In a nutshell, Sannikov’s theories describe how an employer would adjust a contractor’s compensation in real time, in response to changing performance and external conditions. Where most previous researchers only allowed contracts to be updated at regular intervals, Sannikov described what happens when they can be changed infinitely fast.
Not just like hell, but illegal to in the context you describe. Contracts require both parties to agree to the terms--and to subsequent changes to those terms. Sannikov's proposal is DOA, though based on the article it was probably intended more as a thought exercise than as an actual proposal for the future of law.
Don't forget this is already done for trading commodities, stocks, bonds and their derivatives. As well as online ad slots and who knows what else. Sannikov's simply thought about extending that model to employment relationship. It would have, I think, the effect of transforming them in contracting relationships, even before HFTing the labor itself.
Overall cool article, but it seems to blow by the fact that lawyers aren't expected, or usually practically able, to write complete contracts for the solution space (the space being all potential states of the relevant portions of the universe at time of enforcement).
When they do they're for relatively trivial contracts, or for standardized contracts (think a commodities future) that purposely limit the relevant universe to the easily measurable.
I don't think we're going to see HFL compete with human lawyers in the way that HFT competes with human traders anytime soon. To the extent that dynamically rewritten contracts become a thing I think we'll largely view them as an iteration of HFT actually.
On a partially related note I was discussing a law-tech idea with a lawyer friend who pointed out the (US at least) issue that law firms have to be owned by attorneys. A lot of potential legal tech ideas then basically are on thin ice if the tech's behavior is close enough to the line of lawyering. The HFL described would certainly run into that issue (at least enough for one case to set precedence).
Every time I see one of these 'XYZ start-up is going to change the legal world' I shake my head. There seems to be this misguided idea that lawyers are essentially programmers, and that legal work can be entirely automated. There is a huge amount of humanity in law. It is in essence a codified system of dispute resolution.
Legal work involves a huge input of social capital and human-to-human communication: negotiating, interviewing, mediating disputes, arguing, schmoozing clients, networking with lawmakers, smoothing over conflicts with clients and other lawyers, etc... I don't see computers ever replacing this aspect of legal work, and no, AI is not the answer for the same reasons that AI is not a substitute for friends or significant others.
Clients with millions of dollars on the line will never want to place their fate in the hands of an algorithm. They want a reputable team that they can talk to and place their trust in. This is why large corporations continue to use expensive 'name brand' law firms even though much cheaper firms could do the same work. Same goes for criminal law: do you want to place your liberty (and life if you're an American citizen) in the hands of an algorithm? Anyone who has been confronted with the power of the criminal justice system will tell you that its scary. You need a human you can talk to, who can explain and reassure and humanize this otherwise cold and harsh system.
I'm not going to name names, but I've seen some start-ups who offer contract templates and claim that you'll save oodles of money on legal work by using their contracts. First of all, if you haven't studied contract law, how do you that what you're paying for is actually useful? Are you going to ask a lawyer to review it for you? Secondly, if there's ever a dispute over one of these contracts, you're going to have to hire a lawyer anyway. Finally, judges tend to frown on template contracts. A custom contract will always come across as stronger in the mind of a judge or arbitrator.
The law is not binary. It is murky and cloudy and flawed. There is rarely a 'right' answer to legal questions. Judges will frequently make seemingly arbitrary decisions. At this point in time there is simply far too much humanity integrated into law for computers to deal with.
However I do see lots of potential for technology to make legal work faster and more efficient. Programmers are light-years ahead in their work-flow and tools. Whereas programmers take pride in having the slickest workflow, lawyers tend to be allergic to software and hate having to learn new systems. They bang away at buggy old versions of Microsoft Word since they hate having to learn new versions. They use bizarre and inefficient methods to collaborate and share documents. In fact, an IDE for legal work that combined version control, word processing, time accounting for billable hours and access to research databases would be the killer app imo. Also, encrypted email and chat that is extremely user-friendly would be a big big hit among lawyers.
