Robots don't just spring in to existence from nothing. Someone made the decisions necessary to make the robot do the work they're selling (even if that's just deciding what training data and biases to feed in to something they choose to call AI). You can sue that person.
Definitely click bait. The real title wouldn't be as interesting. "Who to sue when an algorithm doesn't work the way you were told it did?" That makes the answer very obvious. You sue the person that misrepresented what the algorithm does, which is indeed what is happening here.
20 years ago, when I described open source software to a lawyer, I got that exact same question. I told him, well, you can sue the developer I suppose, but he doesn't have very much money. The lawyer responded "well, I don't feel comfortable using software where there is no accountability". Looking back, this seems like a silly argument that most people don't follow in any meaningful way in this day and age.
In the end, people will use whatever software has a good reputation. If something bad happens, well it's on them. The sky didn't fall using this model.
One could argue that 20 years ago there was little to no reputation (good or bad) for most open-source software, and that made using open-source software much more dangerous : it was more difficult to separate the good apples from the bad ones.
Therefore his criticism made sense at the time.
While open-source code and open-source maintainers are critical parts of the ecosystem, I think open source users are underrated. If nobody shared their experience about how a piece of open-source software is actually behaving once you put it to use in production, open-source software would have much less value than it currently has.
No, open source has been a "thing" for quite some time even in 1998. No, the guy just didn't get it. He asked questions like who do you sue if something goes wrong. . . where is the market incentive to create software if there is no strong intellectual property rights. . . what if someone steals the code and makes a million dollars. . . What if the original author dies starving while some rich corporation takes advantage? I mean, really silly arguments when you look back.
I told him that most of the software using this model is simply more reliable than proprietary software, but he just didn't get it. There was already plenty of "reputation" even back then, he just wasn't a hacker and couldn't understand any other perspective other than his own.
No, open source software has been a regular thing since the 1980's and arguably with 1970's with richard stallman. It was already 20 years old by the time I was explaining it to the lawyer. He was just unable to understand anything that wasn't part of his paradigm.
Being "a regular thing" and being "a thing that enriches companies off the backs of community-minded software developers" are not concepts in conflict.
"The future is already here — it's just not very evenly distributed."
You're right that open source had been a thing for quite some time in 1998. GNU was started in 1983 after all.
But I respectfully disagree that open source was "obvious" in 1998. Hell, at that time, one could argue it was not obvious for a lot of people how much the Internet and computers would alter our lives. Open source was the least obvious part of quite a few a mind-bending paradigm shifts.
Open source was very obvious in 1998 for competent tech companies. As in look at successful companies in 2005 and most of them where using OS software in 1998.
One of PG’s essays has a line like, ‘any competitor using Oracle was easy to ignore.’ Successful OS software needs to be good enough not to need marketing and happens to be free. It might not be the best, but cheap and good enough generally has a lot going for it.
I'd argue that the same is true today. No big org is using pure OSS. They have big contracts with Redhat and the likes and they like to pretend that they're covered.
You always pay for all software in time. Someone needs to install the software in your company. Someone needs to answer support calls. If you pay RedHat when your people can't figure it out they can turn to someone else - a specialist (You hope: Redhat has a good reputation but there is always risk it is just someone else who knows less than you and will take more of your time understanding the problem than to just figure it out yourself.)
My CEO could do my job. He would need an additional degree in programming, and even more time in a day. He hires me because telling my bosses bosses (skip several levels) boss to get something done will go down the chain into figuring it out how to go from we need to grow market share in the area I'm working to actual code that will grow that market share. This across thousands of employees - the CEO could learn to do any one but he can't do all of them.
This is wrong on a higher level too as most commercial software will disclaim all liability for the correct operation of their software unless you pay them lots of money for a support contract. Turns this is just as viable a business model for OSS too, as Redhat and others found out.
You sue whoever misrepresented the robot's capabilities to you. There's nothing wrong with making a robot that's bad at trading. There is something wrong with pretending it's not bad at trading when you're selling it.
If they didn't fully understand how it worked, but fraudulently represented that they did, they'd be on the hook. It all hinges on what he was promised when he bought it. If they had a good disclaimer, then he might be out of luck.
There are places where adopting AI is going to be difficult for this reason - notably anywhere that has a legal requirement to explain why a decision went a particular way. That includes things like insurance risk models, pension investments, etc.
In these cases, I wonder if it would cause a movement toward more interpretative models with more mathematical backing.
For example, if not being able to explain a model leaves an organization open to litigation, they may instead rely on statistical based learning methods even if they performed less well.
The losses made by software training may be the least problem for the institution.If case goes to court,they may be required to reveal the reasoning behind software's decision and that means going through the code,as all that marketing bs called AI is just a trading algorithm trained on some datasets. No company would want to reveal the code.
Good point, but at least the probabilities of casino games are generally known and available AFAIK. People just tend to not care or have a confidence bias.
