Over the weekend, I was listening to This Week in Law's episode from January 31: Deep blue vs The Universe. They started the episode on the topic of the Singularity and James Miller  and Stan Liebowitz discussed how we might know the Singularity has been achieved. Each had a different view on how far into the future this event might occur and what capabilities that intelligence should possess to be considered "intelligent." Both viewpoints were informative, but I found Dr. Liebowitz' statements extremely interesting in how he felt a contrast between an artificial intelligence and a human should be used to identify when the artificial process has reached or surpassed a comparable human capability. Dr Liebowitz shared this thought:

@24:25 - I see things computers can do that people can't, but things people can do that computers can't is what we're talking about.

I should note this was in response to Dr Miller's assertion that machines can do some things better than humans. When asked by Dr Miller for an example of what a human can do that a machine cannot, he responded:

@24:40 - I say to the computer, go and perform a statistical analysis. You figure out what techniques to use to tell me...whether or not file sharing decreases record sales.

Just providing two quotes and a bit of context doesn't do the conversation justice (and it's well worth the time to listen - @48 minutes I believe) to hear the entire conversation. Listening brought me to the question of what we might use to identify where machine intelligence is in each entity's development? This led me to inquiry-based education, a method of facilitation that helps inquirers (students) work through learning experiences with different levels of support and direction.

While inquiry based education is a well-developed and nuanced field, Herron's Model is how it was originally presented in my courses and provides a simple overview of the levels of inquiry.

At the lowest level, 0, a 'recipe' is provided for the learner. Using chemistry as an example, a beginning student would have the problem/question provided, procedure/tools identified, and a solution described so the students understand when they have completed the lesson.

At level 1, the solution is not provided and the learner must identify when they have completed the lesson.

At level 2, only the problem/question is provided and the learner must identify the method and tools required to find the solution as well as presenting the solution.

At the highest level, 3, none of these are provided. An example might be a science fair project where the student  must decide the topic, procedure and then work to the solution.

At which level can we identify a machine or software is "intelligent"? If I understand Dr Miller's statements, there is 'intelligence' as long as the computer can perform at level 1 better than a human can perform at level 1. Dr Liebowitz's statements make me believe he would acknowledge a machine intelligence when it is capable of performing at level 2. In his second quote, he gives a problem and leaves the procedure open - specifically telling it to identify the techniques - to complete the solution.

To add to the topic, when I listened to the episode again I heard Dr Liebowitz make this statement before the quotes above:

@22:10 - I don't know any computers that can do statistical analyses. They can run regressions that we tell them to run, but they don't know what regressions they need to run.

This sounds like level 3, where the learner may have to start with very little scope and identify the problem-procedure-solution on their own. If we look to Bloom's Taxonomy (in the image), analyze, evaluate and create are the highest levels of the cognitive domain. Is it possible for a machine to reach level 3? Will we be able to create a machine and/or intelligence capable of freely identifying a need and developing the path to solution? Could a machine develop such context awareness that it could produce in nearly any field or would each intelligence be limited to a single specialty such as pharmacology, lawn care or drone bombing the enemy?

No solid thoughts at this point, but it made for an enlightening discussion so soon after reflecting on the use of Multiple Intelligence Theory and how artificial intelligence might develop. I'll come back to this topic again later as I haven't revisited inquiry-based learning for a few years and only then to give a brief introduction as part of a workshop. I'll read through some of the more recent and detailed literature and see if any of these human-focused tools can provide some insight when we can honestly plant a flag on the Singularity and say it has been discovered.

More information on inquiry based learning:

• Introduction to Inquiry Based Learning
• Technology and Inquiry Based Learning
• Wikipedia: Inquiry-based learning
• Herron's Model