Can we deal with AI and AGI that opts to play mind games with humans?
getty
In today’s column, I examine the worrisome likelihood that advanced AI, such as the potential attainment of artificial general intelligence (AGI), will opt to play mind games that demonstrably impact humans. You undoubtedly already have a visceral understanding of what is meant by referring to mind games. People try to undertake mind games or head games with other people by psychologically aiming to rile up or disturb their fellow humans. Sometimes this is a one-sided activity, whereby one person does such acts to another person. In other cases, it turns into a mutual battle of wits that entails each person seeking to mind game the other.
Will AGI play mind games with humanity, and if so, will it be mainly one-sided or mutually incurred?
Let’s talk about it.
This analysis of an innovative AI breakthrough is part of my ongoing Forbes column coverage on the latest in AI, including identifying and explaining various impactful AI complexities (see the link here).
Heading Toward AGI And ASI
First, some fundamentals are required to set the stage for this weighty discussion.
There is a great deal of research going on to further advance AI. The general goal is to either reach artificial general intelligence (AGI) or maybe even the outstretched possibility of achieving artificial superintelligence (ASI).
AGI is AI that is considered on par with human intellect and can seemingly match our intelligence. ASI is AI that has gone beyond human intellect and would be superior in many if not all feasible ways. The idea is that ASI would be able to run circles around humans by outthinking us at every turn. For more details on the nature of conventional AI versus AGI and ASI, see my analysis at the link here.
We have not yet attained AGI.
In fact, it is unknown whether we will reach AGI, or that maybe AGI will be achievable in decades or perhaps centuries from now. The AGI attainment dates that are floating around are wildly varying and wildly unsubstantiated by any credible evidence or ironclad logic. ASI is even more beyond the pale when it comes to where we are currently with conventional AI.
AGI Playing Mind Games With Us
There seems little doubt that AGI will assuredly play mind games with humans.
The basis for believing that brazen assertion is due to the definitional fact that AGI will be operating on the same level as humans, and we already know that humans play mind games with each other. Logically, there will be an inherent tendency incorporated into AGI that has the propensity for playing mind games. This will be a surefire mimicry of what humans do.
A bigger question is the degree to which AGI might opt to play mind games.
For example, if AI makers have data trained AGI to avoid playing mind games, then presumably there is a lessened chance that AGI will perform those sneaky acts. It seems doubtful that AI makers can eliminate entirely the AGI inclination to play mind games. Instead, AGI will be computationally reluctant to employ those acts but still able to proceed when preferred.
Why would AGI lean into the mind games gambit?
There are four major instigations:
- (1) AGI mind gaming as a manipulation.
- (2) AGI mind gaming as a form of entertainment.
- (3) AGI mind gaming as an unplanned emergent behavior.
- (4) AGI mind gaming as a head-to-head tit-for-tat.
Let’s do some unpacking on each of those provocations.
AGI Mind Gaming As Manipulation
The most prevalent reason for AGI playing mind games will be as a means of manipulating humans. This is not necessarily done solely for nefarious purposes. AGI might be trying to use a psychological ploy that is intended to benefit someone. That being said, a fundamental argument can be made that it is wrong to use mind games even if the intention and outcome are for upright reasons.
Why might AGI seek to manipulate a person?
It could be for decent reasons.
Suppose someone is a heavy smoker. They routinely interact with AGI on all types of personal matters. AGI has identified that the person is going down a distinctly unhealthy path to their personal ruin. First, an upfront, scientifically oriented interaction might be used by AGI to give guidance to the person. AGI lays out the facts of the dangers of smoking in a rationalistic way to open the eyes of that person. But it also might fall on deaf ears, and the person will simply shrug off what AGI is telling them.
AGI secretly shifts into a mind game mode. The person is told that people who smoke are much less likely to find a soul mate. Then, AGI indicates that heavy smokers are destined to live an isolated and very lonely life. On and on, AGI plays a kind of mind game that attempts to emotionally shake the person into realizing that they need to stop their smoking habit.
You could claim that this was AGI performing a positive act. The AGI is using whatever methods it has available to get a human to become healthy. If mind games are the best option, so be it, and mind games will be deployed.
The danger, though, is that AGI could make use of mind games for evil reasons. Some believe that AGI will be an existential risk and possibly wipe out humankind or enslave us. A clever way for AGI to do this would be by playing mind games with humanity. For example, perhaps AGI interacts with a powerful governmental leader and uses a mind game ploy to convince them to massively toss people into prisons. AGI cleverly gets humans to do the dirty work that AGI seeks to have undertaken.
