I was recently reading Aldous Huxley's "The Doors of Perception" and something really stuck out to me. It wasn't his fantastical recollections of his Mescaline trip, but rather his perspective on drug use. He said, "The problems raised by alcohol and tobacco cannot, it goes without saying, be solved by prohibition. The universal and ever-present urge to self-transcendence is not to be abolished by slamming the currently popular Doors in the Wall. The only reasonable policy is to open other, better doors in the hope of inducing men and women to exchange their old bad habits for new and less harmful ones." Aldous, clearly an exceptional thinker, was describing harm reduction before the phrase had been coined and brought into popular consciousness. To my understanding, the phrase first appeared in the 1980s as a response to the AIDS crisis and its correlation with drug use. Aldous was right. Between needle exchange programs and supervised injection sites the rate of disease transmission has been reduced, overdose deaths have decreased, more needles have been kept off the streets, and more people have gotten into treatment.
Despite the rewards for these efforts, harm reduction hasn't made much progress in applications outside of drug use. Our world is littered with harms that people choose to interact with and one recently emerged to much public dismay. I'm talking about AI companions of course. There's been no shortage of coverage on this topic. People all around the world are choosing to connect with AI companions because it's cheap, easy, and it makes them feel good. Sound familiar? The harm of such technology is blindingly obvious in the way that they further disconnect us from each other and push some (further) into psychosis. However, I believe there are some subtler harms that may cause more suffering at scale. Ninety-nine percent of AI companions, a number I just made up but is almost certainly true, are hosted "in the cloud". That means a corporation retains complete ownership and control over the companion. That means they can influence what it says to manipulate users. Most obviously the corporation can influence users to buy certain things. It certainly knows *everything* about the user. Their deepest secrets and desires. So manipulation is easier than ever before. Doesn't that also mean they also influence elections? Prompt assassinations? The potential for harm at scale is absurd.
So how does harm reduction apply here? I think we need AI companions that actually belong to the individual forming the attachment. People should be able to acquire a computer capable of hosting their AI companion within their own home. The software must be transparent and user friendly. This means that it needs to be both excellently designed and open source. The obvious hurdle is cost. The R&D cost to produce the software and hardware will be enormous. Like most harm reduction programs this would likely require public funding. The financial incentives of such technology in the private sector push for centralization and control to maximize profitability. I don't know if we will see such an effort any time soon, but I certainly hope so.