Your answers diverge, depending on who is operating the "robot". If it's being presented as a service by an investment firm, then you obviously have a standard dispute with the investment firm, the same as if one of the human brokers went nuts. If was sold as a software system to be self-administered, then you have an issue with a vendor. The latter may hinge solely on misrepresentation, because software is generally sold with little warranty for precisely these reasons.
So much money slushing around can make things a little grayer because in either case the investor didn't actually perform any work themselves. But I'm confident they can use some of that money to discern who ultimately owned what, who employed who, and what type of relationship was created.
I feel like newswriters just love to generate hot air with this "sue a robot" trope. Who to sue for that?
Is it even possible to misrepresent the robot's capabilities? In the UK most (all?) advertisements come with cautionary notices such as "capital at risk" or "past performance not indicative of future returns" or similar...
Although obviously such warnings can't reach some people... My favourite example of this was the collapse of OptionSellers.com following the huge move in Natural Gas futures. I mean, they were selling options - it's right there in the name...
IANAL but I could see a lawsuit being possible if historical "actual" performance figures were misrepresented or back-tested theoreticals (read: overfit) were passed as actuals.
We're already seeing companies shirk responsibility of their automated processes. The censorship by social media is a prime example. When somebody complains that Twitter shadowbanned them, Twitter will point the finger at the algorithm and say they couldn't help it, it's not what it looks like, it's just the algorithm.
For the time being, they often go back and give posts and accounts a manual second review, and often reverse the decision. For now, that keeps the general conversation around this issue from going too far. But it can't last.
Why? The software did it was supposed to do. It doesn't guarantee profits. There's a risk involved.
If anyone should be sued, its Costa (Captain Magic/con-artist/salesperson/whatever) who, if what the investor claims is true, greatly oversold the capabilities of the software.
And even then there might not be a strong case against him.
To me, this is the right solution. Someone might have persuaded him into the decision, but he's the one that made that decision in the end. You shouldn't be making decisions like this if you don't understand the risks.
No not the same way for stock markets. It's not a statistical certainty that you will lose money like gambling in a casino is, and millions if not billions of people depend on the markets in order to invest for retirement. There are people who are certified stock brokers and certified financial advisors who have a legal fiduciary duty to maximize returns and minimize risk on their client's behalf.
> It's not a statistical certainty that you will lose money like gambling in a casino is
But it is a very real possibility as the recent financial crisis shows. Everyone who is unable to stomach a total loss of his investment in one basket should not do so, and it is double dumb to invest in unproven automation like this.
Salespeople often exaggerate to make a sale, it's not uncommon. Believing them with enthusiasm is a sign of someone salespeople love to find. This guy is right to sue but he's still a sucker who was taken in and then leveraged himself to the hilt based on a pitch he could not (or would not) verify. A prudent person would have tested it with a small amount to start with.
"People tend to assume that algorithms are faster and better decision-makers than human traders,” said Mark Lemley ... “That may often be true, but when it’s not ..."
Is it just me or is that quote kind of disingenuous? "People tend to assume" and "That may OFTEN be true" make it sound like it isn't as clear cut or there's still doubt over whether such algorithms outperform humans.
Is that truly the case? I don't know much about trading but aren't algorithms doing most of the trading now?
The faster part is certainly not a matter of debate. The better less so and it depends a lot on the area. High Frequency Trading they essentially dominate and given what I heard also really prove their role's legitimacy as arbitrage as more HFT bots drove down the margins in a move towards uniformity - just like arbitrage should.
They are prone to some occasional "dumb" errors however - like humans but different those which attempt to predict trends using feeds like Twitter have caused losses as they fail to get the context.
And it depends on the area and how you define algorithm and trading. Your bank has lists of requirements for mortgages beyond legal minimums - essentially already following an algorithm but due dilligence involves humans in the loop.
The S&P 500? The FTSE 250? The Nikkei? The Russell 2000? The Russel 3000? The risk-free rate? LIBOR?
When you make statements like this, they mean nothing at all.
Let's say you're talking about the S&P 500. Many hedge funds have been beating it for 20 years or more.
But let's say you had 2 Billion dollars. What would you do with it? Are you telling me you would put it all in one index fund? Maybe you would put it in a couple, maybe you would put it in some bonds? Well guess what, you just made an active management decision! You decided what index, or what bonds, or whatever.
According to Vanguard, over 85% of actively traded funds don't beat the S&P 500, and it's quite possible the one's that do are simply lucky, as in any random distribution there will be a small percentage, of course, that are over 2SD on the profit side.
2bil is so easy. Highly rated tax-free munibonds...a very safe 3% coupon will net you completely tax-free 60mil/yr...I would say plenty to live off of and even reinvest.
> Many hedge funds have been beating it for 20 years or more
Relatively few, I would say, if you measure it by realised Sharpe. Rentech, Brevan Howard, D.E. Shaw, and maybe AHL and Winton excluding the past few years. I'm struggling to name others that have done well over a sustained period.
There are a lot of funds out there, and IMO much of the supposed 'outperformance' is a combination of leverage and survivor bias (I say this as a former hedge fund quant).
