Chatting About Conversational Chatbots and More, with Senior Ontologist and Author, Bill DeSmedt
NTENT: Bill, your title at NTENT is Senior Ontologist. Can you tell us a little bit about what you do in your role?
Bill DeSmedt: Principally, I’ve been engaged in enhancing NTENT’s flagship knowledge representation system by extending coverage of business-critical areas such as industry-specific taxonomies, combinatorial semantic assertions, question-answering logic, etc.
I also provide thought leadership regarding semantics of the future, human-intelligible naming conventions for knowledge identification, default reasoning principles, and core ontology construction.
NTENT: Bill, can you tell us a little bit about your technical background?
BD: I have over a quarter century’s experience in architecting the data structures, algorithms, and heuristics that enable computers to reason about targeted subject-matter domains the way humans do, and in integrating the resulting knowledge engine with lexicons, parsers, discourse analyzers, and language generators to deliver production-ready Intelligent Conversational Agents.
NTENT: I see you are hosting a session at the Business of Bots conference this coming February in San Francisco entitled “MetaLang: Introducing a Knowledge-Based Conversational Agent” — can you tell us what you will be discussing?
BD: My session will feature a presentation comparing and contrasting ELIZA-style chatbots (including so-called “neural chatbots”) with true conversational agents, and delving in detail into the bundle of technologies needed to realize the latter.
The presentation will be followed by a demonstration of several conversational agents created with my own MetaLang engine, which realizes the above technologies. Chief among the real-life examples will be an interactive resume bot, which enables hiring managers to conduct a simulated interview of a virtual job applicant.
NTENT: For businesses looking to deploy conversational agents, what are the top two things they should keep in mind?
BD: 1)- An appreciation for the complexities that lurk behind human conversational abilities generally, and what it will take to enable a machine to simulate those abilities;
2)- An appraisal of where the field stands today, where it should go next, and how to get there.
NTENT: How do you see conversational agents progressing over the next five years? What are they currently capable of and how far can they go? How long do you think it will take the technology to get there?
BD: On this topic, I don’t think I could possibly do better than to reference the schema devised by my good friend and fellow AI researcher Walid Saba of Astound.ai. Walid identifies six levels of virtual agent intelligence, as follows:
- Level 0 — Scripting: predefined input with basic pattern matching, but no actual intelligence whatsoever
- Level 1 — Basic: keyword searching with some superficial semantics
- Level 2 — Contextual: a linguistically sophisticated chatbot capable of deep semantic analysis and contextual word-sense disambiguation
- Level 3 — Adept: “human-like” dialogue for clarification/diagnostic purposes
- Level 4 — Resolution: problem resolution and learning from previous conversational history
- Level 5 — Fluent: the Holy Grail of AI — human-level language understanding including commonsense reasoning
Most chatbots out there — even those from big-name technology companies still seem to be stuck at Level 1, nowadays, with a few beginning to move into Level 2. Level 3 is probably the best we can hope for over the next five years.
NTENT: As we get ready to start 2019, are there any trends or predictions you think we should look out for in 2019?
BD: I think, and I’m not alone in this, that chatbots may be in for a rough ride in 2019 unless vendors make a concerted effort to move up the ladder from Level 1. The reason is simple: even just a few so-called “bad bot” experiences can lead customers to shun such interactions.
NTENT: I know you are an avid reader and a published author of both fiction and non-fiction books– can you tell us a little bit about your book Singularity?
BD: Sure, here’s the elevator pitch:
What if the cataclysmic Tunguska Event of 1908 was caused, not by a meteor or comet, but by the impact of a submicroscopic black hole? What if that fantastic object — smaller than an atom, more massive than a mountain, older than the stars — is still down there, orbiting deep within the solid body of the earth, slowly consuming the planet? What if only a rookie government agent and a brilliant but borderline consultant stand between a renegade Russian oligarch and his plans to use the black hole to rewrite history — or end it?
NTENT: How much of your fictional work plays out in our current reality or potential future?
BD: Well, considering both my books to date (Singularity and Dualism) have tended toward the apocalyptic — a submicroscopic black hole slowly consuming the entire planet, a rogue collective consciousness assimilating the minds of every human being on earth — let’s hope not too much!