DOI: 10.13140/RG.2.2.27962.93125

This article covers the areas of ethics and safety concerns for the growing field of Robotics which use Artificial Intelligence to pursue its tasks. The article covers various kinds of threats AI based Robots may pose to humans and environment. Further, ethical issues and hacking of an AI based Robot and its potential side-effects are discussed. Concluding remarks of what is suggestive to such a problem are proposed herein one among which is inclusion of a law book for entities occupying human like spaces viz. Robots.

Its time General AI ethical issues be addressed. While at the same time science fictions have general AI used well to abundance and it discusses its ethical issues too, howsoever ignored.

1.    Introduction

Robotics is a massively developing field in the current scenario of technology. Mostly every sixth organisation seem to get into it in some way directly or indirectly. The utility areas are widespread as well, robotics for autonomous vehicles, robotics for shopping, robotics for manufacturing industry to even robotics for being a human clone. All that sounds good, as long as the hired Robot know what they are intended to do and perform on. This comes in with the safety mechanism that needs to be put for any kind of Robot hacking. Robot hacking is the term coined to emphasise the capture partial or full of a Robot by an external agent. The external agent trying to hack a Robot can be a self-operating AI software, another machine or even a human manually sending instructions to Robot for accomplishing its own motivated modus operandum. The hackings may be: hardware hacking and software hacking. This article lays emphasis on the kinds of hacking and the ethics concerns that the manufacturer of robotics machines have to put in to ensure the welfare of mankind on large.

The processes followed by an AI enabled Robot are not always very transparent. A completely ethical AI agent and an AI enabled robot needs to be transparent in its workings not just in hypothetical and statistical proof of concepts. Why a black box-based AI agent not well adept to be a good Robot? The answer lies in the fact that one needs to continually ensure whether the ethical Robot has been hacked or any component in it has stopped working to its full potency.  There can be primarily two kinds of hackings a robot can be subjected to; these are given as follows:

1. Software Hackings: The software hackings are committed in the same way as any other software is hacked. For this to be clear, one has to understand that, the Robot is running on a trained software module. The software modules take instructions in form of verbal or visual inputs form environments and produce outputs. Any kind of manipulations in the software of the Robot can be dangerous, from the point of view of the sensitivity of work the robot performs.

• Hardware Hackings: The hardware hacking can be of two kinds, firstly, getting the hold of hardware itself and replacing it, secondly by an induced hardware malfunctioning with help of agents which can be another hardware or a software interrupt call.

For instance, take the manufacturing of chocolates. Any hacking can change the constituents of the chocolate the Robot is self-producing. Taking another example, take an autonomous bike, running for humans, any kind of hacking of the software or hardware can malfunction the very process of automatic driving capabilities of the bike and can harm the human which is assuming the automatic self-balancing mechanism of the automatic bike. This can harm much more than just the navigational error caused by software or hardware hackings.

There are two kinds of Robotic systems which can be developed and classified during testing phase. They are described as follows:

1. Black-Box Robotic System: The black box system verifies only the input and output of the system, validates and test the accuracy of the system on the target classification tasks provided by the dataset to be approximated by the AI based robotic agent.

• White-Box Robotic System: Works on internal details as well as code fragments that are activated given the input instructions to the robot in form of the voice, textual or visual input instructions. The white-box based system asks why, not just validating the accuracy of the system.

Software hackings are a challenge, even now, no complete hack proof software tools are fully functional in the web. Most software is susceptible to some or other kinds of hacking exposures. Hence the need for Robotic system anti-virus. This itself is an issue to be considered as one needs a white box level of analysis for Robotics not just a black-box functioning of Robotics. This is to lay emphasis on the crucial tasks the Robot has to perform.

The key ethical issues in the black-box based Robotic systems can be derived from ethical issues of AI based systems as described in [1]. These are modelled for Robotic system as follows.

1. The AI based Robotic system becomes self-operating. This primarily can happen for the following two reasons:
   1. By hacking of software or hardware units of Robots
   1. By malfunctioning of any component of Robotic machinery
2. The AI based Robotic system dictates to humans hence changing human decisions. This can be particularly harmful in case the Robot is hacked.
3. These things impact the functioning of the society on large.