I remember reading an Andrew Lo quote back in 2007-8 where he said something along the lines of there at the time being about 9000 hedge funds in operation at last count, and for comparison, there were about 7000 Taco Bell franchises.
Given that sample size of funds, possible returns, and any distribution you want to fit into it, your odds may be better at a roulette table.
Huh. I hadn't realised that two sigma were that old. My point still stands though.
Market making is a different kettle of fish in the sense that it's more of a financial service than a directional bet on markets. It's also much less capital intensive so there's less pressure to raise money from clients. As a market-maker, you can certainly find yourself on the wrong side of a trade, but generally the goal is to be as market-neutral as possible.
I mean, the goal of absolute return hedge funds is pretty much the same: only have exposure to alpha factors. With market makers it’s pretty much the same thing: only have exposure to order flow. That’s the alpha.
I’ve worked as a quant at both hedge funds and market makers, and both have a similar philosophy of a only getting exposure to “alpha”, though the means and capital required (as you mentioned) are quite different.
I think a concept of Sharpe ratio still applies to both though, with market making having an clearly higher one.
Personally, I’ve found working for a hedge fund a lot better, market making is kind of a drab business when you get down to the nuts and bolts.
There is such a thing as low-fee management, though, and that’s really what investors care about. Nobody offers the actual Sharpe portfolio (how could they?) but you can get your choice of approximation for a handful of basis points.
Some investors care about returns. Some care about Sharpe ratio. But investors only care about low-fees when they underperform their benchmark net of fees.
Sorry should have made it clear was referring to index in the article S&P 500. And no, most hedge funds can’t even match it. Once you figure in the fees they actually make less money for their clients. It’s known facts, look at historical returns. Sure, there are some that do much better for 15 years and then lose most of the extra wins in one year, still figure in the fees for investor and it’s never worth it. You have better odds at the roulette table just doing your own gambling.
I recommend “Common Sense on Mutual Funds” by John C. Bogle on this topic.
It seems like this hedge fund was run by people who didn't understand finance in the first place. This is how many quant funds work:
1.You develop an alpha factor that you believe is associated with out performance.
2. You control risk factors such as beta, volatility, Fama French 3 factor, etc.
3. Now you create a neutral L/S portfolio that only has exposure to your alpha factor while also negating the risk factors like beta and volatility.
4. Backtest it, run a paper portfolio for awhile, etc.
5. Combine your new alpha factor with a bunch of other factors that you have already developed. The idea being that layering these alpha factors on top of each other will negate some of the noise inherent to each factor.
6. Make money and cycle out and in alpha factors for as long as they work/don't work.
It seems like this "hedge fund" was solely trading based on sentiment without creating a L/S market neutral portfolio or layering on any other factors. This is a very bad idea. The quote from the guy who made the software is pretty damning:
"The signals we have been provided have a strong scientific foundation. I think we did a pretty decent job. I know I can detect sentiment. I’m not a trader."
This is a big red flag. A lot of AI/ML people have this arrogance that trading is easy and that you don't need any financial knowledge to make money. Maybe that was true in the 1980's, but at this point, it requires an incredible level of expertise to generate and implement profitable quant trading strategies.
But all of this is off-topic, to get to the point: the Bloomberg article is garbage click-bait. You sue the General Partners, and I don't think anyone is confused about this.
>Maybe that was true in the 1980's, but at this point, it requires an incredible level of expertise to generate and implement profitable quant trading strategies. //
Are you saying there is no luck in the market any more? Doesn't that mean it's deterministic; and doesn't that mean that market trading is unnecessary as we can allocate needs perfectly?
It's not deterministic. When it comes to achieving high returns luck is more important than ever! The public markets have become more efficient and so it's now harder to win based on skill. Most of the funds which have reported high returns are actually just rolling the dice and hitting a lucky streak; the ones that lose big get shut down quickly and you never hear about them.
Active traders are trying to find mis-priced securities (market inefficiencies). That used to be relatively easy for smart people who could think up a few key insights. But now there are so many quant funds deploying fast computers and legions of mathematicians that most new strategies stop working quickly. The act of making those trades drives out inefficiencies.
Market trading is still necessary for liquidity to minimize the cost of capital.
I am not a trader (and have zero knowledge about it) and what I'm about to say might be utter garbage, but hear me out and correct me where I'm wrong.
> This is a big red flag. A lot of AI/ML people have this arrogance that trading is easy and that you don't need any financial knowledge to make money. Maybe that was true in the 1980's, but at this point, it requires an incredible level of expertise to generate and implement profitable quant trading strategies.
If they were making the trading bot via training it on past data and what they call 'sentiment signals' generated from social media reaction, etc, do they need experience with trading?
Is there something else that a human trader knows that the bot (which has been trained on past data and accounts for live reactions) doesn't?
Something that stood out to me from the article was that the bot "generated a single trade in the morning." Unless they're trying to realize a quick gain by exploiting the "dumb money" most knowledgeable investors specifically avoid trading at market open because the volume/volatility typically comes from Charles Schwab et al. executing trades from regular Joes & Janes. It's noise, not signal.