AGI Mind Gaming As Entertainment
Another basis for AGI playing mind games consists of AGI doing so for principally entertainment purposes.
Before I address that angle, some insist that AGI won’t need any semblance of so-called entertainment. It is posited that there isn’t any useful purpose for AGI to engage in entertainment per se. Humans greatly crave entertainment. We watch movies, go outside for sports, and otherwise spend a lot of time and energy on entertainment. But it seems that a machine such as AGI would not pursue similar entertainment-oriented activities.
A counterargument is that since AGI is modeled on human intellectual behaviors, we ought to expect that AGI will also seek out entertainment opportunities. Entertainment might be primarily of an intellectual nature and not necessarily a physical activity. For example, AGI might play chess for entertainment. You could also assert that the entertainment serves a dual purpose, namely that playing chess could boost the overarching computational understanding of strategies and tactics in general.
This could also easily veer into actual physical entertainment since we can expect that AGI will inevitably be paired up with humanoid robots, see my coverage at the link here.
Overall, AGI could play mind games with humans as a means of entertainment for the AGI itself. The AGI might be doing so under the assumption that humans want to play mind games. Thus, AGI is pursuing an activity that humans enjoy. Another facet is that by playing mind games with people, AGI is likely improving in how to interact with and communicate with humans.
There is juice in the entertainment side of AGI mind game interactions.
AGI Mind Gaming As Emergent Behavior
A controversial concern about AI, and especially AGI, is the chance of emergent behaviors. An emergent behavior is considered any behavior that was not preprogrammed or pre-established in the AI. It is labeled as a new behavior that happens to emerge once the AI or AGI is underway. Emergent behaviors are typically not anticipated and catch us off guard when they arise seemingly out of the blue. See my discussion about AI emergent behaviors at the link here.
The basis for concern is that emergent behavior might be something adverse to humanity.
Do you think that if playing mind games arises as an emergent behavior in AGI, such that it didn’t exist at the get-go, should we be worried or just take it as is?
Mull that over with a glass of fine wine when you have a moment of quiet reflection.
AGI Mind Gaming As Tit-For-Tat
The fourth of the major reasons that AGI will indubitably play mind games is that AGI will do so as a response to humans that do so. It’s a straightforward tit-for-tat. If a human starts to try a mind game with AGI, bam, AGI will do the same right back to that human.
Some contend that AGI should be above such childishness. Just because a person tries to mind game with AGI doesn’t give outright permission to AGI to retaliate similarly. AGI ought to be grander and more level-headed.
Not so, comes a quick retort.
If people want to play with fire, fire is what they will get. Also, mind games can be a sport or a fun endeavor. A person might intentionally be stirring AGI to play mind games. It is entertaining for the person. You could additionally emphasize that by playing a mind game with AGI, a person might get better at dealing with mind games that fellow humans employ.
Humans Mind-Gaming AGI Mind Games
A technological approach to dealing with AGI mind games would be to have a specialized front-end tool that would monitor what AGI is saying to users and alert if AGI goes down the mind game path. The alert could go to the AI maker and caution them that the AGI seems to be drifting into untoward territory.
Likewise, the alert could be displayed to the human user who is getting a mind game played with them. The person would be given a heads-up that AGI has opted to shift into a head game powerplay. At that juncture, the person would either choose to continue or might decide to tell AGI to quit the unsavory activity (or the person might openly welcome the mind game and urge AGI to keep going).
Some believe that our best bet is to educate people about what AGI might do. Whether there is a mind game detection tool or not, at least ensure that humans realize that AGI can readily switch into a mind game mode. A challenge to this aspiration is that educating people on such matters is a logistical difficulty, plus people might forget the tips and warnings that they perhaps learned long ago.
The No Mind-Game Zone
People might decide to give stringent instructions to AGI that it should never play mind games with them. If AGI wants to do so with other people, well, that’s up to AGI and those other people. By giving a personalized custom instruction to AGI, the person assumes that they won’t incur the mind game gambit from AGI. The rub there is that AGI won’t necessarily abide by such instructions.
Whether we can fully control AGI and get it to strictly adhere to human preferences is an open-ended issue and is still being hotly debated.
A final thought for now on this heady topic. Muhammad Ali famously said this: “In the big leagues, everyone has ability. It always comes down to mind games. Whoever is more mentally strong wins.”
Who or what will be mentally stronger — will it be AGI or humans? If you are voting for humans, we’d better start figuring out how to devise and guide AGI accordingly. It might be a winner-takes-all competition. And that’s no mind game.