Tentacular Artificial Intelligence
[subsumes (our approach to): Multi-AI/Robot Problem-Solving]

Tentacular AI (TAI) enables artificial agents to problem-solve in ways that exploit the true potential of the Internet (I), the Internet of Things (IoT), edge computing, and cyberspace. These agents achieve a ubiquitous problem-solving power by stretching their ``tentacles’’ across heterogenous environments, sensors, effectors, and machines covering Earth and — increasingly — outer space. Six specific properties distinguish TAI agents; one of the six is that all offered solutions are accompanied by an explanation, justification, and certification of safety and/or ethical/legal correctness. When the sixth property (which asserts that the problem-solving leverages I/IoT etc.) is dropped, but the remaining five affirmed, the result is a distinctive approach to multi-robot/AI problem solving.

1. Capable of problem-solving. Whereas standard AI counts simple mappings from percepts to actions as bona fide AI, TAI agents must be capable of problem-solving — even when the problems to be solved are wholly anomalous, and therefore no relevant data of the sort that enables ML is available. This may seem like an insignificant first attribute of TAI, but a consequence that stems from this attribute should be noted: Since problem-solving entails capability across the main sub-divisions of AI, TAI agents have multi-faceted power. Problem-solving requires capability in these sub-areas of AI: planning, reasoning, learning, communicating, creativity (at least relatively simple forms thereof; Property 5, below), and — for making physical changes in physical environments — cognitive robotics. (Cognitive robotics is defined by Levesque and (2007) as a type of robotics in which all substantive actions performed by the robots are a function of the cognitive states (e.g. beliefs & intentions) of these robots. Hence, all TAI agents can plan, reason, learn, communicate; and they are creative and capable of carrying out physical actions.
2. Capable of solving at least important instances of problems that are at and/or above Turing-unsolvable problems. AI of today, when capable of solving problems, invariably achieves this success on problems that are merely algorithmically solvable and tractable (e.g., checkers, chess, Go, all of which are only EXPTIME).
3. Able to supply justification, explanation, and certification of supplied solutions, how they were arrived at, and that these solutions are safe and ethical. We thus say that the problem-solving of a TAI agent is rationalist. This label reflects the requirement that any proposed solution to the problem discovered by a TAI agent must be accompanied by a justification that defends and explains that the proposed solution is a solution, and, when appropriate, also that the solution (and indeed perhaps the process used to obtain the solution) has certain desirable properties. Minimally, the justification must include an argument or proof for the relevant conclusions. In addition, the justification must be verified, formally; we thus say that certification is provided by a TAI agent.
4. Capable of ``theory-of-mind’’ level reasoning, planning, and communication. For an introduction to this concept, see (Bringsjord, Govindarajulu, et al. 2014), available online here. In short, this level of machine intelligence entails that the AI in question has a model of the psyche of other agents.
5. Capable of creativity, minimally to the level of so-called m-creativity. Creativity in artificial agents, and the engineering thereof, has been discussed in a number of places by the Bringsjord (e.g. Bringsjord & Ferrucci 2000), but recently Bringsjord & Sen (2016) have called for a form of creativity in artificial agents using I/IoT.
6. Has ``tentacular’’ power wielded throughout I/IoT, Edge Computing, and cyberspace. This is the most important attribute possessed by TAI agents, and is reflected in the ‘T’ in ‘TAI.’ To say that such agents have tentacular problem-solving power is to say that they can perceive and act through the I/IoT and cyberspace, across the globe. TAI agents thus operate in a planet-sized, heterogeneous environment that spans the narrower, fixed environments used to define conventional, present-day AI, such as is found in (Russell & Norvig 2009).

Many technologies form the backbone of TAI, and many will be developed in its fold as the project progresses. Immediately below, technologies developed solely by RPI researchers prior to the commencement of, and during this project are highlighted in red, technologies developed solely by IBM researchers, in blue, and collaboratively developed technologies, in yellow. In addition, companies and services that in part enable/supply blue technology are colored grey.

