Supercomputer vs. Slime Mold

When I was a pre-tween, my grandfather and I used to ride the Boston public transit system for fun. We had no reason to be anywhere specific, we just liked ‘riding the rails’, exploring what for me at least were new sections of my city.

Caveat: Don’t try this today. Anyone riding Boston’s public transit for fun in this century would be locked up…and rightly so!

Sometimes our outings would be ‘random walks’ across the city’s neighborhoods. Other times, we would map out our excursion in advance and in great detail. Either way, our objective was always the same: Visit as many interesting places as possible without ever retracing a single step.

Fortunately, in those days Boston Transit (MTA/MBTA) had a robust system of transfers that enabled us to ride all day for just one nickel fare. (Remember ‘Charlie on the MTA’: He would ride forever ‘neath the streets of Boston.)

There were virtually no islands: each station (node) connected directly to several other stations (nodes) making  ‘return trips’ unnecessary. A led to B and B led to C, D, and E, at least one of which connected back to A without returning through B.

Little did we know, Gampa and I were working to solve one of mathematics’ most daunting challenges: the Travelling Salesman Problem (TSP). According to TSP, a salesperson needed to visit a fixed list of cities and wanted to find the shortest possible route to touch all cities once but none more than once. Without realizing it, Gampa and I were doing graduate level math.

***

The City of Tokyo is famous for its commuter rail service. Again without realizing it, their transportation planners had been working on TSP for decades: how to connect a given number of destinations (e.g. 25) in the most efficient possible way. 

Imagine doing that without a computer! Impossible IRT: a reasonable approximation was all that could be hoped for. Today’s supercomputers can design a perfectly optimal flow…but even the most powerful would need several hours to map the Tokyo rail network efficiently. There are 10^32 ways to connect Tokyo’s 25 transit stations. 10^32 – that’s 1 with 32 zeroes after it. Granted, most are absurd, but a computer has to try everyone.

***

You fancy yourself a bit of a home chef and you pride yourself on a spotless kitchen. But one day, those darn grandkids spilled a box of cereal and left 25 bits of oat flakes scattered on the top of your ‘island’…along with a single cell of Slime Mold.

Unlike you and me, Slime Mold lives to eat and eats to live; it obeys The Gastronomic Imperative. So our intrepid unicellular immediately sets to work on a plan to visit all 25 oat-nodes in the most efficient possible sequence. This is precisely the same problem that Tokyo’s supercomputer had to solve. And Slime Mold was also able to ‘solve’ it, also in just a few hours.

One difference: the supercomputer’s solution is perfect, 100% optimal. Slime Mold’s solution is between 91% and 96% optimal. So if you happen to have a supercomputer like Deep Blue in your kitchen, it will outperform, marginally, that pesky mold you’ve been trying to get rid of since you bought the house.

Or will it? Grandkids being the way they are, those oat flakes are likely to get moved around (but not cleaned up) several times in the course of a day. If you wanted your optimal map updated each time the configuration shifted, Deep Blue could give it to you…but it would require several hours run time…every time. 

Each time there is a change in the data points, no matter how slight, the supercomputer must start from scratch. Given the speed of grandchildren, it is likely that Deep Blue’s initial map will already be inaccurate by the time it’s ready. 

Slime intelligence works differently: it never has to start from scratch. Once it conducts its initial environmental mapping, it retains that map indefinitely in its ‘memory’ and, when the territory changes, Slime Mold changes its map only as much as necessary to accommodate that change. Most adjustments can be made in a matter of minutes keeping the map evergreen.

Slime’s 4% - 9% fault tolerance creates the ‘wiggle room’ that allows it to arrive at a functional solution within a practical time frame. That wiggle room makes slime’s solution more flexible and therefore more adaptable to changes in underlying  conditions. 

Example: Suppose Tokyo decided to close a station, a supercomputer would have to start from scratch; Slime Mold might not need to make any adjustment at all (to remain within its 91% fault tolerance), but if it did, it would require minimal time and energy. Slime builds on past experience; computers have no past. 

So Slime Mold can adjust to changes in a matter of minutes; Deep Blue will need several hours, every time. According to Karen Alim, a theoretical physicist at the Technical University of Munich, Slime Mold appears able to learn, remember and make decisions — all without a brain.  Apparently, it uses variations in the flow of cytoplasm and configurations of its tubular network to record its experience. 

Consciousness is the ultimate opportunist. It will co-opt virtually any configuration of material to facilitate conscious experience: your neurons, Scarecrow’s straw, slime’s cytoplasm. Anywhere it can create a feedback loop, it will. It’s as though self-awareness was an Ontological Imperative: Know thyself! 

That said, consciousness manifests in different ways in different contexts. You are conscious (I hope); so is your pet goldfish. But you two have very different intellectual capacities and very different experiences of being-in-the-world. If your AI Bot now or later becomes conscious, it will bring its own unique skills to the party and it will experience being-in-the-world in its own unique way.

Deep Blue has almost limitless intelligence but no wisdom. Slime Mold’s intellect is limited to just a few functions, but it has the wisdom of Solomon. 

***

Ernst Haeckel’s Plate 85, “Ascidiae” (1904) from Art Forms in Nature presents sea squirts with an almost architectural elegance, highlighting the intricate logic of their forms. His rendering makes these simple marine organisms feel like deliberate designs, revealing a quiet intelligence in the geometry of living structures.