Deconstructing humans is a common starting point for building human-like AI. Take your fingerprints: When holding soapy dishes, we instinctively alter our grip based on the structure of our fingerprints. It just doesn’t occur to us because we chalk it up to reflex – and for the longest time, scientists did as well. No one had any equations to figure out how this works since, well, it didn’t really important. However, the rise of automation has complicated matters.
For a robot to achieve this, we must first figure out exactly what is going on and then convert that information into writable code. Now, deciphering fingerprint grip is important, and researchers are finally attempting to develop a new physical law to explain it.
Physical knowledge and the capacity to code human qualities are, in some ways, requirements for robotic programming… This raises a critical question about the future of lifelike AI. Is it possible that some components of human consciousness will never match these criteria? According to certain thinkers, there may be.
And you might agree after reading through two extremely mind-blowing thought experiments. Or you might not.
What Mary didn’t know
A lady named Mary lives in a small house. She’s never gone anywhere. When she looks about her home and out the windows, everything appears to be black, white, or a shade of grey. Mary is colourblind, yet she frequently wonders, “What do those people on my black-and-white TV mean when they talk about red roses?”
Assume Mary’s room has a magical library. This fictitious location has books containing every tidbit of information on the colour red. And by everything, I mean everything. Mary reads everything to satisfy her need for knowledge.
She learns about red electromagnetic wavelengths, how crimson makes people feel, the most accurate descriptions of scarlet, cherry analogies, and whatever else she can think of. Plus much more. Mary is an expert on the colour red. Then she finishes her book… and decides to leave her house.
Mary is surprised to see the colour. She had never been colourblind. Her home, furniture, and equipment were all black and white, and her windows filtered the outside world in the same way.
Then something significant occurs. Mary notices a red apple, which is the colour of her skill. Her jaw drops open. She discovers something new about red. But… that’s strange. Why wasn’t this information in her library? It had everything there was to know about the colour red, right?
This narrative is based on philosopher Frank Jackson’s famous thought experiment from 1986, “What Mary Didn’t Know,” and the intangible piece of knowledge about red that Mary has just gained is known as qualia.
What is qualia?
Simply described, qualia is the knowledge that can only be obtained by conscious experience.
It’s similar to the subjective knowledge you received when you first heard your favourite song. You might have had shivers run down your spine and said to your pals, “You have to listen to it to understand.” Magically bombarding their minds with the depths of music theory and acoustic research is unlikely to work. They wouldn’t recognise the song until they heard it, unlike you.
Qualia may explain why even the best neuroscientists, psychologists, and poets are unlikely to be able to fully express the sorrow of heartbreak to someone who has never experienced it.
And, returning to Mary, all of the physical information in the world about red was insufficient to educate her what it felt like to stare at the colour. “She will learn what it is like to see anything red, say when she is let out of the black-and-white room or given a colour television,” Jackson writes. “This is appropriately regarded as learning; she will not say ‘ho, hum.'”
Though the hypothesis has been debated for many years, it remains a common explanation for how some knowledge is incomprehensible by language and unique to human cognition. That is, if qualia is a real force, its inner workings would be extremely difficult to write down and so programme. It is possible that it will act as a barrier between humans and AI.
On the other hand, it’s possible that it won’t. Perhaps we can decode it in some way, similar to how we’re gradually learning about fingerprint grip dynamics.
Can we do it?
The short answer is that we don’t know. Experts have argued in both ways, and some have proposed new perspectives. Most, however, are trapped behind conceptual barriers, and the fact is that qualia have no scientific explanation.
Making a Qualia robot
Okay, based on what I’ve said about Mary, you’ve undoubtedly come up with some objections to qualia. You’d be in good company: Loopholes are common in thought experiments, and Mary’s room is no exception.
Some counterarguments claim that colour pigments could have been conveyed by shadows in the room. Others argue that the “magic library” would have provided Mary with information in ways we can’t fathom. To the latter point, philosopher Daniel Dennett offers a remarkable refutation – one that is surprisingly germane to our major question about AI.
