Hacker News Reader

Comment by patio11 | Hacker News Reader

The cryptocurrency enthusiasts call this a Distributed Autonomous Corporation and they envision a future where they're capable of being entirely self-sustaining by using APIs to a) create value for customers (human or, more interestingly, not), b) transact with customers, and c) purchase the resources they need to continue living.

(A rather smart gentleman who may not appreciate me attributing the conversation explained his theory that Bitcoin is, itself, one of these. He and I agree more on that than we disagree, though he meant it as a compliment and I'd phrase it as a "self-organizing Internet boiler room with very impressively engineered viral spread among people susceptible to certain flavors of memes.")

View on HN

There are a variety of names. DAC is something Ethereum came up with. Before then they were being called autonomous agents:

https://en.bitcoin.it/wiki/Agents

I was never totally clear on the difference, but as people often refer to Bitcoin itself as a DAC, I feel the term is too vague to be useful.

The next step to make these REALLY useful for agents is to set up Intel TXT on them. This would likely involve some voodoo as TXT is appallingly badly documented. But, if you make sure the hardware is compatible and that hardware virtualisation integrates properly using tboot it should be possible.

This step is a game changer because then an agent can get a hardware-rooted proof that the remote system is secure including against Kordless or whoever else owns the hardware. The agent then knows it's safe to install a copy of itself on the new machine, have the copy create a wallet, and send money to the new box such that the operator can't steal the money but the new agent can still pay for its own hosting costs.

Once done the agent would effectively be autonomous if it had a way to earn income e.g. reselling disk space. Scary stuff!

I didn't see anything in the TXT whitepaper about securing programs run by a hypervisor against attacks by the hypervisor. Seems more focused on making sure that managed VMs can't compromise their neighbors or do side-channely things.

If I provide your VM with a simulated "TPM" that actually emails me the private keys, how is that supposed to be secure?

If I understand rightly, this sort of security is impossible until fully homomorphic encryption is invented. Any actual cryptos available to comment?

The TPM is un-simulatable, that's the purpose of it.

The hypervisor is included in the remote attestation. The theory is, if you trust the signing keys used in the TPM (i.e. Intel/the hardware), then you can get a message from the remote computer that attests what code is running including the hypervisor. Thus you can conclude that the remote environment is safe against all but quite sophisticated hardware attackers (assuming a modern setup where you can lock cache lines and the TPM is integrated with the northbridge).

I had a rather engrossing conversation about a runaway version of one of these eating all our compute. One of the individuals in the conversation came up with a way to incentivize efficiency in the network, so we all calmed down a bit afterwards. I still wonder about it though.

Good news: Bitcoin can't go Clippy because you can't get > X compute for X compute's cost in BTC, at virtually any margin, to say nothing of the scales required to e.g. meaningfully impact availability at any major cloud provider. Ask if you care about the economic or engineering details, but typically, people running Bitcoin mining on public clouds are doing so to exfiltrate value from accounts or payment methods which they have stolen, in a way which is fairly difficult for law enforcement to combat. (That or they're enthusiasts who either don't particularly care about monetary incentives or should have spent a wee bit more time reading about currently successful mining strategies.)

To eat all your compute it would have to outbid everyone else, which would probably mean it was creating an incredible amount of value.

Which is also a roundabout comment on VC-backed vs. bootstrapped companies.

If the "DAC"'s wallet ever got stolen a DOS of epic proportions could take place! You could literally bankrupt the company by buying up their services with their own money.

Using a file hosting DAC as an example - if you kept uploading junk data the DAC would continue to procure server space using the stolen money from its own account. Seems an ideal way to rid yourself of competitors.

I have trouble comprehending the ideas behind DACs and find myself perplexed whenever I read about them. How exactly are they different than a regular corporation (except the whole it's completely automated thing)? I don't know why, but for some reason, DACs make me of 19th century banks issuing their own own notes.

More than being automated, DACs operate autonomously across an infrastructure built to run them for a fee they pay to be run.

Example: A Bitcoin address is created with a simple script [1] in it that causes it to pay for it's 'existence' by periodically sending BTC to a local printer it has entered into an arrangement to print flyers with it's own Bitcoin address on it. It also pays a Task Rabbit employee periodically to go to the printer and take the flyers and put them in people's mailboxes. The flyer then 'sells' itself by asking people to donate a small amount of money to a cause that was found to be somewhat important to the people who live in the area that was canvassed. It then gives X% to the cause, and pockets the rest so it can pay for more flyers to be printed and more Task Rabbiters to deliver.

Once more feature rich blockchains like Ethereum come online, you'll be able to code up a bot that is able to do all this by itself without human intervention. It'll know where the API to the printer is, will know how to pay for research APIs for BTC and find the best couriers to deliver. It'll also provide bounties for humans to improve its software so it can make more money.

I'm thinking this is maybe 10-15 years out, with hints of it happening in the next 2-3. Going to be interesting, that's for certain.

[1] https://en.bitcoin.it/wiki/Script

Reply on news.ycombinator.com