Artificial Intelligence (AI) through the Lens of Islamic Economics – IV

Text analysis may be useful, for instance, for a waqf organization to know what the neighborhood feels about the public goods (education, healthcare, orphan care etc.) it provides. An association of zakat organizations in a country concerned about a high rate of “donor attrition” may feel the need for a “feel-good analysis” of donors. A microfinance institution serving micro and small businesses may be interested in a “business sentiment analysis” of its clients. 

Sophia has admirers and critics. (for an introduction to Sophia, see my maiden blog in this series) Many experts who have reviewed her open-source code state that she is best categorized as a chatbot with a face. Others strongly disapprove of her over-hyped presentation and feel that she is no where close to having human-equivalent intelligence. I do not have the privilege of having any interaction with Sophia. However, I was fortunate enough to find another robo-lady to engage in a conversation. Meet Mitsuku, or Kuki to her close friends. She is a record-breaking, five-time winner of the Loebner Prize Turing Test and the world’s best conversational chatbot (according to folks like Google AI Research). She has been featured in the New York Times, WSJ,  BBC, Guardian, Wired, and Radiolab. Wearing the hat of an Islamic economist, I had the following conversation with her.  My comments are within brackets. You may skip it, if you find it boring.

So, this was a sort of Turing Test for the machine called Kiku. This test, named after Alan Turing, a famous computer scientist, is a method of inquiry in AI for determining whether or not a computer is capable of thinking like a human being. I leave it to you to judge, whether Kiku passed the test or not. I must admit that the conversation lasted longer that what I report here in the interest of brevity. In many cases, Kiku didn´t seem to understand my questions. Rather than asking for clarification (a very human approach in this situation), she responded in a way that appeared to be random answers, which kind of exposed her to be a machine. She was polite, occasionally trying to add humor. Often, she was lost and gave philosophical-sounding responses or a reply straight out of an encyclopedia, reinforcing one’s understanding that she was far from being human-like.

However, the fact remains that she and her likes can always be made to learn more, to acquire good knowledge about Islamic economics, banking and finance and learn to respect the cultural and religious sensitivities of Muslim clients through their conversation. If we want computers to talk to Muslims, they must learn a few things about what a Muslim knows as part of her/his beliefs, norms and culture, how s/he thinks, feels, and speaks in her/his natural language. This is a possibility within the domain of natural language processing (NLP).

Natural language is how we talk. How we speak in turn, is related to how we think. Making the machines understand us and to talk back to us is a very challenging idea. Natural language is highly complex because it can be loosely structured and words usually have multiple meanings. We may be playing with words or saying things that we don’t mean literally. We tend to have different accents and unique ways of uttering words. These can create a lot of ambiguities.

We can break up natural language processing into several activities. First, the machine needs to be able to recognize speech when it is spoken to. That’s converting language from its speech form to its textual form. Then the machine needs to extract meaning from that text to understand. Second, the machine needs to be able to articulate a response as a string words put together. Third, the machine may need to turn that into spoken words to synthesize speech. AI applications may include one or more of the above processes.

There are AI applications that take care of the first step, extracting text out of audio or video recordings. There are applications that go further and do an analysis of text contained in tweets or messages or emails or responses on a portal. Such an exercise may be useful, for instance, for a waqf organization to know what the neighborhood feels about the public goods (education, healthcare, orphan care etc.) it provides. An association of zakat organizations in a country concerned about a high rate of “donor attrition” may feel the need for a “feel-good analysis” of donors. A microfinance institution serving micro and small businesses may be interested in a “business sentiment analysis” of its clients. All these would require analysis of texts in the form of responses of stakeholders in various channels, e.g. tweets, emails, comments and queries at portal.

Web based chat bots, such as, Kiku go one step further. They are able to formulate a response to individual customer messages and post that back as an instant message. Automatic response systems of the type we find with a few Islamic banks are similar, except that they need to understand spoken requests in the first place. So, they need to incorporate a speech recognition element. There are of course, some basic ones that won’t talk back beyond playing some standard pre-recorded messages. They simply note what the customer issue is. There are however, the Kikus, the Siris and the Alexas that are fully interactive voice assistance systems that cover the full range of activities.

There is no doubt that the Kikus of tomorrow will be much smarter than today’s. A robo-scholar or a robo-Shariah-auditor or a robo-regulator, well-versed in Islamic law & economics as well as in national laws and regulations is a distinct possibility. As of now, however, we seem to be pretty much comfortable within the advisory domain with robo-advisors in place for Shariah-compliant investments and wealth management. The issue of accountability of such machine-advisors is an important issue that we turn to in the next blog.

(To be continued)

______________________________________

Chat with Mitsuku at https://www.pandorabots.com/mitsuku/