2.    AI based Robotics and Ethical Issues

AI and Ethics in “Perspectives on Artificial Intelligence” by Stahl, B. C. [1, 2] lays emphasis on the following things from ethical point of view for characterization of AI:

a.           Narrow AI,

b.           General AI and

c.           The mixed systems, which use combinations of things

In the context of AI based Robot we can have an equivalent characterization of AI enabled Robotic machinery:

1.  Narrow AI for Robotics

The term is coined for Robots which use machine learning based algorithms as their Artificial Intelligence based guiding mechanisms. Several points have been presented in [1] for ethical analysis, however they have been not enough for Robotic AI systems. The points in [1] have been modelled, elaborated and expanded, here in this article for consideration for AI based Robotic systems.

1. The mechanisms and processing of Robotic AI machines should be transparent and not to be a black-box based analysis.
2. The outputs of the AI based Robots should be reproducible under similar artificial sensory inputs.
3. Test data consisting of simulations on Robot for real time, audio, video and other sensory inputs need to be accumulated.
4. Evaluations need to be performed on these sensory inputs in simulating environments for the Robotic machinery and performance need to be computed for same.
5. Non-predictivity is present due to presence of point 1 above and hence have to be eliminated form an AI based Robotic system.
6. Huge amount of training data is required and some can be generated by machines themselves to be tested on Robot for the desired outputs which are validated by humans.

2.  General AI for Robotics

General AI or General Artificial Intelligence also known as “good old-fashioned AI” (GOFAI) [1]. The General AI for Robotics relies on the basis that robotic world entities could be used to solve problems just the same way as humans solve the problems. And it is known to all that Robotics is not at complete AGI stage yet. Decades since its inception [1] not much progress has been made in AGI. The author [1] attributes it to fact that only symbols can’t represent complete problems. As per author [1] general AI cannot cause ethical issues, since it is not developed in full-fledged composure. While at the same time science fictions have general AI used well to abundance and it discusses its ethical issues too, howsoever ignored.  These pays emphasis of ethics talks in AI for robotics. Also, to add to the details, hacking needs to be considered in AGI even more. This is because once a AGI induced Robot is hacked, it becomes even more dangerous, as a competitor can use it for its best use against the maker. This cannot be easily diagnosed as well, since the Robot is hypothetically declared as AGI equipped. Hence, stringent regulations for AGI enabled Robots are the need of the hour.

• Socio-Technical Systems

These are systems that combine both the above designs, to the best possible to solve some social problem. Ethical issues in this aspect are discussed in [1], however from point of view of AI based Robotic system the following are considered:

1. The Robotic system becomes self-operating for purpose of some pre-defined perception of the maker of the Robotic system. In this case, the Robot may cause un-foreseen harm to people and other living species apart form damage to human things be them be consumables or unconsumables.
   1. The Robots starts to guides human hence changing human decisions, which is completely unethical as humans have a right to choose their decisions.
   1. These things can impact the society on large. As the law and order need to be enhanced to not just humans but also to Robots. Yes, the law book for Robots may be needed if the issues are not handled well on time.

3.    Conclusion and Future Works

The conclusions of this article lay emphasis on the need for law book for Robots. Robots occupy space like humans in society and hence become part of the law books. Robots should not be allowed to change decision making of humans by compulsive reactions rather should limit its boundaries to suggestion provider only. Further, measures should be taken to avoid Robotic hacking in any kind of situation be it AI based Robot or an AGI based Robot. The future work comprise of stringent testing of any Robotic systems for possible threats and dangers.

References

[1] Stahl, B. C. (2021). “Perspectives on Artificial Intelligence” Artificial Intelligence for a Better Future: An Ecosystem Perspective on the Ethics of AI and Emerging Digital Technologies (p. 124). Springer Nature.

[2] Stahl, B. C. (2021). “Concepts of Ethics and Their Application to AI” Artificial Intelligence for a Better Future: An Ecosystem Perspective on the Ethics of AI and Emerging Digital Technologies (p. 124). Springer Nature. [3] Joachim von Braun, Margaret S. Archer, Gregory M. Reichberg, and Marcelo Sánchez Sorond. AI, Robotics, and Humanity: Opportunities, Risks, and Implications for Ethics and Policy.  J. von Braun et al. (eds.), Robotics, AI, and Humanity, https://doi.org/10.1007/978-3-030-54173-6_1

#artificialintelligence#machinelearning#ethics#safety#law#robots#hacking