All the news right now is about Artificial Intelligence (AI). We have entered the AI Hype period. It is something like the internet in the 1990s, when the hype and the fear were all over the place, and it seemed that the internet was going to replace all our jobs.
The internet didn’t take our jobs, but it certainly did change them. AI is going to change our jobs too, in ways that are hard to see right now. The question on everybody’s minds right now is “Will AI replace relocation services? And when?”
Benefits of AI
We already benefit from AI in many ways that we don’t even realize. From Siri and Alexa to robot vacuum cleaners and cancer screening, AI has been gradually entering our lives. A common example is that when you write a text message, the keyboard suggests the word that you might be trying to write or the word that you might use next.
The new development that has everyone’s attention is ChatGTP, the chatbot that was released in November and the OpenAI technology that is being integrated into Microsoft’s Bing. This hype is going to continue with the further integration of AI into other Microsoft products, like Office, and into other systems that we use.
But let’s step back and ask what we mean by AI. Artificial Intelligence at this point mainly means for us two things: machine learning and generative processing.
How Does AI Work?
The revolution of machine learning was to get a system to look at a large amount of data in order to identify patterns. For example, you could feed the system 1000 or even 5000 CVs (resumés), and tell it which applicants you eventually hired. Based on that, the AI could predict which of the next 5000 would be good applicants, based on previous patterns.
This is called machine learning. Instead of telling the system what to look for, it looks at the historical data and figures out the patterns.
That is great if it helps you to filter 5000 applicants to find the best candidates. But how was that result reached? The truth is that we often don’t know. This is often called a “black box” because we can’t see inside the machine to understand how it arrived at its output.
In a famous case, researchers discovered that the system identified malignant skin lesions by the fact that there was a measuring stick next to them. Why? Because in the training data, skin lesions that were malignant often had a measuring stick next to them.
The point is that the training data represents the patterns and biases of the past. If a company tended to hire men into IT roles in the past, then the AI will likely flag men for IT positions in the future. It won’t select men on purpose, but it might give more weight to applicants that played football (soccer). In this way, AI can reflect the best and the worst of past patterns in our society.
In generative processing, the dataset is now basically all of the information on the internet. This is the difference between what you might hear as weak AI, which is meant for a narrow task, like cancer screening, and the beginnings of stronger, more general Artificial Intelligence.
An AI tool like ChatGTP looks at all the information that has been written about a subject and tries to produce text that answers a question by looking at past patterns and putting together sentences that try to match human language. It can even write decent computer code and solve math problems.
This is a huge advance. You can ask it, “How can poverty be solved?” and it can provide a couple of paragraphs about solutions proposed by the United Nations, World Vision, and other institutions. But this shows that it doesn’t really bring any new ideas to the discussion. It only presents current ideas in a form of human language rather than simply showing you the top search results.
AI in Relocation
Relocation isn’t about writing code, solving math problems, or researching an essay topic. Relocation is about people and their needs. This means that AI has some very large limitations in the area of relocation.
The great weaknesses of AI at this point are that it doesn’t understand, it doesn’t empathize, and it doesn’t listen.
Relocation is about a specific person or family from a specific place with specific needs moving to another specific place. It is not about large datasets or generalized information.
If I ask, “What is it like to live in Berlin”, then the system will give me a couple of paragraphs about living in Berlin, and even some tips about living as an expat in Berlin. This is because the AI is really just synthesizing search results and using an advanced form of autocorrect to put the results into sentences and paragraphs.
The AI has limited knowledge about the world. In fact, it has no true knowledge about the world. It only has “facts” that may or may not be true and that reflect the cultural bias that the internet has in general. In this case the system assumed that I was and expat rather than a student or a refugee.
Because of this, AI is utterly incapable of empathizing. It has not lived a life with experiences and relationships. Humans have the amazing capability to put themselves “in another person’s shoes”, as we say in English, or “in another person’s skin”.
I can understand something about who you are, what you might be feeling, and what your concerns might be. I can try to understand how cultural and social norms might be affecting what you are saying or the expression on your face. But I can also understand that I probably don’t know all of your feelings, concerns, or worries.
AI cannot relate back to its own experience to a time when it felt lost or lonely, unsure or frustrated. No doubt, systems will become better at trying to read voice patterns and micro expressions to deduce a person’s emotional state, but this is far from empathy.
More important, however, is that the AI doesn’t listen. It only answers. It is like the person that has all the answers but doesn’t take the time to understand the question.
We have all learned the value of listening. Listening uses empathy with the other person to try to understand what they are not saying. That is why one of the most important strengths of a person in relocation is to listen actively.
You can look at all of the information about a person to try to understand what their concerns might be and then ask relevant questions. Are they concerned about their children, or practicing their faith, or maybe their lifestyle in a new culture and location?
People could, of course, construct their search queries more carefully and less ambiguously, but that’s been the problem all along hasn’t it? We, and those we support, haven’t been waiting and hoping for Google or Bing to produce longer, better written results. What we have truly wished for is that they really understand our searches, who we are, and what we really wanted to know.
People already rely on apps and online information to relocate. They run the risk that they will make mistakes or forget something important. AI doesn’t change that fact. Worse, it can obscure the complexity and the often unwritten rules of complicated processes. Irene Bunt at Settle Service has already written about how many answers are incorrect or incomplete.
AI will become part of the tools that we use. Matching properties to renters’ requirements could become quicker and more better. Applications for immigration and other administration will no doubt become more efficient.
At some point AI might even help us understand patterns of behavior, characteristics, and past choices to help us better support transferees. But AI will not replace the understanding, experience, relevance, and value that people add to the relocation process.