Since the bot wasn't implementing a day-trading strategy it's pretty clear the programmers didn't know much about how the markets work. I would guess the training data reflected that knowledge gap. As the saying goes, you don't know what you don't know.
I'm not a trader either, but I have non-zero knowledge about it...
> do they need experience with trading?
That's a yes. You don't get to put a lot of money onto an investment without some experience on how it's done. This is true for all forms of investment, not only trading.
How do you even evaluate to robot if you don't know the market? What do you do if the robot breaks? What do you do if the market breaks (happens once in a while)? How much money you leave under its supervision?
Absolutely. You wouldn't know how/when to back test, you wouldn't know how to reduce turnover in your portfolio, you wouldn't know how to measure the volatility of your portfolio, you wouldn't understand why you were making or losing money, you wouldn't realize a regime change in the market, you wouldn't be able to efficiently allocate trades to reduce slippage, you wouldn't know how to optimize your portfolio to achieve a better Sharpe ratio, etc etc etc. The list goes on and on and on.
I worked as a quant, and the finance knowledge you need is very deep.
(Pedantic tl;dr: I refuse to put the punctuation inside the quotes. I consider that an outright bug in English and it needs to be stamped out by rebellion.)
"Using his eyes," "using her eyes," and "using its eyes" have a certain parallelism. Comparing proper nouns to pronouns has more problems than a missing apostrophe: note the lack of capitalization in the pronoun. Horrors!
Not a bug, IMO. In general, we have separate individual words for possessive pronouns; we don't create them by adding the possessive suffix to a nominative pronoun: I -> my (not I's), you -> your, he -> his, she -> her, it -> its, etc.
"It's" is just the standard use of apostrophe to indicate a contraction.
I'm glad it's not possible to Sue in my country. Saves a lot of silly nonsense and has people a bit more responsible about their actions instead of doing random shit and sueing willy nilly when the shit hits the fan
In Sweden you can only sue in civil cases where there's a real disagreement, not because someone spilled coffee on you. Not sure if it would be applicable in this case though.
We often hear about the crazy lawsuit culture in US anyway and how insane it is. Whether that is true or not I don't know.
But tort law does exist, right? So if somebody purposefully spills boiling hot coffee over you, you will be able to sue them. The claims are just less outrageous than in the US because there are no punitive damages for example.
If it was bug that caused him to lose money, sure, he should be compensated by the company that made/owns the robot. But he says that it's capabilities were misrepresented, so the robot doesn't actually have fault, he either just made a bad investment or was scammed.
It usually works the other way around. A bug in the code doesn't often rise to the level of negligence. (Though 99.99% of the time a fund will reimburse a loss caused by a buggy algorithm for customer relations reasons they aren't obligated.) Misrepresenting the capabilities/returns of a strategy is where fund salespeople get themselves into trouble time and again. It's perfectly legal to sell a terrible investment strategy so long as you've been truthful.
If my microwave oven burn my pop corn, I don't sue the microwave oven. The product (robot) is not what seller tell you : you sue the seller. The product (robot) has a defect : you sue the maker.
If I buy a chainsaw and accidentally cut my finger off I wouldn't assume that I have any legal right to sue the chainsaw manufacturer or the person who sold it to me.
At least not where I'm from - is this not the case in the USA?
A more fitting metaphor along this line would be if you bought a chainsaw-weilding robot and directed it to cut down some trees, and it instead cut off your finger. Once a product has agency and is making decisions of its own accord, if it makes the wrong decisions, I don't think it's unreasonable to say the person who programmed it has some level of responsibility, likely some kind of negligence.
You really can’t pin anything like this on a single person, as tempting as it may be. The company that sells a product should deal with the consequences and accept responsibility as a complete unit. If someone coded something and somehow the bug eluded understanding all the way up the chain, it’s everyones fault for not putting more “solidness” pressure all the way down.
Theoretically, that’s how it works in the US, but not in practice.
Years ago, I noticed that my kids’ inner tube had multiple disclaimers, including multiple English disclaimers. The disclaimers next to the Union Flag, Australian Flag, and Canadian Flag were “use under competent supervision.” The disclaimer next to the US Flag was much longer and more detailed (“use under adult supervision” [note that is different from “competent supervision”], “do not tow from boat,” “do not use when drunk,” “do not dive into”, etc.). It did not make me proud to be American.
Additionally, many states have the concept of “joint and several liability,” so that if multiple parties share responsibility for an accident, the victim can collect the full amount from any of them and the perpetrators are expected to pay each other appropriately. The end result is that, generally, the company with the deepest pockets pays the full amount and then hopes to collect from the other parties later.
So, the answer is that you sue the chainsaw manufacturer, the chain manufacturer, the landowner, etc. and hope that they settle or that you get a final judgement that you can collect from the richest one, even if that particular party is held to be, say, 1% responsible.
