Hacker News Reader

Comment by dcolkitt | Hacker News Reader

Much of the world literally does run algorithmically. Literally every day you trust your life, privacy and money to algorithms that make decisions without human intervention. The financial system trades trillions of dollars in automated systems that make split-second decisions in a way that precludes human supervision. Hedge funds and banks already trust algorithms that if broken could lose billions.

The existence of flawed software does not mean that it’s impossible to make reliable or trustworthy software. If you’ll never trust your money to a smart contract no matter how vetted, then I’d recommend you never fly a plane or store personal data in the cloud.

View on HN

It literally does not. The world runs advised by algorithms, but not governed by them. It's a fundamental difference. When algorithms in the real world create lose-lose outcomes, people override them, which is why when your credit card gets stolen you don't end up paying for stuff. You can bake that logic into a software contract, but if the design of your system is that the totality of software contracts are the final word, you have the same problem; you can't predict all the corner cases. It turns out a lot of the human beings that work for companies engaged in the financial system actually do stuff to solve problems.

> When algorithms in the real world create lose-lose outcomes, people override them, which is why when your credit card gets stolen you don't end up paying for stuff.

An exploit where I take all your bitcoins isn't a lose-lose outcome, and neither is that. If I steal your credit card and you have to pay for my stuff, you're the only party who loses.

If you steal my credit card and buy things with it, I will call my bank and I won't need to pay for it when I settle my statement balance at the end of the month. The bank will follow up with the merchant and/or law enforcement and, in most cases, the bank will get its money back somehow. In the cases where it doesn't, I'm sure the bank is insured against losses. Banks and insurance companies probably have a complicated model for estimating $X/yr in theft.

If someone steals my wallet's private key (= stealing my credit card), I'm done. There is no recourse AFAIK.

> If you steal my credit card and buy things with it, I will call my bank and I won't need to pay for it when I settle my statement balance at the end of the month.

This is an epically terrible argument for why you shouldn't have to pay for things someone else purchased with your stolen credit card. "I shouldn't have to because I don't have to"?

Compare tptacek's claim:

>>> which is why when your credit card gets stolen you don't end up paying for stuff.

There are four parties involved: you, the thief, the bank, and the merchant.

When you can afford to pay, here's a win/loss table in the case where you have to pay:

    you:      lose
    bank:     win
    merchant: win
    thief:    win

And in the case where you don't have to pay, but the bank collects from the merchant:

    you:      win
    bank:     win
    merchant: lose
    thief:    win

You don't have to pay, and the bank can't collect from the merchant:

    you:      win
    bank:     lose
    merchant: win
    thief:    win

When you have to pay, but you can't afford it:

    you:      win
    bank:     lose
    merchant: win
    thief:    win

Where's the lose/lose scenario?

I was sure someone would make this argument and debated dumping more words into my comment to address it preemptively, but figured instead I'd wait for someone to spell it out first.

So, just to clear this up: you, the customer, and your credit card company agree that it's lose-lose for retail customers to be on the hook for credit card fraud. It makes it much harder to actually use the credit card transaction processing system, which you want to do because it's way better and safer than carrying cash, and the credit card companies want you to do because they're credit card companies.

In any given situation, a merchant on the hook for stolen stuff does not agree that this is lose-lose. Them getting paid with stolen money is a win-lose for them. But the credit card company takes the long view: if customers were always on the hook for this stuff, customers would stop using credit cards as much. It's better for everyone if it's easy to buy things, and not terrifying to shop online.

Merchants that disagree with this assessment can refuse to take credit cards. They can just take Bitcoins, The Currency Of The Future. The consensus view here is not hard to discover.

As the holder of the credit card I can't think of a good reason why I would/should pay the debt off for a purchase I didn't make. I did not initiate the sale with the merchant, I did not verify I was the rightful card holder, I did not authorize the funds to be transferred, and I will not send money to that bank to make up for any of that. Banks are in the business of encouraging transactions and my card number could be skimmed from a physical device, or from hacked website on the internet, or I used it at Target https://money.cnn.com/2013/12/18/news/companies/target-credi..., or someone physically stole my wallet.

