AI Society Meetup: “How Artificial Intelligence Is Breaking Finance”
AI has become a real trending topic in 2016 as this article shows, and it will even more in the coming years. With this in mind, a group of PALO IT members (always eager to learn and know more!) chose to attend the following presentation on the 24th January, 2017 : “AI Society: How Artificial Intelligence Is Breaking Finance”.
Here is the planned agenda:
- Benefits of using AI in Fintech
- Practical Examples of AI in the Financial Industries
- Real Life Demo of Financial Assistant Chatbot
- Threats of AI in the Financial Sector
- Future Outlook.
We will try to sum up what came out of it.
1/ Benefits of using AI in Fintech
In finance, the goal is to have an edge over your competitor, even if you get a 1% advantage over the other, you will make more than others companies. That is why finance companies had a calculator before other, why they had computer before most of us and why when it comes to AI, they have been using it for the last 2 decades. But to keep the edge they don’t share what they exactly do. However, there are many fields in Fintech where AI could help us. And in today’s conference, they focused on chatbots and how they can help to improve customer service? (which is known to be terrible in Hong Kong).
2/ Real Life Demo of Financial Assistant Chatbot
(The schedule was changed due to personal imperative of speakers) Claire.ai is a company that helps to build chatbots for various customer and presented us with some advantage of using the chatbots:
- Personalized experience: your chatbots can adapt to each of your customer from the data you have or from the previous conversation;
- Fewer apps: you don’t need to go through one app for each of your query, you can have only one chatbot that could connect to various services for you;
- 24/7 support: a chatbot don’t need to take a break and rest! It is always 100% efficient and it can be multilingual.
However, it was also reminded that chatbots bring their own issues:
- User experience: sometimes people get unhelpful information which leads to high degree of frustration. Especially if they weren’t warned they were interacting with a bot;
- Potential risk on security: especially for sensitive information like account number, secret code, etc.
On top of those, they gave some estimation about time to market for a chatbot projects wich is:
2 months to build a test prototype
Week 1 to 5 : Design & Development
Week 6 to 8 : Testing & Training
- Bot’s NLP is trained per your specifications;
- NLP expert develops test cases for initial validation and will tune and refine NLP to ensure 90% accuracy;
- Bot provided to User Acceptance Testing (UAT) testers;
- Bot’s NLP and other functional behaviors continually monitored by our NLP expert and bot developer.
End of Week 8 : Launch
- Both officially deployed into production environment.
- Continues to monitor logs for issues found by production users and makes corrections to the bot configurations as needed;
- Continually tunes NLP per user feedback.
They also provided a video of a demo of their chatbot for the personal banking.
3/ Practical Examples of AI in the Financial Industries
We got a quick view of the current state of AI field, reminding us of that chatbots are really a small field in AI. Currently we have massive research in machine learning, deep learning, expert systems etc. The speaker, Gerardo Salandra, also quickly reminded us of the difference between Retrieval-based model AI that is mostly used nowaday and Generative models that are more promising but has no business application so far (a definition can be found here). Then the various application we may have in Finance currently and the company that built them. The speaker chose to break them in the following categories:
AI assistant: assist the user/customers on inquiries
Kasist, Trim, Penny, Insurify, Sure.
Fraud detection: to detect fraudulent credit card transaction or fraudulent health claims
Personal banking: could be used to gather customer financial status that can be used in anticipating the financial needs or identifying insights and provide solutions and actions for the customers
Quantitative trading: could be used to help the traders to be in the best position in the stock market
Sentient technologies, Clone Algo, Alpaca, Walnut Algorithms.
Predictive analytics: trying to produce forecast from all the previously known situation
Ayasdi, Digital Reasoning, α context relevant, H2O, Kensho, Cortical.io, Numenta, Turi, Datarobot, Nervana.
Sentiment analytics: could extract sentiments from various sources (social media or more classic media) and identify opinions expressed
Indico, Acuity, Lucena, Numerai, Dataminr.
Credit storing: could offer faster and much accurate information for credit scoring. It is mainly used for credit evaluation and classify people with good or bad credits
TypeScore, αire, ADF, Wecash, Creditvidia, Zest finance, Cream finance.
4/ Threats of AI in the Financial Sector & Future Outlook
Of course, AI are currently reshaping various industries especially in finance since they can help on various tasks that are redundant. And you can easily picture yourself in a couple of years where, for example: you or your partner is pregnant. First, it is possible that your personal assistant for banking spots it, as your purchasing pattern will change, so it will guess that you or your partner is pregnant.
Knowing this information it will propose you the following: “I can guess from your buying pattern that your family will soon get wider. However, your current car is too small and unsafe for a family, you probably want to buy a new one. But as you don’t have the cash on hands you will need a loan. However, your current credit score is not that good. But if you cancel those music and tv subscription that you seldom use for 3 months, that you pay your credit card bills on the 11th of each month it will allow you to get a loan at a better interest rate. Do you want me to proceed?”
This is both awesome and also very scary. But it will probably be something happening in the near future.
Another aspect of AI that will change our world is unfortunately the amount of jobs. Since it will both destruct jobs and create them. Unfortunately for every 10 jobs it destroys it will create only 1 (the ratio may be even worse). And this is clearly a concern.
To tackle this the speaker referred us to 2 ideas:
- The first one was the universal basic income (or UBI), which would imply for states to give, without any condition, a certain amount of money to all their citizen. While some experimentation have been done and are globally positive, we have no clear long term impact of such measure on a wide population.
- The second one is regulation about AI, for that he referred us to “The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies” by Erik Brynjolfsson and Andrew McAfee.
To conclude, and this is a personal opinion, I would say that such presentation was pretty disappointing for technical persons. Whether they were in IT or in Finance and interested about AI. If you had no real idea about how AI is changing the landscape, it would probably have been much more interesting and giving you an idea of the current changes that are happening right now.
AI has been fantasised about for decades now, and it is taking an increasing role in our lives, everyone should use those events to learn more and make an idea by themselves about what is happening and how they could benefit about it.