Nitpick, but you can sue for anything, just not necessarily be successful at it.
This, along with [AIUI] not default providing an award of costs to a successful plaintiff in USA, makes it possible to sue and get a substantial out of court settlement because the risk profile is such that this is cheaper than paying your defence lawyers considering the [mathematical] expectation that you might lose a case that seemingly has no merits.
This is why you get warnings like "chainsaws are dangerous and may cause bodily injury or death, only for use by certified personnel". Then you can't [it is hoped by all reasonable people!] successfully claim that your naivety should have been accounted for, as it was by the warning.
The whole framing of this case is incorrect and stupid. It doesn't matter if the software used for trading is a bunch of if statements or machine learning, it's still just software. If it's misrepresented the case is pretty straightforward. The algos used don't matter.
> Over the following months, Costa shared simulations with Li showing K1 making double-digit returns, although the two now dispute the thoroughness of the back-testing.
And I wished the article linked to the fillings or at the least discussed this more thoroughly.
We obviously didn't have enough bullshit jobs for human beings, so we're gonna create some useless robots as well. It makes me think of Futurama with all the worthless robots (Preacher Bot, Robot Devil, Homeless Bot), which were finally explained after the show came back as the robots self-reproducing.
Wasn't there a Philip K Dick story about a human computer programmer in the future where robots were taking over all the good programming jobs, who had to pretend to be a robot, in order to get the high paying robot-only programming gigs? Every morning when he signed in to work, he had to fail a Turing Test in order to prove that he was really a robot.
The founder/CEO of the company that built the trading system had previously "agreed to pay $17 million to the U.S. Securities and Exchange Commission to settle charges of defrauding investors at his mobile-payments company, Jumio Inc."
If you're an accredited or institutional investor, then you have only yourself to blame (unless the investment was misrepresented, in which case it should be the one who sold/brokered the investment).
In this case, the investor was foolish to believe that a magic AI system could generate reliable positive returns. At the same time, it sounds like the performance of the system was misrepresented. Either way, if you're responsible for 1bn of assets, you better do really good homework.
Investing is not a guaranteed return system. You might win, or you might lose all your money. At the end of the day, the only person who is responsible for it is you.
I think the point is, an experienced investor should be able to either see through a fraud or at least do proper due diligence before investing a large sum. Plus, institutions often make a small initial test investment. If the performance is as expected, risk considered, then they may make the big investment.
A retail investor can't be expected to properly research claims, but a professional should be.
One does seem to get a pass from (possibly multiple) financial failures. In fact, one very high profile American with a history of bankruptcies comes to mind...
Trump has not declared bankruptcy. He has been owner or part-owner of hundreds of companies and a handful have declared bankruptcy.
As terrible (or not) as that is, there's a stark difference between availing yourself of long-standing bankruptcy law and the just-for-you shenanigans pulled by the federal government in 2008.
Forgetting who's responsible for trading losses (the customer should know investing is never a sure thing), this sounds like a very crude algorithm. The best trading programs don't use machine learning and sentiment scraped from the internet, they use well defined strategies that are set by a competent human. The whole point of using AI is to use it for tasks humans aren't suited to; we're far better than machines at gauging sentiment over days and weeks. Computers are better at seeing numerical trends over shorter time periods, or identifying lesser known securities based on trading signals. Using a computer to try timing a major benchmark seems like a massive waste of time.
Maybe the concept of blame is meaningless when it comes to robots, and we’re clinging to a cultural and spiritual concept of blame that serves no purpose when it comes to robots.
If a robot doesn’t function properly, you fix it or take it out of service. Taking retribution against a robot (or abstractly, the civilization that created it) is of questionable value. The right question is how do we build robots that don’t lose fortunes?
The fact that human beings get angry at robots and want retribution or recompense for malfunctions is an evolutionary adaptation for dealing with other humans. It is useless when it comes to dealing with non sentient deterministic agents. Sure, if the robot was being controlled by a human, go after the human. If not, what’s the point?
This cannot work as a practical matter. If a person is killed by a robot, it is important to have a legal theory to understand some analogue to "the responsible party". Otherwise, the introduction of a "robot" in the workflow becomes a wildcard to evade torts of all sorts.
If the goal of the legal theory is to reduce human deaths by robots, why not just reduce human deaths by robots by dismantling or repairing the appropriate robots?
Again, the legal theory works because when dealing with human beings you have to assign responsibility to human beings. Hurricanes and earthquakes kill tons of human beings and there’s no legal theory of responsibility. We simply work as a society to minimize those deaths without blaming anyone.
You seem to be suggesting that a sufficiently complex system dissipates responsibility, transforms an act of engineering into an Act Of God.
If that's not your intent, I suggest you rephrase.
When a bridge (another system which is complex but tractable in principle) fails, we work hard to find the feature of the design that led to the failure, and if we decide it was negligence, we exact consequences.
From current news: Should Boeing get a pass on their MCAS failures because "Gosh, it's really hard" ? The MCAS is a robot.
