What is AI?
In my explorations of AI I’ve found the term “AI” is typically divided into three distinctive categories. Because of this phrases such as “The Rise of AI” don’t say much. As you’ll see below “The Rise of ANI” has already happened but in contrast “The Rise of ASI” would (will…) be scary.
Typically AI is divided into three areas:
- Artificial Narrow Intelligence
- Artificial General Intelligence
- Artificial Super Intelligence
This short intro explores these three types of AI with the aim of helping you (and me) understand which parts of our world can be altered by AI currently and which cannot.
Artificial Narrow Intelligence
Artificial Narrow Intelligence (ANI) is simply when a machine is programmed to do a task it can repeat by itself. We use ANI all day every day. Every price comparison website, every supermarket checkout, every time we unlock our phone with Face ID.
ANI is when a computer is trained to perform specified tasks. We feed it data and program it to produce certain results based on that data. The more data the better as this will improve the ANI’s accuracy. For example, the more examples an Amazon Alexa or Google Echo has of different voices asking it things, the better it will be at responding accurately. Alexa or Echo cannot process questions that they have not been programmed to understand. Similarly, they might not understand an accent that they were not trained to recognise.
Artificial General Intelligence
In contrast, Artificial General Intelligence (AGI) is human-level intelligence in machines.
I believe most people think we are talking about AGI when we say “the growth of AI”. And that’s why it seems so scary.
AGI can make assumptions and reach conclusions without analysing thousands of data points.
In theory, we are still some way from AGI, although there was recently an interesting case at Google where an engineer believes that a chatbot has become sentient…(link)
The concern is that once we reach the point of AGI there is no going back. Once computers have AGI, humans will be less able to intervene in their activities, just as we cannot control other humans.
Ideally, we’ll have strict ethical international standards before AGI becomes a reality. Potentially this will need UN-level intervention to protect the interests of humans.
ANI vs AGI – An Example
The below example from Andrew Ng (an AI OG) seriously helped me get clarity on the differences between ANI and AGI.
ANI vs AGI Example: Lung Cancer X-Rays
We can feed a computer thousands of x-rays of lungs, some healthy and some with cancer. The computer is taught to recognise the typical signs of the lungs with cancer. It can then start to identify with increasing accuracy whether an unidentified x-ray shows cancer.
This could speed up detection of common cancers where we have a lot of data.
However, ANI could not currently look at a diagram of lung cancer in a textbook and then identify it in an X-ray. ANI cannot generally jump across mediums.
In contrast, AGI could “read” a textbook explaining a medical condition it had never witnessed and diagnose it correctly. Just as a human doctor can accurately diagnose an illness they’ve never seen first-hand before.
At this stage computers left alone to detect cancer in x-rays would miss some of these rare or early cancers, so we always need a human pair of eyes to review.
However, imagine there was a universal database where every hospital across the world registered every lung x-ray. It wouldn’t take too long to improve the AI so that it became increasingly accurate.
Perhaps it could get to the point of “red-flagging” cases which are then bumped up an appointment wait-list. However, I do expect it will be a while (if ever) before humans will be open to letting computers fully diagnose our maladies.
ANI can learn to recognise patterns in lung xrays
if it’s shown 1,000s of labelled examples
ANI could not learn from a model and apply that
knowledge to an xray but AGI could
Artificial Super Intelligence
Artificial Super Intelligence (ASI) refers to the point beyond AGI where computers become more intelligent than humans.
It’s hard to say more about it as a computer with ASI will be able to do things that we cannot conceive with our tiny human brains.
At this stage, it is most useful as a concept as frankly once we are at this point, us humans most likely won’t be around to know about it.
AI & Real Estate – Initial Thoughts
I believe the lack of data in the real estate world is the reason automation (or AI processes) have not been adopted at scale yet.
A price comparison site for flights has 1,000s of flights to compare every single day. By contrast in the central London commercial real estate market, there might be 100 sales in one year. The data about those sales is held by different people and often kept private.
The only public data set in the UK is the sold price. The size, rent & physical condition are mostly kept in private records so it is currently difficult to have an accurate public record of true rental or investment value.
I do believe we’ll get to a point where AI can make tangible improvements to the real estate industry but this data point has helped me to understand why we’re not there yet…
People | Spaces | Places 