Question: There are so many preditions about what AI will be in the future, if they're going to develope their own free will or if they will kill off humans because it is them thinking they are following out their code in the best possible way. You work very closley with AI, what do you think will actually happen? And what do you think the benifits of AI will be?

Fran Zuch answered on 6 Nov 2017:

Hi Miranda

That is a great question. It turns out that it actually is very difficult to simulate how our brain works. We are doing baby steps at the moment. AI is actually a very fancy word for algorithms. Now, what are those?

I always see them as another fancy word for mathematical equations and a way of reaching a decision. There are different options how you can reach a decision and that is usually what makes them different. Basically, what we will do, we will first train our AI based on training data and then she can go off and decide on her own based on what she learned. Later, we will then tell the AI if the decision was right or wrong. The AI can then learn and make a better decision next time.
I give you an example: In the beginning, we will mainly use the AI to answer client questions. As you probably know, when people are typing in messengers they are not always very specific, so the AI will hit some bumps when the clients are not using the correct grammar or make spelling mistakes or use local words.

Eventually, she will build up the knowledge, she will also have a copy of all previous conversations with each client and should be able to answer the questions without having to ask the client a lot about their history. This is one thing clients really hate when you talk to someone in a call centre and you have to tell different people the same thing over and over again. Well, I do 🙂
So I think this is an amazing benefit. Also, clients can ask the question anytime, on weekends and holidays and they will get an answer.

This is just one benefit of AI, it can do so much more, for example:
– Can help GPs with diagnosis of patients, if they are lost for ideas they can enter the symptoms and use the knowledge and experience of other GPs and training data
– Can help people with mental problems to interact/get help (AI might be better in reading between the lines)
– Can help lonely people to feel connected to the world (using chat bots and similar)
– and so much more…
So here is what I think needs to happen for people to trust that AI is not taking over the world:
– Transparency who you are talking to: As a client/person I want to know when I am talking to an AI and not a person
– Transparency of the Algorithm: As a company, you should show how and why decisions are being made especially so you can prove they are not biased or prejudiced
– Audit of Algorithms: Audit is a process where you check that a company is doing what they tell you they are doing, say that the meat used at McDonalds is actually meat and similar. I and some other people more involved with the industry propose this for your algorithms
– FailSafes: We are writing the software, we can build those in and should
– Government / Law oversight: There are already meeting in the UK discussing and setting our rules what robots can do and cannot, similar this needs to happen for AI as well

I do hope this makes sense. Feel free to ask me further questions if not 🙂