Intelligence Must Be Time Independent

Why is it practical and not just fun philosophy to assume time doesn’t exist?

Logical and Efficiency

if you want yourself or your machines to learn and adapt faster and better than other people and machines then you should not organize signals by default in a time (observed causal) order.

But do we do this? no and yes. The human body does not. It has evolved several systems that work well to not do this.

But our computers (all of them) do order everything by recency. From social media and tv news feeds to the raw bits on disk and circuits on chip. When we invented practical computers (not theoretical computers which are time agnostic) we built in time preference. From ram to disk storage algorithms to user interfaces. We did this because at the time we were inventing computers we mostly were doing simple chained math calculations where there is a logical and efficiency case to be made for order of operations being time sequenced.

And so for 50 years we’ve built one time sequenced layer upon layer. And now we’re at an existential chasm for democracy (an attempt to fairly distribute resources) while were also hitting a wall for machine learning (from spam detection to self driving cars to game playing bots). Our systems are unable to detect the deeper patterns and to detect them fast enough because all of the systems are having to de-couple time from the data before finding new patterns.

Consider what a federal criminal investigation actually does? It first has to disconnect from the socially experienced real time. And then has to build up a fact pattern of possibilities. No single witness or account or evidentiary trail will report causal connections the same. Prosecutors must allow for all sequences. Only at the end can they do a probability analysis of the most likely causal ordering. But if an investigation assumes a given causal ordering facts will be missed. This is why investigations “take time”. They have to simulate a huge space of possible causal chains.

Now consider what a computer does. On disk the most recently used data is stored close together in ram or even on disk. Most recently used short cuts and browser histories etc etc are at the fore. The operating system is constantly indexing data and files and polling the cloud putting things it assumes it will need sooner into a cache etc. the entire system is always trying to stay recent. This is why when you’re trying to find something from years ago or something you rarely access or is bizarrely named etc it “takes time”. Every aspect of a computer is organized around recency. Now take this basic thing and consider how it plays out in a self driving car. Same challenge. Car computers are referencing the immediate and recent to make decisions. This is why currently the cars must stay in places they’ve been a bazillion times. They aren’t learning. They are actually building up deep histories of a repeated recency.

Guess what happens to a human that only sees the recent stuff? They build up a deep history of repeated recency. They reduce their adaptability. They make it harder/costlier to find non time dependent concepts.

When we do breakthrough to new governing ideas and new levels of learning and adaption in our cars, computers and selves it will because we process signal in a time agnostic way.

Practically speaking or in get rich in tech terms the breakthrough in AI is going to come by storing and retrieving data (organizing data) independent of time. As signals come in store them by their geometric/algebraic shape. And do not attempt to aggregate them into caches or lossy compressed versions immediately. And to not just put signals within a device but also consider its ecosystem.

Biological systems process and organize signals this way. They do not encode time. They intake coincidental signals. And as signals build up in the brain, cells, environment any of the signals that show up frequently in a creature and its ecosystem stick around…. dripping into lower and lower layers from sensory organ structure to neural networks to cell processes to genetics and the geological environment itself. Signals processed this way better represent the probability space of reality because Einstein clearly showed us… causal observations can be very different for different observers. The key is learning the time independent physics/possibilities.

This, I believe, is why painting/art/narrative is so universal to humans. For example, a painting is a huge collection of signals that is not “stored” or “read” in any particular time sequence. Sure it may be constructed stroke by stroke in a time sequence but that is not how its meaning or information coheres and is perceived by an observer.

So in a way I am suggesting directly that we organize our machines much more like biological creatures painting paintings.

Bank it.