> Clients with millions of dollars on the line will never want to place their fate in the hands of an algorithm.
They will want, if it will unlock multi-billion market opportunities. It happened multiple times already, e.g. algo trading robots replaced some human traders.
Business service robots will bring forward
the day when we need to think hard about
how to spread the massive wealth created by
the automation of everything.
So instead of allocating half of all wealth to the top 1% we'll allocate it to the top 0.1%?
17 comments
[ 2.6 ms ] story [ 39.8 ms ] threadWorth the read.
In a nutshell, Sannikov’s theories describe how an employer would adjust a contractor’s compensation in real time, in response to changing performance and external conditions. Where most previous researchers only allowed contracts to be updated at regular intervals, Sannikov described what happens when they can be changed infinitely fast.
So... Next step: Giving power of attorney to a software agent. (What could possibly go wrong?)
But the time and money saved is just too appealing
When they do they're for relatively trivial contracts, or for standardized contracts (think a commodities future) that purposely limit the relevant universe to the easily measurable.
I don't think we're going to see HFL compete with human lawyers in the way that HFT competes with human traders anytime soon. To the extent that dynamically rewritten contracts become a thing I think we'll largely view them as an iteration of HFT actually.
On a partially related note I was discussing a law-tech idea with a lawyer friend who pointed out the (US at least) issue that law firms have to be owned by attorneys. A lot of potential legal tech ideas then basically are on thin ice if the tech's behavior is close enough to the line of lawyering. The HFL described would certainly run into that issue (at least enough for one case to set precedence).
Legal work involves a huge input of social capital and human-to-human communication: negotiating, interviewing, mediating disputes, arguing, schmoozing clients, networking with lawmakers, smoothing over conflicts with clients and other lawyers, etc... I don't see computers ever replacing this aspect of legal work, and no, AI is not the answer for the same reasons that AI is not a substitute for friends or significant others.
Clients with millions of dollars on the line will never want to place their fate in the hands of an algorithm. They want a reputable team that they can talk to and place their trust in. This is why large corporations continue to use expensive 'name brand' law firms even though much cheaper firms could do the same work. Same goes for criminal law: do you want to place your liberty (and life if you're an American citizen) in the hands of an algorithm? Anyone who has been confronted with the power of the criminal justice system will tell you that its scary. You need a human you can talk to, who can explain and reassure and humanize this otherwise cold and harsh system.
I'm not going to name names, but I've seen some start-ups who offer contract templates and claim that you'll save oodles of money on legal work by using their contracts. First of all, if you haven't studied contract law, how do you that what you're paying for is actually useful? Are you going to ask a lawyer to review it for you? Secondly, if there's ever a dispute over one of these contracts, you're going to have to hire a lawyer anyway. Finally, judges tend to frown on template contracts. A custom contract will always come across as stronger in the mind of a judge or arbitrator.
The law is not binary. It is murky and cloudy and flawed. There is rarely a 'right' answer to legal questions. Judges will frequently make seemingly arbitrary decisions. At this point in time there is simply far too much humanity integrated into law for computers to deal with.
However I do see lots of potential for technology to make legal work faster and more efficient. Programmers are light-years ahead in their work-flow and tools. Whereas programmers take pride in having the slickest workflow, lawyers tend to be allergic to software and hate having to learn new systems. They bang away at buggy old versions of Microsoft Word since they hate having to learn new versions. They use bizarre and inefficient methods to collaborate and share documents. In fact, an IDE for legal work that combined version control, word processing, time accounting for billable hours and access to research databases would be the killer app imo. Also, encrypted email and chat that is extremely user-friendly would be a big big hit among lawyers.
They will want, if it will unlock multi-billion market opportunities. It happened multiple times already, e.g. algo trading robots replaced some human traders.
Roomba, Intel, ibm etc.
There won't be revolution. There will be just rich and poor.