From my perspective as a corporate lawyer (ret.), I can easily imagine how the kind of AI described could work in drafting of M&A agreements, registration statements and the like. Many firms, including mine, are already using document assembly software to do some of this, although, to this point, it doesn’t approach the sort of AI capabilities described.
There has been resistance even to this much. I knew late Silent Generation/early Boomer partners who, as recently as ten or fifteen years ago, were still having their secretaries print out their new emails every morning, and then, having read them in hard copy, dictating their responses for the secretary to type up and send out.
Partly this kind of thing is just innate guild conservatism, but there are also professional responsibility issues. How good does the AI need to be so that choosing a reasonable AI package and asking it reasonably adequate questions is a defense to a malpractice claim if the AI makes a mistake or commits some other infelicity that the human in the loop doesn’t catch? Do you have to tell clients that you are using AI? What if a client prefers that you do things the old fashioned way?
The dumb AI that litigators are already using for “electronic discovery” raises similar issues. If you’re using software to do the first review of millions of documents, and the software misses something important, is that your fault? You can’t very well have the program review a million documents and then also review them yourself, just to be sure. You can establish quality control sampling protocols, but then the adequacy, or not, of your protocol can become an issue.
It is little understood outside my particular area of practice, but the availability of public databases, particularly the SEC’s EDGAR platform, that put all the material agreements of every public company and, more importantly, every material agreement that the opposing law firm has drafted for any of its public company clients, online, revolutionized the practice of corporate, M&A and securities law over the last twenty-five years.
Before EDGAR became mandatory for public companies in the mid-90s, every corporate negotiation was bespoke, because it was virtually impossible to find specimen agreements, outside of your own firm’s archives. We fought out agreements draft by draft on the basis of first principles--“we should do it our way because [reasons]”--and negotiating skill. Now, and for at least the last twenty years, negotiations are almost entirely about what is “market” for the particular term or condition under consideration. Opposing counsel knows exactly what you have previously agreed to on that point for your current client, and all your clients, and the entire universe of public company agreements can be accessed an analyzed to determine what people usually do in this situation, which is what is “market.”
It dramatically narrows the scope, tenor and timeline of negotiations. It’s also less fun for the lawyers, which is of course beside the point.
AI - translation software has been like this for some years (and like everything in this vein can be expected to improve steadily). Having the base document in a machine translation saves a good deal of tedium but leaves quite a lot of the more interesting bit still to be done. But I've used it to prepare published documents, where the final stage is to query an expert fluent in the language and knowledgeable about the subject matter, on one or two interesting points.
Some of what is described is just more machine translation - from English to English, or from English to a program language, or from generic English to our own English.
For some purposes the machine translation is perfectly adequate - notably for online reading in general where one may want to go back and check a really botched phrase, or one may not care.
I believe translator's rates per word have gone down since machine translation came in (quite some time ago). What this means for their income is less clear. But the utopia of having half your job done by a computer is more likely to become a shareholders' utopia.
Twitter is an efficient, and for some profitable way to reach others. Nevertheless, posting on Twitter is unethical because it supports one who is acting to destroy democracy and liberalism. Therefore, I am out of here. I hope to see you on the other side.
From my perspective as a corporate lawyer (ret.), I can easily imagine how the kind of AI described could work in drafting of M&A agreements, registration statements and the like. Many firms, including mine, are already using document assembly software to do some of this, although, to this point, it doesn’t approach the sort of AI capabilities described.
There has been resistance even to this much. I knew late Silent Generation/early Boomer partners who, as recently as ten or fifteen years ago, were still having their secretaries print out their new emails every morning, and then, having read them in hard copy, dictating their responses for the secretary to type up and send out.
Partly this kind of thing is just innate guild conservatism, but there are also professional responsibility issues. How good does the AI need to be so that choosing a reasonable AI package and asking it reasonably adequate questions is a defense to a malpractice claim if the AI makes a mistake or commits some other infelicity that the human in the loop doesn’t catch? Do you have to tell clients that you are using AI? What if a client prefers that you do things the old fashioned way?
The dumb AI that litigators are already using for “electronic discovery” raises similar issues. If you’re using software to do the first review of millions of documents, and the software misses something important, is that your fault? You can’t very well have the program review a million documents and then also review them yourself, just to be sure. You can establish quality control sampling protocols, but then the adequacy, or not, of your protocol can become an issue.
It is little understood outside my particular area of practice, but the availability of public databases, particularly the SEC’s EDGAR platform, that put all the material agreements of every public company and, more importantly, every material agreement that the opposing law firm has drafted for any of its public company clients, online, revolutionized the practice of corporate, M&A and securities law over the last twenty-five years.
Before EDGAR became mandatory for public companies in the mid-90s, every corporate negotiation was bespoke, because it was virtually impossible to find specimen agreements, outside of your own firm’s archives. We fought out agreements draft by draft on the basis of first principles--“we should do it our way because [reasons]”--and negotiating skill. Now, and for at least the last twenty years, negotiations are almost entirely about what is “market” for the particular term or condition under consideration. Opposing counsel knows exactly what you have previously agreed to on that point for your current client, and all your clients, and the entire universe of public company agreements can be accessed an analyzed to determine what people usually do in this situation, which is what is “market.”
It dramatically narrows the scope, tenor and timeline of negotiations. It’s also less fun for the lawyers, which is of course beside the point.
AI - translation software has been like this for some years (and like everything in this vein can be expected to improve steadily). Having the base document in a machine translation saves a good deal of tedium but leaves quite a lot of the more interesting bit still to be done. But I've used it to prepare published documents, where the final stage is to query an expert fluent in the language and knowledgeable about the subject matter, on one or two interesting points.
Some of what is described is just more machine translation - from English to English, or from English to a program language, or from generic English to our own English.
For some purposes the machine translation is perfectly adequate - notably for online reading in general where one may want to go back and check a really botched phrase, or one may not care.
I believe translator's rates per word have gone down since machine translation came in (quite some time ago). What this means for their income is less clear. But the utopia of having half your job done by a computer is more likely to become a shareholders' utopia.
Twitter is an efficient, and for some profitable way to reach others. Nevertheless, posting on Twitter is unethical because it supports one who is acting to destroy democracy and liberalism. Therefore, I am out of here. I hope to see you on the other side.