The Independent Discovery of TCP/IP, by Ants | Hacker News Reader

The Independent Discovery of TCP/IP, by Ants

priceonomics.com

180 pointspknerd33 comments

Article is about rate adjusting, not packets.

\Theory We can only perceive nature's algorithms after we've worked them out for ourselves, because we don't know what to look for in the noise. Therefore, we are surrounded by incredibly clever algorithms, unbeknownst.

Also, I would write by humans, not by ants.

One could also have a deep philosophical discussion whether ants actually can discover anything (how much consciousness/ability to abstract/... is needed for that, and do ants posses it? Or maybe, do ants lack it, but ant colonies have it? Or is it DNA that discovered it using a genetic search algorithm?)

Have you heard of the China Brain thought experiment?

I assume you are referring to Searle's "Chinese Room" [1] thought experiment

[1] https://en.wikipedia.org/wiki/Chinese_room

No, China Brain. It is more analogous here.

https://en.m.wikipedia.org/wiki/China_brain

It's interesting that both "thought experiments" take essentially the same form - layer a fully working consciousness (defined either by being able to pass the Turing test or by actually emulating a brain on the physical level) onto a bizarre and improbable substrate that, strictly speaking, is too small to contain it; then gawp at the improbability of the arrangement you've constructed.

Dennett has a good rebuttal to this whole class - he notes that is not at all intuitive that a brain should be conscious. Open up someone's skull and look at the pulsing meat, and try and convince yourself that there's a mental universe in there. You may as well say "it's just stupid atoms bumping around - how could that be conscious?"

You either believe in a) supernatural souls, b) the mystical consciousness-granting power of meat, or c) minds are independent of their substrate.

I think a and c could be the same. Why not believe it?

Yeah I thought that as I was typing it, but I was procrastinating and didn't have time to fully flesh it out so I hit submit anyway!

In 1899 the equations of relativity or quantum mechanics may have sounded quite "mystical" too.

Lots of things sound incomprehensible, or fairy-tale-like, until you actually understand their intricacies.

Probably. However, more important, we are surrounded by incredibly clever medicine that we haven't discovered yet. When we tear down rain forests and diversity, those potential discovers are lost.

Since evolution has had a lot of time to optimise, and humans relatively little time to build the tooling and science to understand what's going on, I would assume there is a lot left to discover

We do embody a many of those algorithms. How does that Henn and Egg problem work out? I think you are right for the beginning of language learning and the rest is just synthesis and guesswork.

\Theory That's a hypothesis.

Discovery of TCP/IP is a significant overstatement. The ants perform flow control based on the number and frequency of ants returning from a known source.

They of course have no concept of out-of-order delivery, retransmitting lost ants or even having to deal with ants accidentally duplicated as they walk their paths.

> retransmitting lost ants

That's not an issue to ants and their problem.

Actually it is. The article seems to describe an algorithm with a catastrophic failure mode where no (or not enough) ants return. Without retransmission the ants just hang around waiting for a returning ant. So there must be some sort of retransmission.

With no retransmission you limit the number of dead ants though. If you continue to send ants you're going to get lots of dead ants in certain situations, so you'd need a more complex algorithm than simple retransmission.

The wait-state is an interesting adaptation, I have wondered about the small ants that invade my house, they don't appear to have any kind of similar requirement. Which does lead to a lot of ant death (I feel bad about it, but I don't want them getting into my food). Which has always struck me as maladaptive - to recover a few calories of crumbs, the colony will send hundreds of ants in, yet I am sure that growing those workers took more resources than they bring back. It may be that in nature there aren't any similar situations (for these ants, at least - no anteaters in North America).

I doubt that ants care about one of them dissapearing?

They try to optimise food-collection and their algo seems quite performant there. No? They won't wait for an ant endless but they won't go searching for it neither (like someone else explained it here)

Interesting on one hand, nothing new on the other.

Complexity through a few iterations of a very simple algorithm is one of the corner stones of Stephen Wolfram's "A New Kind of Science"[1] and has independently been proven feasible by VPRI's "STEPS" project [2].

I had to google (again) for documentation on STEPS and it's not the document I was looking for, but one of the for me astonishing insights was that correct text flow in typesetting is just a special case of a cellular automaton. If anyone finds the document that highlights this, I'd be grateful for a link.

[1] https://en.wikipedia.org/wiki/A_New_Kind_of_Science

[2] http://www.vpri.org/pdf/tr2012001_steps.pdf

> If anyone finds the document that highlights this, I'd be grateful for a link.

probably this one : http://www.vpri.org/pdf/m2010002_lobjects.pdf ?

Not the one I meant, but still interesting. Thanks :)

Comparison with TCP is forced. What ants actually do is regulate how many workers to send based on how quickly they return. TCP does this, but also much more.

Even just looking at the flow & congestion control by it self, it's not similar to TCP. Or at least the article didn't go into details. How does the decrement and increment of the rate is being done. In TCP we have the congestion window. Which is increasing and decreasing as a function of the state of the algorithm (slow start, congestion avoidance) and the occurrence of packet-loss or perceived packet-loss by the algorithm. In the article it's not clear how Ants are adjusting the amount of ants out there.

I thought ants discovered how to build java applications?

BT Labs at Martlesham and Prof Peter Cochrane was exploring packet routing based on similarities to ant colonies back in 1994.

It is the other way around. IP routing was inspired by the emergent behaviour of the ant farms.

Regardless of whether that's true, this article is talking about flow control, not routing.

Wouldn't the technical term for this be "End-to-end congestion control" and transmit window adjustment, instead of "TCP/IP"?

Kermit supported sliding windows, not x/y/zmodem. Best technology analogy is probably:

http://www.kermitproject.org/

> This article was first published on February 3, 2015

Ah! I thought I had seen it on HN before... :)

The story has been submitted many times:

https://hn.algolia.com/?query=TCP%20Ants&sort=byDate&dateRan...

No previous submission have generated any discussion, or garnered many points.

Useful when you're lost in a Jungle ;)

Reply on news.ycombinator.com