technology." (Clarke, 1993, 1994)

In an attempt to improve the 3 Laws of Robotics, Asimov later wrote in the 'Zeroeth Law of Robotics':

0.  A robot may not harm humanity, or, by inaction, allow humanity to come to harm.

How to quantify such a concept, was something which Asimov discussed in his 1986 novel ‘Foundations and Earth’;

Trevize frowned. "How do you decide what is injurious, or not injurious, to humanity as a whole?"
​
"Precisely, sir," said Daneel. "In theory, the Zeroth Law was the answer to our problems. In practice, we could never decide. A human being is a concrete object. Injury to a person can be estimated and judged. Humanity is an abstraction."

​Through anthropomorphizing AI, it can sometimes be forgot that tools embedded with computer coding do not think in the same way n which human beings do and they do not have a subconscious mind. For Asimov's '3 Laws of Robotics' to work in practice, the AI involved would first need to have a thorough understanding of morals and methods of harm in order for it to determine when a human is in danger, when to disobey an order for it conflicting with Law 1 and when to decide to allow itself to become destroyed in order to save a human being, (Law 3). Even then, would these laws really be enough to protect all peoples in such a complex, un-coded world?

Asimov's laws are rather good and are well-intentioned. Unfortunately, it is still science fiction. And often sci-fi, (although albeit entertaining and an inspiration to this field), jumps ahead in inventions in many ways, assumes that the law and social responsibility has caught up and neglects the nuances outside of the sci-fi world. 

Although perhaps Asimov was never really intending his ‘3 Laws of Robotics’ that he designed to stretch into the real world innovations of AI to begin with, and only meant them as a starting point of ethical conversations that need to be had when designing AI that interacts with human beings?

The main issues with the laws in real life are:

– AI is currently unable to understand the connecting concepts around such laws, such as ‘harm’ and ‘humanity’ and arguably, ‘human’.

 – There is currently no secure AI focused legal system to back up Asimov's laws and offer safeguards for human beings should something go wrong.

​So if Asimov's proposed laws do not suffice past sci-fi, what exactly is the answer?

Firstly, there needs to be an established legal responsibility should something go wrong readily in place, one which has the capabilities to assert issues and failures and mitigate accordingly and fairly.

Secondly, human rights and social awareness risk assessments need to be thoroughly and pre-emptively utilized to assess any possibly dangers. Safeguarding restrictions should be implemented into the design and coding of AI so to ensure that it is not capable of causing harm to living beings in the first place.

Lastly, for Asimov's laws, or such alike, to be implemented into AI successfully, consultation from experts within a range of multi-disciplinary fields – from philosophy to social care – would need to be utilized in every step of designing a suitable ethical coding for which is relevant to the AI and is realistic in its capabilities and understanding.

*Consciousness in AI will be discussed fully at a later date in this blog.

​
Barry, Woods (2013). The Environment. In: G. M, ed., Human Rights Politics and Practice. Oxford: Oxford University.

Clarke R (1993, 1994) Asimov’s laws of robotics: Implications for information technology. Published in two parts, in IEEE Computer 26, 12 53–61 and 27, 1, 57–66

McCauley, L. (2007). AI Armageddon and the Three Laws of Robotics. Ethics and Information Technology, 9(2), pp.153-164.

Vanderelst, D. and Winfield, A. (2018). An architecture for ethical robots inspired by the simulation theory of cognition. Cognitive Systems Research, 48, pp.56-66.