• The paper “Toward Cognitive and Immersive Systems: Experiments in a Cognitive Microworld,” under revision for Advances in Cognitive Systems, is available here.
• TAI was publicly unveiled at the FAIM Workshop on Architectures and Evaluation for Generality, Autonomy & Progress in AI (AEGAP 2018), held in conjunction with IJCAI-ECAI 2018, AAMAS 2018, and ICML 2018, July 2018, Stockholm, Sweden (Bringsjord, Govindarajulu, et al. 2018). A preprint of the paper can be obtained here; slide deck and demo (in video form) available on the present page in the obvious categories (BibTex for this paper is below).
• We presented a discussion of the ethical and legal implications of TAI, and a method to navigate them, at the International Conference on Robot Ethics and Standards (ICRES 2018), held in Troy, New York, USA (Sen et al. 2018). A preprint of the paper may be obtained here. (BibTex for this paper is below.)
• We presented an application of TAI to smart cities, at the European Conference on Ambient Intelligence (AMI 2018), held in Larnaca, Cyprus (Sen et al. 2018). The paper may be obtained here. (BibTex for this paper is below.)

• A poster summarizing the theory and applications of TAI may be found here. This was presented at the AI Research Collaboration (AIRC) Poster Social, February 21 2019, at RPI.

• A final, summative presentation, wrapping up the exploratory phase of the TAI research program, can be found in Keynote form here, and in static pdf form can be obtained here.
• The presentation “Toward Smart Cities at the Mental Level via Tentacular AI (TAI) Agents,” was given at the Smart Cities Workshop at ICRES 2019, London UK, by Mike Giancola. The presentation can be found in Keynote form here, and in static pdf form here. The abstract follows.

Tentacular AI, or for short simply ‘TAI’ (rhymes with ‘tie’), is a new form of distributed, multi-agent AI. Of the six distinguishing marks of a TAI agent, one is that it’s able to recruit interconnected, subsidiary, “lesser” agents in order to achieve goals in the realm of the purely mental, in the minds of human persons. (Often this realm is said to be at the level of “Theory of Mind.”) For instance, a TAI agent might strive to bring it about that a human person for whom it works has certain knowledge, or emotions, or certain beliefs about the mental states of other people (where those other people in turn might well have particular mental states themselves). When people speak of “smart cities,” almost invariably the goals to be obtained by relevant AI agents are non-mental. For instance, in non-smart cities, car parking is chaotic, public transportation is less coördinated, energy use is wasteful, and so on; all these negative things to be rectified are indeed bad, but are in the realm of the inanimate/non-mental. Now, what might it be like for a human person, say Alfred, to live in New York City, supported by powerful TAI agents able to bring it about that Alfred has the states of mind that he seeks? This is the question explored in this presentation. The exploration is based on the availability of certain tailor-made-for-TAI computational logics, and cutting-edge automated-reasoning and automated-planning technology that brings these logics to life.

• A presentation that synthesizes both the AEGAP 2018 presentation in Sweden (see the relevant paper) on July 15 and a presentation given at a session on July 31 devoted to AIRC projects at RPI is available here in Keynote.
• For a review on Aug 17 2018, the relevant slide deck is available here in Keynote, here in pdf, and here in PowerPoint.
• The presentation at AMI 2018 may be found here.

• The capstone demo for the exploratory phase of TAI r&d, showing a blending of TAI:multi-robot problem-solving and CVQ+AJV, can be found here.
• In this demo, a TAI agent saves a family from dying by solving an anomalous problem the arises in their home. The TAI agent sits atop lesser agents from Nest, Amazon, and Apple, and shows the possibility of a level of AI that can rationalize the rather chaotic world of networked but non-interoperable devices in homes.

