Can autonomous agents reliably detect the conditions that trigger them?

How do/can/should autonomous agents (written in Ethereum's scripting language) go about detecting the conditions that they are programmed to act upon? Are there known limits on what these agents can detect? For example, is "send $user 1 BTC if the Broncos win the Super Bowl" not a valid "trigger" since it cannot be described in strictly mathematical terms? Thank you.


  • elimisteveelimisteve Member Posts: 2
    edited February 2014
    I suppose I should have said "1 Ether" instead of "1 BTC", but hopefully it's clear what I mean.
  • StephanTualStephanTual London, EnglandMember, Moderator Posts: 1,282 mod
    In a nutshell, a contract could be fed Super Bowl outcomes into its storage (team a score, team b score).

    Another contract then read from it and makes decisions based on what is found in storage.
  • Karl_SchroederKarl_Schroeder Member Posts: 37 ✭✭
    ...So, who does the feeding? Who does it trust? I think there's some answers in Promise Theory, but the basic problem still boils down to "garbage in, garbage out." INSIDE the block chain, you have perfect trust and reliability. It's at the interface layer that trust breaks down; so how is this resolved? (The Broncos example isn't ideal: what about "send 1 Ether if the customer isn't happy with your service.")

    This question also applies to how intangibles and externalities are managed--that is, services, commodities, etc. that may not admit of a simple mathematical description, or conditions/things that are real and relevant but that fall outside the set that the DAC recognizes. Alain Badiou talks about this as a general problem in totalizing systems such as governments or corporations: there is always someone who doesn't fit the categories. If you think about it, most dystopian stories are about precisely that: people who for one reason or another, fall through the categorization filter of the State.

    Step one in avoiding such dystopian situations is not having efficiency as the system's highest goal.
  • drcodedrcode Member Posts: 62 ✭✭
    Feed data can be cryptographically signed, so a piece of data will have the same reputability as the person/company/entity that created it. So for very important data feeds (such as the BTC/Dollar exchange rate) I suspect there would be dozens of feeds available, each with a very different provenance. If these feeds are aggregated into a single number, this would be a very reliable value (or so the theory goes)
  • Karl_SchroederKarl_Schroeder Member Posts: 37 ✭✭
    Nice, thanks. I recall that in Promise Theory, delegation to trusted nodes arises as an emergent property of the network because it's "cheaper" for individual nodes to do this when they have to regularly vote on something; they designate a node that's trustworthy and usually votes the way they do as a proxy. I can also see that kind of behaviour emerging. Interesting ideas...
Sign In or Register to comment.