In a nutshell, Dennett contends that if Mary truly knew everything there was to know about red, wouldn’t she be omniscient? She wouldn’t only know about the hue like any other human. In theory, she should have studied about red “qualia,” if such a thing exists, as part of her red literature, right? And from this, we can deduce that qualia is absolutely ready to elevate AI, but we just haven’t worked out how to harness it yet.
Perhaps, but this appears to be a dead end. We don’t know for sure if Mary would have such abilities. We don’t have omniscience, thus we don’t even know what those abilities would look like. As a result, Dennett proposes that we forget about Mary’s humanity in order to remove these limitations.
“Thinking in terms of robots is a good exercise,” Dennett argues, “because it removes the excuse that we don’t yet know enough about brains to tell exactly what is going on that might be significant, allowing a sort of foggy romanticism about the mysterious powers of brains to cloud our judgement.”
What RoboMary knows
Welcome to the second thought experiment.
RoboMary is an iteration of a class of bots known as Mark 19, however, she was developed without colour vision and is now awaiting an upgrade. Until then, RoboMary’s “eyes,” or video cameras, could only relay data in black and white.
“RoboMary’s black-and-white cameras stand in perfectly for human Mary’s isolation,” Dennett writes. “We can let her wander at free through the psychophysics and neuroscience journals reading with her black-and-white-camera eyes.”
Essentially, she’s perusing her own version of the magic library. RoboMary, on the other hand, takes things a step further.
Dennett describes how she studies how Mark 19 colour inputs operate before “using her huge knowledge, she builds some code that enables her to colourize the input from her black-and-white cameras.” She can change herself in ways that human Mary cannot.
This new setting enables her to use her black-and-white vision to look at an apple, for example, and then accurately visualise it as the correct colour code for Mark 19 bots. As she explores the world, RoboMary begins automatically applying the setting to everything. But it’s here that she truly distinguishes herself. She observes other Mark 19s in action, dissects how they react to different hues, and modifies herself accordingly.
At this point, RoboMary recognises every colour input and responds to it in the same manner as any other Mark 19.
The big day has here. RoboMary’s colour sensors are activated.
“When she eventually instals her colour cameras, turns off her colourizing software, and opens her eyes, she observes… nothing. In fact, she needs to double-check that she has the colour cameras installed “Dennett writes. “She already knew what it would be like for her to see colours in the same manner that other Mark 19s do.”
I get chills just thinking about it. RoboMary appears to have simulated qualia for herself by adjusting her settings. But I can’t help but imagine a much frightening scenario.
What if RoboMary opened her eyes… and everything changed?
We are limitations as humans
The storey of Mary does not end here.
Despite numerous other changes – some of which come from Jackson himself in order to refine the original argument – Dennett’s entire book is also extremely careful.
He discusses the arguments you may have when thinking about Mary and RoboMary, and later into a difficult situation that does not enable RoboMary to tamper with her settings at all, to test if qualia are still preserved. There’s even a follow-up essay to Dennett’s titled “What RoboDennett Still Doesn’t Know.“
But, as with other philosophical thought experiments, Mary and RoboMary’s goal isn’t to tell you the truth. Its purpose is to force you to consider your options and come to your own conclusions regarding the truth.
Here are a couple of ideas I’ve had: To achieve qualia, AI may need to be constructed similarly to RoboMary. Or, if we can find a way to mathematically explain consciousness as a whole, perhaps robots can be taught to be conscious in a broader sense. A “aware” robot may theoretically explore the world in the same way that we do, and hence obtain qualia like us.
Or it’s possible that qualia aren’t what we think it is. Jackson’s account makes a compelling case that when Mary stares at red, something happens. It has been given the label “qualia” and is associated with learning something new, but what if it is a mash-up of many things with many names and has nothing to do with learning?
Or… just maybe, qualia is an impenetrable, unprogrammable barrier between human awareness and AI.
These are merely the tip of the iceberg, and they could (and probably will) be disproven in the next years, if not already. But remember why RoboMary was created in the first place: to envisage an entity that defies human boundaries.
It’s a far-fetched thought experiment because we’re limited as humans. We can only speculate.