@book{brutus,
        Address = {Mahwah, NJ},
        Author = {S. Bringsjord and D. Ferrucci},
        Publisher = {Lawrence Erlbaum},
        Title = {Artificial Intelligence and Literary Creativity: {I}nside the Mind of Brutus, a Storytelling Machine},
        Year = {2000}}

@ARTICLE {nuclear_deterrence_logic,
 AUTHOR = {Bringsjord, S. and Govindarajulu, N.S. and Ellis, S.
           and McCarty, E. and Licato, J.},
 YEAR = 2014,
 TITLE = {{Nuclear Deterrence and the Logic of Deliberative
           Mindreading}},
 JOURNAL = {Cognitive Systems Research},
 VOLUME = 28,
 PAGES = {20-43},
 URL = {\small{http://kryten.mm.rpi.edu/SB\_NSG\_SE\_EM\_JL\_nuclear\_mindreading\_062313.pdf}}}

@ARTICLE {creative_cars,
  AUTHOR = {Selmer Bringsjord and Atriya Sen},
  TITLE = {{On Creative Self-Driving Cars:  Hire the Computational
            Logicians, Fast}},
  JOURNAL = {Applied Artificial Intelligence},
  VOLUME = 30,
  ISSUE = 8,
  PAGES = {758-786},
  YEAR = 2016,
  URL = {http://kryten.mm.rpi.edu/SB\_AS\_CreativeSelf-DrivingCars\_0323161130NY.pdf},
  NOTE = {{The URL here goes only to an uncorrected preprint.}}}

@inproceedings{Levesque07cognitiverobotics,
        Address = {Amsterdam, The Netherlands},
        Author = {Hector Levesque and Gerhard Lakemeyer},
        Booktitle = {Handbook of Knowledge Representation},
        Publisher = {Elsevier},
        Title = {{Chapter 24: Cognitive Robotics}},
        Url = {{\scriptsize http://www.cs.toronto.edu/~hector/Papers/cogrob.pdf}},
        Year = {2007},
        Bdsk-Url-1 = {%7B%5Cscriptsize%20http://www.cs.toronto.edu/~hector/Papers/cogrob.pdf%7D}}

@book{aima.third.ed,
        Address = {Upper Saddle River, NJ},
        Author = {S. Russell and P. Norvig},
        Publisher = {Prentice Hall},
        Title = {Artificial Intelligence: {A} Modern Approach},
        Year = 2009,
        NOTE = {{Third edition.}}}

@INPROCEEDINGS {tai_introduced_aegap,
  TITLE = {{Tentacular Artificial Intelligence, and the Architecture
            Thereof, Introduced}},
  AUTHOR = {Selmer Bringsjord and Naveen Sundar Govindarajulu and
            Atriya Sen and Matthew Peveler and Biplav Srivastava and
            Kartik Talamadupula},
  BOOKTITLE = {Proceedings of the Architectures and Evaluation for
               Generality, Autonomy \& Progress in AI Workshop
	       (AEGAP 2018)},
  ADDRESS = {Stockholm, Sweden},
  COMMENT = {{And IJCAI-2018 workshop.}},
  MONTH = {July},
  YEAR = 2018,
  URL = {http://kryten.mm.rpi.edu/TAI\_AEGAP2018\_cc.pdf}}

@INPROCEEDINGS {tai_ethics_icres,
  TITLE = {{For AIs, Is It Ethically/Legally Permitted That Ethical Obligations Override Legal Ones?}},
  AUTHOR = {Atriya Sen and Paul Mayol and Biplav Srivastava and
            Kartik Talamadupula and Naveen Sundar Govindarajulu and Selmer Bringsjord},
  BOOKTITLE = {Proceedings of the International Conference on Robot Ethics and Standards (ICRES 2018)},
  ADDRESS = {Troy, New York, USA},
  MONTH = {August},
  YEAR = 2018

@INPROCEEDINGS {tai_smartcities_ami,
  TITLE = {{Toward a Smart City Using Tentacular AI}},
  AUTHOR = {Atriya Sen and Selmer Bringsjord and Naveen Sundar Govindarajulu and Paul Mayol and Rikhiya Ghosh and Biplav Srivastava and Kartik Talamadupula},
  BOOKTITLE = {Proceedings of the European Conference on Ambient Intelligence (AMI 2018)},
  ADDRESS = {Larnaca, Cyprus},
  MONTH = {November},
  YEAR = 2018