4 ms·

Programmer will probably should not become lawyers. But an AI will (after more advances in machine reading, and NLG). I.e. it will not replace lawyers, but wil

Programmer will probably should not become lawyers. But an AI will (after more advances in machine reading, and NLG). I.e. it will not replace lawyers, but will enable 10x productivity, which means that you will need 1/10 less lawyers.

dragonwriter 7y ago
> I.e. it will not replace lawyers, but will enable 10x productivity, which means that you will need 10x less lawyers. Lawyers are countable, so “fewer”, or, even better “1/10 as many”. But the argument is flawed other than grammatically, as the quantity of legal services demanded isn't fixed, and if they are ten times more productive that means they become viable for tasks where they wouldn't have been worth employing before...and the total quantity of legal services consumed goes up.
- streetcat1 7y ago
  Sure. My bad. Been coding for 12 hours... So introducing AI is not unlike the innovator dilemma, it will push human lawyers up and up. However, I think that even if the demand is not fixed, most of it is for basic legal services.
  - dragonwriter 7y ago
    > However, I think that even if the demand is not fixed, most of it is for basic legal services. I dunno, public defenders’ work is in a way the most basic of legal services, and from everything I've seen their time is rationed pretty heavily because there simply aren't enough resources. There’s a lot of room to translate more productivity into more getting done rather than fewer people doing the same level of work. (OTOH, I'm not all that optimistic about AI short of AGI helping that much in the aspects needed in that role, because I think the limiting factors there are the harder to automate aspects of lawyering, not the more mechanical ones.)
    - streetcat1 7y ago
      I personally think that the problem with AI today is deployment and not research. You see a lot of very busy lawyers not because AI cannot help, but because it is not deployed. The same applies to health care. Even if we come up with super human models for cancer detection, they are not deployed. The problem is that in order to deploy AI you need a lot of very costly talent (costing as much as the lawyer itself). This will be solved first by AutoML / Auto deploy. I.e. once automl is working, and each professional can create and deploy models, than will see the productivity gains.
- dllthomas 7y ago
  Related: https://en.m.wikipedia.org/wiki/Jevons_paradox https://en.m.wikipedia.org/wiki/Jevons_paradox
  - streetcat1 7y ago
    I am afraid that this does not apply to intellectual work since the marginal cost of new AI is 0. Hence, in theory, any new work given to lawyer, will be soon covered by more AI models. Another effect of deploying AI, is that once a model learn something, ALL the other models become better - this does not occur in nature. I.e. once a human doctor fix a mistake other doctor can repeat this mistake. However, once an AI doctor fix a mistake, all the other AI doctors learn from this mistake.
    - dllthomas 7y ago
      The context was (explicitly) not AGI that can fully replace lawyers, but AI technologies that make human lawyers more effective. In that setting, the Jevons paradox would still seem to apply (until we actually saturate demand for anything human lawyers can do).

dragonwriter 7y ago

> I.e. it will not replace lawyers, but will enable 10x productivity, which means that you will need 10x less lawyers. Lawyers are countable, so “fewer”, or, even better “1/10 as many”. But the argument is flawed other than grammatically, as the quantity of legal services demanded isn't fixed, and if they are ten times more productive that means they become viable for tasks where they wouldn't have been worth employing before...and the total quantity of legal services consumed goes up.

streetcat1 7y ago
Sure. My bad. Been coding for 12 hours... So introducing AI is not unlike the innovator dilemma, it will push human lawyers up and up. However, I think that even if the demand is not fixed, most of it is for basic legal services.
- dragonwriter 7y ago
  > However, I think that even if the demand is not fixed, most of it is for basic legal services. I dunno, public defenders’ work is in a way the most basic of legal services, and from everything I've seen their time is rationed pretty heavily because there simply aren't enough resources. There’s a lot of room to translate more productivity into more getting done rather than fewer people doing the same level of work. (OTOH, I'm not all that optimistic about AI short of AGI helping that much in the aspects needed in that role, because I think the limiting factors there are the harder to automate aspects of lawyering, not the more mechanical ones.)
  - streetcat1 7y ago
    I personally think that the problem with AI today is deployment and not research. You see a lot of very busy lawyers not because AI cannot help, but because it is not deployed. The same applies to health care. Even if we come up with super human models for cancer detection, they are not deployed. The problem is that in order to deploy AI you need a lot of very costly talent (costing as much as the lawyer itself). This will be solved first by AutoML / Auto deploy. I.e. once automl is working, and each professional can create and deploy models, than will see the productivity gains.
dllthomas 7y ago
Related: https://en.m.wikipedia.org/wiki/Jevons_paradox https://en.m.wikipedia.org/wiki/Jevons_paradox
- streetcat1 7y ago
  I am afraid that this does not apply to intellectual work since the marginal cost of new AI is 0. Hence, in theory, any new work given to lawyer, will be soon covered by more AI models. Another effect of deploying AI, is that once a model learn something, ALL the other models become better - this does not occur in nature. I.e. once a human doctor fix a mistake other doctor can repeat this mistake. However, once an AI doctor fix a mistake, all the other AI doctors learn from this mistake.
  - dllthomas 7y ago
    The context was (explicitly) not AGI that can fully replace lawyers, but AI technologies that make human lawyers more effective. In that setting, the Jevons paradox would still seem to apply (until we actually saturate demand for anything human lawyers can do).