If they did try to collect then that sounds like a great reason to go back to cash because even if I get mugged, criminals can't spend me tens of thousands of dollars into debt by stealing my cash on hand. They are limited to what's in my wallet.

> As the holder of the credit card I can't think of a good reason why I would/should pay the debt off for a purchase I didn't make.

This applies to everyone involved, but someone's going to pay for the loss anyway, unless you can recover from the thief.

There is no reason for the merchant to pay for the purchase either -- he is even less culpable than you are. But that's what everyone is advocating here.

There obviously is a reason for merchants to eat the losses, or else virtually every merchant in the modern world wouldn't accept credit cards. The reason is that merchants (on the whole) agree with credit card companies that it is better for customers to feel safe and comfortable shopping, and thus increase transaction volume, even if doing so means that merchants will occasionally eat losses from fraud.

You are discussing a different level of "reasons" than frumper is. In your sense, there is a reason for merchants to take the losses, and there is also a reason for customers to take the losses. But in the sense that there is no reason why customers should take these losses:

>>> I did not initiate the sale with the merchant, I did not verify I was the rightful card holder, I did not authorize the funds to be transferred

there is also no reason why merchants should.

The fact that the system currently operates in a certain way is not actually evidence that it cannot operate in a different way.

It's pretty good evidence that there's no consensus that favors structurally irreversible transactions. But again, if you do, there's a Coin for that!

But there is a price to pay for the responsibility you burden off onto the bank. Banks charge fees, pay almost zero interest (at least currently) and are usually very unpleasant to deal with when you want to borrow money.

I’d say it depends. You probably shouldn’t manage your life savings using a crypto wallet. Keep some money at a bank. Keep some money in cryptos.

Cryptos are not a solution to every financial problem out there but they add new and exciting options for us. Don’t think black and white. But embrace more freedom to choose.

Crypto transactions have fees, crypto wallets don’t accrue interest, and my exchange won’t loan me coins. So I’m not sure this is the slam dunk on banks that you were going for.

> crypto wallets don’t accrue interest

The Coinbase wallet integrates the major lending protocols. Ethereum fees are reasonable these days. So lending USDC or DAI and earning interest has become a great alternative to saving accounts.

If you hold fair amounts of Bitcoin or Ethereum long term, you can use it as collateral and borrow USDC. This way you can not only let your money work for you, you can also use it to borrow dollars for a short term credit — like getting a mortgage but without a bank.

Furthermore, these days there are other emerging financial tools that banks simply can’t provide. Earning exchange fees on liquidity pools, for example, or staking cryptos and earn validation fees like a miner for Proof-Of-Stake networks.

Point is: Lots of innovation, here. When I look at banks nothing really has been happening there for decades.

Replying to adtac:

But what if someone does not want the insurance of the bank? There is no opt out. With something like bitcoin you ideally have the option to take the risk of final settlement on your own or delegate it to some insurance deliberately

You are aware that the vast majority of financial market trading is run by algorithms that govern not advise vast amounts of capital in an irreversible way.

Sometime these algorithms break or are exploited. Often in spectacular disasters. Knight lost half a billion dollars in 30 seconds. Nobody reversed though, even though it was clear error.

If you have a 401k I guarantee you that a significant amount of your money was traded by an algorithm that if broken would have lost everything. If the question is whether people would ever trust capital to irreversible algorithms, the answer is they already do. It’s called the stock market.

The world is more than just automated trading systems

The matching engine on the exchange is also automated. The settlement system is electronic. Etc. etc.

Yes and if any of those make mistakes, the trades can be rolled back with human intervention.

> Sometime these algorithms break or are exploited. Often in spectacular disasters. Knight lost half a billion dollars in 30 seconds. Nobody reversed though, even though it was clear error.