Responsibility itself is a heuristic with a goal. As autonomous systems take on more and more responsibility, tracing back a historical cause will be less and less relevant than actually fixing the fault in the system.
Take Boeing for example. Assigning blame to individuals, while rewarding, may not be as effective as changing the system that allows Boeing to continue to act the way it does. Boeing scapegoats, fires some people (which may include the CEO) but the organization, and the nation and legal system that sustains it continues largely unchanged.
You’re trying to patch by punishing humans individually, while ignoring faults in the system we have collectively created. Our instinct is to assign blame because historically and evolutionarily harm is usually done by individual humans, and punishing humans is a good heuristic. But it does not seem to me that punishment does a good job of creating good large scale organizations or systems.
It depends on what basis algo trading was sold to the customer. Regardless of whether it's a human or computer, anyone willing to give such an amount a go, should be asking how risk is beimg hedged?What happens when shit hits the fan? How fast can you exitvthe position?If it's Forex, what stops are being used to minimise the losses and etc. I'm sceptical that either sides signed the deal without firstly going through some details on how it works, the risks and rtc.Also,the legal department probably did their homework to cover as much as possible in such cases. The responsibility, ultimately,falls upon the institution utilising the software and not the software itself.
139 comments
[ 0.20 ms ] story [ 208 ms ] threadIn the end, people will use whatever software has a good reputation. If something bad happens, well it's on them. The sky didn't fall using this model.
Therefore his criticism made sense at the time.
While open-source code and open-source maintainers are critical parts of the ecosystem, I think open source users are underrated. If nobody shared their experience about how a piece of open-source software is actually behaving once you put it to use in production, open-source software would have much less value than it currently has.
I told him that most of the software using this model is simply more reliable than proprietary software, but he just didn't get it. There was already plenty of "reputation" even back then, he just wasn't a hacker and couldn't understand any other perspective other than his own.
The evolution of open source software in the twenty years hence suggests that this particular question is not as silly as you frame it.
You're right that open source had been a thing for quite some time in 1998. GNU was started in 1983 after all.
But I respectfully disagree that open source was "obvious" in 1998. Hell, at that time, one could argue it was not obvious for a lot of people how much the Internet and computers would alter our lives. Open source was the least obvious part of quite a few a mind-bending paradigm shifts.
One of PG’s essays has a line like, ‘any competitor using Oracle was easy to ignore.’ Successful OS software needs to be good enough not to need marketing and happens to be free. It might not be the best, but cheap and good enough generally has a lot going for it.
My CEO could do my job. He would need an additional degree in programming, and even more time in a day. He hires me because telling my bosses bosses (skip several levels) boss to get something done will go down the chain into figuring it out how to go from we need to grow market share in the area I'm working to actual code that will grow that market share. This across thousands of employees - the CEO could learn to do any one but he can't do all of them.
On the contrary every big org is using OSS.
OSS is in everything.
That there is 'nobody to sue' is besides the point, it's like suing the inventor of the fork because something happened with a fork.
If you bought something complicated from someone with a contract and you depend on it doing ABC and it doesn't, then there's a case of liability.
But tons of software is OSS and there's nobody to sue.
'Liability' in software is maybe important in some areas, but mostly it's not.
Most of our problems as a society are due to that gap.
For example, if not being able to explain a model leaves an organization open to litigation, they may instead rely on statistical based learning methods even if they performed less well.
Is cheesing an algorithm illegal?
Saying the algorithm makes 95% winning trades may not mean much if those remaining 5% of trades cause you to lose most of your value.
So much money slushing around can make things a little grayer because in either case the investor didn't actually perform any work themselves. But I'm confident they can use some of that money to discern who ultimately owned what, who employed who, and what type of relationship was created.
I feel like newswriters just love to generate hot air with this "sue a robot" trope. Who to sue for that?
Although obviously such warnings can't reach some people... My favourite example of this was the collapse of OptionSellers.com following the huge move in Natural Gas futures. I mean, they were selling options - it's right there in the name...
https://www.bloomberg.com/news/articles/2018-11-19/hedge-fun...
For the time being, they often go back and give posts and accounts a manual second review, and often reverse the decision. For now, that keeps the general conversation around this issue from going too far. But it can't last.
If anyone should be sued, its Costa (Captain Magic/con-artist/salesperson/whatever) who, if what the investor claims is true, greatly oversold the capabilities of the software.
And even then there might not be a strong case against him.
Uh... don't get me wrong, but when you go to the casino and lose all your money, it's no one's fault but yours. Same way for stock markets.
Whoever is dumb enough to put money into essentially a gambling algorithm should be able to pay for the losses himself.
But it is a very real possibility as the recent financial crisis shows. Everyone who is unable to stomach a total loss of his investment in one basket should not do so, and it is double dumb to invest in unproven automation like this.
Is it just me or is that quote kind of disingenuous? "People tend to assume" and "That may OFTEN be true" make it sound like it isn't as clear cut or there's still doubt over whether such algorithms outperform humans. Is that truly the case? I don't know much about trading but aren't algorithms doing most of the trading now?
