Original article was published by Mark Gulino on Artificial Intelligence on Medium
Developing Simulated Consciousness
I want to talk about some ideas for improving the development of simulated consciousness. After all, why not try to bring something to the table, right? Maybe I’m way off base — Regardless, I’ll give it a shot.
Personality Development Systems.
As a writer, creating a personality is not a foreign concept for me. Fiction writers are very familiar with the process(es) behind developing characters. The extent to which these characters can change and how much readers (or viewers) can learn about them is entirely in their creators’ hands.
I believe design teams can use randomization and procedural generation methods to create systems capable of producing unique, artificial personalities. Establishing parameters can help maintain the desired ranges to which attributes are limited.
For example, if you’re designing a digital character that likes animals, you might want it to select which one will be its favorite randomly. If you’re going to focus the available options to only popular animal types, you might set limits on which kinds they can choose.
Of course, allowing the system to select a personality’s preferences and attributes shouldn’t stop there. I would implore designers and dialogue writers to answer “Who, what, where, when, why, and how” questions more often.
When the artificially intelligent character says its favorite animal is a dog, and the user asks why they like dogs, the response shouldn’t be “I don’t understand.” There should be a simple follow up.
“The more design teams flesh out the details; the more believable these characters will be.”
There are many ways a technique like this can be applied to add depth to A.I. characters. The more design teams flesh out the details; the more believable these characters will be.
Continuous Creative Expansion.
Developing artificially intelligent characters will likely feel like going down the rabbit hole. With each detail, trait, and attribute established, more questions may arise that need answering.
“Putting a constant focus on the finer details and expanding the ways A.I. can appear to think, feel, and respond will encourage further consumer engagement.”
Putting a constant focus on the finer details and expanding the ways A.I. can appear to think, feel, and respond will encourage further consumer engagement. The more creative this effort becomes, the more convincing it will be in the long run.
Interpersonal Contextual Memory.
Contextual memory is a considerable part of cutting-edge virtual assistants. The ability to retain information from one interaction and correctly interpret its connection to another is essential.
For example, if you ask Siri or Google Assistant what the weather forecast is, they will tell you the forecast for the day you requested. If you then ask, “What about tomorrow?” they’ll respond by providing the next day’s forecast.
There are many ways contextual understanding makes artificial intelligence smarter and more effective. Unfortunately, I don’t often see it implemented conversationally.
But what if it was?
“Being able to remember and interpret key points in segments of user-facing dialogue would drastically improve conversational awareness.”
Being able to remember and interpret key points in segments of user-facing dialogue would drastically improve conversational awareness.
Let’s say you ask a virtual companion how it’s doing, and after it responds, it returns the question: “I’m great, thanks for asking. How’s your day going?” You answer, saying, “Not so great.”
Instead of ending the conversation there, what if it asked you what’s wrong, simulating interest, and then followed up with “I’m sorry you’re having a bad day,” simulating empathy.
Now jump to the following day.
What if this plays out again, however, this time instead of asking you how you’re day is going, it instead says, “I hope you’re having a better day today.”
It remembers you had a bad day and encourages you the next.
That’s what I refer to as interpersonal contextual memory.
I’d like to see proactive engagement — or, in other words, activity initiated by virtual assistants rather than the user. With information gathering and contextual memory, there’s no reason A.I. companions can’t reach out to users directly.
“What if it could ask why you’re having a bad day?”
Imagine the scenario I mentioned above, where an A.I. companion remembers you had a bad day. Now imagine it tries to do something about it. What if it could ask why you’re having a bad day?
What if when you say that you’re stressed, it offers to play some relaxing music or pull up your favorite show on Netflix? Are you feeling lonely? Perhaps it could offer to play built-in games together (like Google Assistant’s ad-libs mode).
These are just a few ways we can make A.I. personalities more proactive, and once again, increase product engagement.
Technology is continually advancing, and there’s no better example than smartphones. Smartphones are an essential platform for virtual assistants. Not only are these devices with us throughout our day, but they harness a wide variety of sensors that consume and interpret data.
Conditional awareness is not a new concept. Artificially intelligent processes can use hardware (like G.P.S.) to consider both internal and external information. Virtual assistants already use variables like location data, for example, to prompt notifications and reminders.
I would like to see more implementation of these techniques.
“We’re missing opportunities to create fun new ways to interact with digital personalities.”
With all the different ways smartphones can sense conditions and interpret data, we’re missing opportunities to create fun new ways to interact with digital personalities.
These are the “proactive engagement” opportunities I mentioned in the previous segment.
With devices that sense motion, detect eye movements, monitor location, and use facial recognition, there are plenty of options available to trigger interactions.
User Influenced Growth.
Machine learning is a crucial factor in the growth of artificial intelligence. The ideas I discussed above are ways users can input data, and developers can use these methods to output more authentic responses and solutions.
In short: Proactive engagement and conditional awareness can increase the flow of user-relevant information. Interpersonal contextual memory will improve personalized responses. User influenced growth evolves the process as a whole, making conversational A.I. smarter and far more personal.