This actually blew my mind. The exchanges bust trades all the time. I don't why they didn't do it in this case.

I think quite a few people who find crypto objectionable will also consider many, even most of the weirder technological outgrowths of the stock market to be every bit as objectionable.

As we've repeatedly learned, to the detriment of most people in the world.

In a previous lifetime not so long ago, I was paid to find vulnerabilities in exchange infrastructure --- order routers, match engines, FIX gateways, all that stuff. When I was getting started, any interesting finding was super exciting. "This is a billion-dollar vulnerability! Suck it, Mark Dowd!"

Then I learned about out-trades. In reality, these heavily automated markets screw up (or, rather, customers of these markets screw up) somewhat regularly. Surprise: one of the reasons lots of humans work both at the exchanges and at companies that program directly to them is that mistakes happen, and are often resolved not by code but by hashing things out between cooperating blobs of electrified fat tissue. Gross, I know!

Obviously, there are big screwups where the rules coded into the algorithm are the last word. But those rules aren't always the final word, and, more importantly, the rules themselves are easily changed when the electric meat blobs want them to.

// Knight lost half a billion dollars in 30 seconds. Nobody reversed though, even though it was clear error.

You cited Knight twice in this thread, and it's just not the same concept. I think your point is that people should accept irreversible transactions in whole because there are some edge cases where they exist today, and that just doesn't make sense.

First, there's a difference between an average person and Knight in the degree of sophistication of tools that they can wield. So the fact that Knight is allowed to operate like this is in no way connected to whether this mode should be the default for an average person.

Second, the Knight example can be interpreted as an exception that proves the rule. They had an unforced error that (from what I recall) looked to the rest of the world like real trading. But there's a whole other slew of issues (clearly bad prices, exchange-errors, etc) that cause trades to be unwound. Some exciting reading if you care: https://www.investopedia.com/terms/e/erroneous-trade.asp

You have a point that yes to some extent we're all subject to underlying risk of irreversible activity but the bigger point is that we as society try to minimize it where possible. So moving the system in the other direction, especially where it makes it easy for Joe-Blow to fuck it up, is not the way.

There’s no categorical difference between Knight and Vanguard. If Vanguard fucks up its trading, the exchange is not under any obligation to make it right. The only difference is that Knight was a lot more aggressive in terms of pushing the envelope on its algorithms to reap excess profit. I’m sure Vanguard is a lot more politically sympathetic, but I’d trust that “guarantee” a hell of a lot less than I’d trust battle tested software.

Say the fund managers running your 401k decided to say “fuck it, turn off all the circuit breakers.” Would you just shrug your shoulders and say “no worries, I’m sure FINRA’s got my back if anything goes wrong. Blindly trusting the people running institutions to keep you safe is as misguided as assuming that all software is infallible .

I agree with the individual bits of data but I don't agree with your ultimate conclusion.

The point is - do you aim to make the system fixable by human judgement or not? This thread is full of examples we like: the credit card protecting us from fraudulent charges, FDIC insurance, etc. Are there places in our life that society is forced to operate without such backstops? Yes. But we want to minimize, not maximize those - especially when it comes to consumer-facing capabilities.

> If you have a 401k I guarantee you that a significant amount of your money was traded by an algorithm that if broken would have lost everything

How so? My 401k is invested in a fund that just owns a broad index of stocks, not an algorithmic trading hedge fund.

Even that has an "algorithm" behind it - they have to manage inflows & outflows, the heartbeat trades if it's an ETF, etc. Vanguard doesn't have a person sitting there buying a little bit of AAPL every time someone invests in VTSAX.

OP is trying to say that even something as simple as VTSAX could have a problem and, I don't know, accidentally sell all its AAPL and so it's proof that even for the current market there are rules that govern.

Except, of course, if VTSAX dumped all their AAPL somehow the trade would be unwound, as trades occasionally are. Selling stock on the market is not nearly as irreversible as it may seem.

Knight took closer to an hour to melt down than 30 seconds.