They are prone to some occasional "dumb" errors however - like humans but different those which attempt to predict trends using feeds like Twitter have caused losses as they fail to get the context.
And it depends on the area and how you define algorithm and trading. Your bank has lists of requirements for mortgages beyond legal minimums - essentially already following an algorithm but due dilligence involves humans in the loop.
The S&P 500? The FTSE 250? The Nikkei? The Russell 2000? The Russel 3000? The risk-free rate? LIBOR?
When you make statements like this, they mean nothing at all.
Let's say you're talking about the S&P 500. Many hedge funds have been beating it for 20 years or more.
But let's say you had 2 Billion dollars. What would you do with it? Are you telling me you would put it all in one index fund? Maybe you would put it in a couple, maybe you would put it in some bonds? Well guess what, you just made an active management decision! You decided what index, or what bonds, or whatever.
There is no such thing as passive management.
2bil is so easy. Highly rated tax-free munibonds...a very safe 3% coupon will net you completely tax-free 60mil/yr...I would say plenty to live off of and even reinvest.
Relatively few, I would say, if you measure it by realised Sharpe. Rentech, Brevan Howard, D.E. Shaw, and maybe AHL and Winton excluding the past few years. I'm struggling to name others that have done well over a sustained period.
There are a lot of funds out there, and IMO much of the supposed 'outperformance' is a combination of leverage and survivor bias (I say this as a former hedge fund quant).
Given that sample size of funds, possible returns, and any distribution you want to fit into it, your odds may be better at a roulette table.
Also if we're talking about market makers, Jane Street, Susquehanna, Fortress, and Citadel very well for themselves.
Market making is a different kettle of fish in the sense that it's more of a financial service than a directional bet on markets. It's also much less capital intensive so there's less pressure to raise money from clients. As a market-maker, you can certainly find yourself on the wrong side of a trade, but generally the goal is to be as market-neutral as possible.
I’ve worked as a quant at both hedge funds and market makers, and both have a similar philosophy of a only getting exposure to “alpha”, though the means and capital required (as you mentioned) are quite different.
I think a concept of Sharpe ratio still applies to both though, with market making having an clearly higher one.
Personally, I’ve found working for a hedge fund a lot better, market making is kind of a drab business when you get down to the nuts and bolts.
I recommend “Common Sense on Mutual Funds” by John C. Bogle on this topic.
1.You develop an alpha factor that you believe is associated with out performance.
2. You control risk factors such as beta, volatility, Fama French 3 factor, etc.
3. Now you create a neutral L/S portfolio that only has exposure to your alpha factor while also negating the risk factors like beta and volatility.
4. Backtest it, run a paper portfolio for awhile, etc.
5. Combine your new alpha factor with a bunch of other factors that you have already developed. The idea being that layering these alpha factors on top of each other will negate some of the noise inherent to each factor.
6. Make money and cycle out and in alpha factors for as long as they work/don't work.
It seems like this "hedge fund" was solely trading based on sentiment without creating a L/S market neutral portfolio or layering on any other factors. This is a very bad idea. The quote from the guy who made the software is pretty damning:
"The signals we have been provided have a strong scientific foundation. I think we did a pretty decent job. I know I can detect sentiment. I’m not a trader."
This is a big red flag. A lot of AI/ML people have this arrogance that trading is easy and that you don't need any financial knowledge to make money. Maybe that was true in the 1980's, but at this point, it requires an incredible level of expertise to generate and implement profitable quant trading strategies.
But all of this is off-topic, to get to the point: the Bloomberg article is garbage click-bait. You sue the General Partners, and I don't think anyone is confused about this.
Are you saying there is no luck in the market any more? Doesn't that mean it's deterministic; and doesn't that mean that market trading is unnecessary as we can allocate needs perfectly?
Active traders are trying to find mis-priced securities (market inefficiencies). That used to be relatively easy for smart people who could think up a few key insights. But now there are so many quant funds deploying fast computers and legions of mathematicians that most new strategies stop working quickly. The act of making those trades drives out inefficiencies.
Market trading is still necessary for liquidity to minimize the cost of capital.
> This is a big red flag. A lot of AI/ML people have this arrogance that trading is easy and that you don't need any financial knowledge to make money. Maybe that was true in the 1980's, but at this point, it requires an incredible level of expertise to generate and implement profitable quant trading strategies.
If they were making the trading bot via training it on past data and what they call 'sentiment signals' generated from social media reaction, etc, do they need experience with trading?
Is there something else that a human trader knows that the bot (which has been trained on past data and accounts for live reactions) doesn't?
Since the bot wasn't implementing a day-trading strategy it's pretty clear the programmers didn't know much about how the markets work. I would guess the training data reflected that knowledge gap. As the saying goes, you don't know what you don't know.
> do they need experience with trading?
That's a yes. You don't get to put a lot of money onto an investment without some experience on how it's done. This is true for all forms of investment, not only trading.
