Can a human-centred AGI simulate a solution to rectify the emotional shocks to humanity brought about by historical events that disrupt social interaction?
Image credit: Natasha Connell, unsplash.com
Subash ChapagainOct 08, 2020
What kind of particular biomolecular and cognitive manipulation (say dopamine/serotonin/and-an-army of neurotransmitters) would be required to regulate behavioural outputs so that we would actually be emotionally resilient in the moments when we are subjected to unprecedented emotional turmoil? How do we make forced changes to behaviour less stressful?
Anecdote No. 1: I had been watching football (soccer for Americans) for quite a time -and so have us all (at least some kind of sports). There used to be stadiums full of people cheering up and a lot of noise and the game's exclusive playlist. There used to be Mexican waves and standing ovations, to imagine a player's point of view. There also used to be a lot of cussing and extrication of angry rants. Sometimes there used to be fist-fights on the bars in the middle of the cities. To cut it short, there used to be a lot of emotional expenditure and behavioural output.
Anecdote No 2: In pre-COVID classrooms, there used to be first-benchers, back-benchers and the monitor. Now, there is a host and (some) co-hosts. There used to be a lunch break and the tussle to eat first and go to the playground. Now it's all gone.
Artificial General Intelligence (AGI)
Aimed at building the machines that can think, AGI is an emerging field of study geared towards developing general-purpose systems with intelligence level comparable to that of the human mind. Since the mainstream AI research has turned out to be domain-dependent and problem-specific, AGI has emerged as a new science to indicate research that pursues the dream of ‘Human-Level AI’.
Human-centered AGI
Human-centered AGI is the kind of AI that learns from extensive human-based input with the goal of understanding human language, emotion, and behaviour hence bridging the gap between machine and human being. In business economics, this has already been used to understand customer behaviour by using qualitatively loaded data and contextual analytics to combine data with human science (mainly neurobiology). This makes it drastically easy for algorithms to predict mass behaviour, aspirations, and hence provide tailored services.
Coming back to the premises of the question again:
Humans are machines running on emotions. Emotions are more often attached to real tangible things, places, and people . Every-time human beings get forced to behave differently, there is an emotional constraint to it .
Post-COVID, we have been keeping inside. We have been locked down; everyone is watching the matches from home. Everyone attends Google class. But it doesn't quite feel right. There's a gap: players miss the atmosphere of the stadium and fans miss the collective bonding brought about by the experience of going to the games; students miss being around friends, teachers sometimes genuinely miss being physically around their students. We might pretend that it is not much different in the 'NEW NORMAL', but we know that would be untrue.
On such grounds, if there ever were to be an optimally approximate artificial general intelligence (AGI) capable of filling in the kind of emotional chasm brought about by the unprecedented happening of events like COVID-19 pandemic, what could be the key design components to simulate the emotional manifestations? What could be the variables to analyze? In other words, what data points from me as a biological machine would the AGI algorithm need for it to be capable of making itself more emotionally resilient and temporally dynamic?
[1]Trampe, D., Quoidbach, J., & Taquet, M. (2015). Emotions in Everyday Life. PLOS ONE, 10(12), e0145450. https://doi.org/10.1371/journal.pone.0145450
[2]Berkman, E. T. (2018). The neuroscience of goals and behavior change. Consulting Psychology Journal: Practice and Research, 70(1), 28–44. https://doi.org/10.1037/cpb0000094
Internet is full of emotions and can be used to guide AGI
J
JuranOct 19, 2020
Summary
Managing emotions could be done by applying up-to-date methods in situations of unprecedented emotional turmoil to reduce the negative effect of the event and regulate the behavior
A human-centered AGI maybe doesn`t need the biochemical background to understand the emotional process; maybe all it needs is the trigger and an outcome
AGI could be learned to influence browsing content a person sees online in order to therapeutically help
I saw two important questions in your session (tell me if I am wrong):
How to reduce the stress from emotionally stressful situations by hormone/other-kind-of manipulation?
It is a very important question in today`s dynamic world of changes that hit us, mostly unprepared. Following @Anja`s tracks, I would say the emotions, although still mostly unknown, can be partially influenced and thus, regulated using various methods:
Drugs
Illegal drugs (cocaine, LSD, heroin) are often used to play with your mind and emotions
e.g. Ketamine - used to be given to wounded Vets to "get them away" and observe as a third-person; recently approved by the FDA to treat depression
e.g. acetaminophen (Tylenol) and ibuprofen (Advil) reduce emotional pain in addition to physical pain
antidepressants make you feel numb
Professional psychological approaches
Physical ways of managing emotions using light, sound, or e.g. optogenetics, introduced by @Anja
Spiritual ways of reaching the emotionless state or a state where you can control your emotions (something like "The choiceless awareness" theory )
Personally, I don`t think emotions can be deleted, just "managed" or an event can be (un)intentionally misinterpreted (e.g. sociopaths).