NYSE and the SEC _chose_ not to reverse those trades but they could have. There are plenty of ways the exchange can undo those and also trade busts happen daily on every exchange.

Almost certainly if a large index etf or fund lost everything to a software glitch the trades would be reversed and even if for policy reasons they weren’t they could be. The facility exists.

> Much of the world literally does run algorithmically.

I think you're missing the parent's point, which is around the world running on legal contracts.

It's that legal disputes are settled non-algorithmically. If someone harms you through fraud or other illegal action, a judge can order a transaction reversed, etc.

None of this has anything to do with algorithmic trading, or using algorithms in finance generally for efficiency.

> The existence of flawed software does not mean that it’s impossible to make reliable or trustworthy software.

It also doesn't mean that it is economically feasible to make 100% reliable or trustworthy software.

> If you’ll never trust your money to a smart contract no matter how vetted, then I’d recommend you never fly a plane or store personal data in the cloud.

Isn't vetting of contracts the opposite of having a no human touch algorithmically run world?

And who does the vetting of the code? Who does the vetting of the people the vet the code? Who does the vetting of the people that run the code?

To have your money in smart contracts right now is tantamount to seeing those 737 max plane crashes in the past, then plugging your ears with your fingers and saying I can't wait to fly on a 737 max tomorrow. It's kind of a mess out there right now [1].

[1] https://cryptoslate.com/binance-smart-chain-sounds-alarm-ove...

I'd say most of the systems you mention either a) have human failsafes b) have years of pre-algorithmic precedent such that the algorithm matches some well-known accepted process or c) are speculative in nature and therefor can assume the risk of a faulty algorithm.

Algorithmic stablecoins are mostly in category "c" so far. There is no human failsafe and there is no predetermined process that can shape the algorithm.

> The financial system trades trillions of dollars in automated systems that make split-second decisions in a way that precludes human supervision.

Nope. Regulations require human supervision, in some way, of that trading.

Furthermore, those trades have exactly the kind of non-algorithmic softness that omk talks about: if an algorithm makes a trade which is obviously incorrect, you can ask the exchange to bust it.

>Literally every day you trust your life, privacy and money to algorithms that make decisions without human intervention. The financial system trades trillions of dollars in automated systems that make split-second decisions in a way that precludes human supervision.

... but when that stuff goes wrong, which it does, we don't just say "geez that's too bad, the code is the code".

Look at the 2010 flash crash: something like 20,000 trades were broken after discussion between FINRA and the exchanges based on how far they were from the reference price.

Similarly, I don't end up eating the entire loss if my bank's anti-fraud system mistakenly approves a transaction on a stolen credit card, because I have 60 days from the statement date to report the problem under the FCBA.

The point is that in most cases where physics is not involved (e.g. flight control systems), the real-time behavior of the systems is backstopped by processes to deal with exceptions in a slower, more considered way.

Knight lost half a billion dollars in 30 seconds and not a single trade was reversed. If anything similar happened to any major hedge fund or market maker, nothing would get reversed.

Every hedge fund and market maker knows that if their algorithms break, their money is gone. Period. No backsies. The point is billions in capital is already completely trusted to algorithms with no human fail safe.

>Knight lost half a billion dollars in 30 seconds and not a single trade was reversed. If anything similar happened to any major hedge fund or market maker, nothing would get reversed.

I am pretty sure that NYSE canceled trades in six names after Knight; and the LULD pause rules that were introduced after the 2010 flash crash (unfortunately too late to save Knight) also create opportunities for human intervention.

But in many cases, there are ways to dispute and/or retrospectively fix or amend a transaction that is mis-classified.

In the exceptionally rare case that something is exploited within a cryptocurrency, there is no recourse for the victim.

Sure. But when you get pissed off at those algorithms, you can turn to the government to force the entities behind the algorithms to change the result. That's the whole point. It's not algorithmic in the end. Code is not law, and if it is, that's a bug, not a feature, to most people.

Reply on news.ycombinator.com