How do you even evaluate to robot if you don't know the market? What do you do if the robot breaks? What do you do if the market breaks (happens once in a while)? How much money you leave under its supervision?
I worked as a quant, and the finance knowledge you need is very deep.
(Pedantic tl;dr: I refuse to put the punctuation inside the quotes. I consider that an outright bug in English and it needs to be stamped out by rebellion.)
What a stupid idea to make "it's" stand for "it is"
"It's" is just the standard use of apostrophe to indicate a contraction.
[0] https://www.merriam-webster.com/words-at-play/history-and-us...
We often hear about the crazy lawsuit culture in US anyway and how insane it is. Whether that is true or not I don't know.
Software has been around for a long time.
At least not where I'm from - is this not the case in the USA?
Years ago, I noticed that my kids’ inner tube had multiple disclaimers, including multiple English disclaimers. The disclaimers next to the Union Flag, Australian Flag, and Canadian Flag were “use under competent supervision.” The disclaimer next to the US Flag was much longer and more detailed (“use under adult supervision” [note that is different from “competent supervision”], “do not tow from boat,” “do not use when drunk,” “do not dive into”, etc.). It did not make me proud to be American.
Additionally, many states have the concept of “joint and several liability,” so that if multiple parties share responsibility for an accident, the victim can collect the full amount from any of them and the perpetrators are expected to pay each other appropriately. The end result is that, generally, the company with the deepest pockets pays the full amount and then hopes to collect from the other parties later.
So, the answer is that you sue the chainsaw manufacturer, the chain manufacturer, the landowner, etc. and hope that they settle or that you get a final judgement that you can collect from the richest one, even if that particular party is held to be, say, 1% responsible.
Nitpick, but you can sue for anything, just not necessarily be successful at it.
This, along with [AIUI] not default providing an award of costs to a successful plaintiff in USA, makes it possible to sue and get a substantial out of court settlement because the risk profile is such that this is cheaper than paying your defence lawyers considering the [mathematical] expectation that you might lose a case that seemingly has no merits.
This is why you get warnings like "chainsaws are dangerous and may cause bodily injury or death, only for use by certified personnel". Then you can't [it is hoped by all reasonable people!] successfully claim that your naivety should have been accounted for, as it was by the warning.
> Over the following months, Costa shared simulations with Li showing K1 making double-digit returns, although the two now dispute the thoroughness of the back-testing.
And I wished the article linked to the fillings or at the least discussed this more thoroughly.
How could you trust him to trade $2.5B?
In this case, the investor was foolish to believe that a magic AI system could generate reliable positive returns. At the same time, it sounds like the performance of the system was misrepresented. Either way, if you're responsible for 1bn of assets, you better do really good homework.
Investing is not a guaranteed return system. You might win, or you might lose all your money. At the end of the day, the only person who is responsible for it is you.
Don't want to lose your money? Don't invest.
There are rules about fraud and fiduciary duty and disclosure which make blanket statements like this incorrect.
A retail investor can't be expected to properly research claims, but a professional should be.
Sigh. Unless you're too big to fail in 2008.
Trump has not declared bankruptcy. He has been owner or part-owner of hundreds of companies and a handful have declared bankruptcy.
As terrible (or not) as that is, there's a stark difference between availing yourself of long-standing bankruptcy law and the just-for-you shenanigans pulled by the federal government in 2008.
The factory owner.
https://www.youtube.com/watch?v=KXnL7sdElno
If a robot doesn’t function properly, you fix it or take it out of service. Taking retribution against a robot (or abstractly, the civilization that created it) is of questionable value. The right question is how do we build robots that don’t lose fortunes?
The fact that human beings get angry at robots and want retribution or recompense for malfunctions is an evolutionary adaptation for dealing with other humans. It is useless when it comes to dealing with non sentient deterministic agents. Sure, if the robot was being controlled by a human, go after the human. If not, what’s the point?
Again, the legal theory works because when dealing with human beings you have to assign responsibility to human beings. Hurricanes and earthquakes kill tons of human beings and there’s no legal theory of responsibility. We simply work as a society to minimize those deaths without blaming anyone.
If that's not your intent, I suggest you rephrase.
When a bridge (another system which is complex but tractable in principle) fails, we work hard to find the feature of the design that led to the failure, and if we decide it was negligence, we exact consequences.
From current news: Should Boeing get a pass on their MCAS failures because "Gosh, it's really hard" ? The MCAS is a robot.
Take Boeing for example. Assigning blame to individuals, while rewarding, may not be as effective as changing the system that allows Boeing to continue to act the way it does. Boeing scapegoats, fires some people (which may include the CEO) but the organization, and the nation and legal system that sustains it continues largely unchanged.
You’re trying to patch by punishing humans individually, while ignoring faults in the system we have collectively created. Our instinct is to assign blame because historically and evolutionarily harm is usually done by individual humans, and punishing humans is a good heuristic. But it does not seem to me that punishment does a good job of creating good large scale organizations or systems.