Back to the question. From all the above mentioned, we can conclude there are ways to regulate behavioral outputs in order to become emotionally resilient. Not all of them are safe, but it is possible. The "safe ones", such as sound or visual effects that trigger the emotion, could and are already being used to add a subconscious emotional dimension to our perception of various products and situations (colors, symbols, textures, shapes and smells in marketing make you think differently).
Managing emotions could, therefore, be done by applying these methods in situations of unprecedented emotional turmoil to reduce the negative effect of the event and regulate the behavior. While following a class from your couch, the sounds of e.g. creaking chalk or chairs scratching the wooden floor could be introduced to increase the fidelity of the virtual in-class experience and reduce the anxiety from missing friends and the in-school environment. In the context of a stadium experience while watching the game, increased sounds effects and online platforms where people can jointly watch, comment, and share game experience could bring the power of mass emotion to your home.
(Sorry if I missed the topic with this one.)
2. How to translate the real emotion to an AGI-readable signal (and back)?
An extraordinary idea you have here! There are tons of applications where you could use the ability of AGI to mimic human emotions.
Some of them which immediately crossed my mind where, for example:
more realistic and sympathetic virtual assistants like Siri and Alexa that can then share our happiness or help us when feeling sad or depressed
emotion-powered decision making in the AI industry (human-like robots that "feel")
novel therapeutic approaches to treat psychological disorders
preventions or diminution of unwanted public reactions (fights after the match, protests, ) by "pre-sensing" the emotion-triggers and reduce negative reactions
The most intuitive things to be considered are, as @Shubhankar mentioned, biochemical mechanisms underlying emotions. Enriched with an integrative network of biochemical processes, emotions, and behavioral outputs, AGI could mimic, predict, and create tailored solutions for emotion-driven problems.
But, if we look at emotions as an end-product of a biochemical process, triggered, memorized, and often retriggered by the extracorporeal stimulus, a human-centered AGI maybe doesn`t need the biochemical background to understand the emotional process. Maybe it just needs to know two things: the trigger and the outcome. The trigger should be then reduced to "tags" (stimulus components that make it a unique experience) and introduced into a database of human behaviors (if there is such). Triggers can nowadays be easily found online. People's emotions can be seen in words, phrases, sentences, emoticons, and GIFs they use while commenting or texting. Phone calls, Siri requests, browsing history, and music can be used to perceive what kind of emotion the person is experiencing at that moment. Even the pauses between messages can, accompanied by the user activity status, be used as an emotion source. Therefore, human-centered AGI could be able to learn about emotions by using the right algorithm, I guess.
Following this principle, AGI could be learned to use the same strategy (online therapist @Shunhankar) as mentioned in the answer to the first question, to influence browsing content a person sees online in order to therapeutically help.
Tweaking the levels of serotonine, dopamine, and such is already possible through drugs for various mood stablizations, but I am not familiar with whether there are other usages, closer to what you have suggested. However, I remembered optogenetics, as a big discovery that is being looked into and developed further. The basic idea is a sort of a mimicry of brain electric impulses. The discovery itself relates to light-sensitive microbial proteins that also move ions that cause the same changes when they are genetically engineered into neurons. It has been proven to work on different specimens. I remembered this one of a bird species called zebra finch, where the use of optogenetics helped implant false memories and thus result in altering their song, which manifests in the tone length, etc.
So, rather than toggling with physiology on this primary and expected level, and also rather than going that much further as to implant false memories, perhaps it will be possible to use something like optogenetics for an "in between" the two previously mentioned usages and teach the AGI to recognize behavioral patterns of an individual, thus stimulating it lightly in both spoken word/provided images (especially this), etc. I would stress "an individual" here, because for a more sophisticated experience it would always need to be so, fine-tuned for a specific person.
Study human behavior, emotion, and chemistry together
Shubhankar KulkarniOct 09, 2020
Human behavior is largely regulated by emotions. To understand emotions and the underlying chemistry, human-centered AGI needs to study those together. These data will form the basis for the reactions the AGI will provide later during its application. One way to study emotions and chemistry together is to study the neuronal firing in the brain using an imaging system and to record the turnover of the hormones in the brain using an implanted electrode. This detection system should work simultaneously and collect real-time data on the emotional state of the person (based on the person's conversation with others) and associate it with the changes happening at the neuronal and the chemical level in the brain.
Studying humans this way will tell the AGI what each hormone does - when it is upregulated or downregulated and what are its effects. This information can be later applied to simulate a solution to rectify the emotional shocks.
This can be used as therapy for patients showing suicidal tendencies and other psychiatric disorders.
Subash Chapagain Oct 08, 2020
Anja M Oct 13, 2020