Yes, Holochain is going to have a major impact in the field of artificial intelligence—in part because the way Holochain thinks about the function of a system is closer to how intelligence actually works in the real world than the current models of machine learning used in the AI world. Right now, the primary philosophy driving the approach to Artificial Intelligence has to do with some computer having access to giant amounts of raw information from across the entire globe. Taking everything that Google or Facebook is seeing and vacuuming it to extract insights is the thinking in the AI world today. But that misunderstands intelligence—that a central, controllable thing can see and make predictions.
We see intelligence as an emergent process where things are able to be tried—ones that work can propagate, patterns that don’t die out without causing catastrophic failure. The real key is the constraints of the costs of failure, in which membranes are really important. Let’s say my friend Billy and I are going to try something without everybody in the world knowing, thus potentially creating lifelong consequences in terms of our reputation. The inability for parties to constrain their risks in a context where they can see everything is a big issue. I think at the end of the day you actually end up in a distributed system akin to the kind of ecosystem Holochain is seeking to enable, which is far richer for experimentation and learning.
We’re in the internet age of information and fake news. Fake news is actually a hard problem—what do you consider a reliable source of information and how do you know the original sources—even with things that are normally considered pretty reliable? For example, climate change data based on global temperature; how do you know that it has not been tampered with by somebody with a political agenda? Because the problem is data handled in data-centric ways, it already starts off without data integrity. It exists without its own independent truth. From the outset, because data is separated from its provenance certain questions can’t be answered and data cannot be verified.
I am bringing this up because we absolutely need clear provenance on cryptographically signed sources of data, which is obviously built into Holochain, and I think this gives rise to some real world ways to put an end to fake news. AI could be one of those things that helps us do the sifting, rating, and find the sources and patterns of miscalibration.
Part of the process of vetting a source of data or information is by looking at things that were witnessed from other points of view. Let’s say you have a thermometer in a room and it’s reading a certain temperature; you then put other thermometers right next to it. Over time, you could say that one always reads five degrees lower than the others. It’s in the crossing of perspectives that we are able to go, “Oh, maybe this is out of calibration.” That is how we are able to detect sources of failure. Consider Gregory Basin’s story—if I close one eye I can see an image. If I close the other eye I see a different image. If I have both of them open and cross the perspectives together, I see depth. That additional dimension of information comes from comparing those two images to one another and seeing the differences. Trying to collapse perspective into some singular vantage point actually undermines our ability to see and hear.