Chief justice centers Supreme Court annual report on AI's dangers
thehill.com
thehill.com
Document review is already being done by AI, lawyers are using AI to beef up closing arguments and review arguments, AI researchers to go through the relevant cases for citations have been worked on for awhile.
Lawyers being lawyers have immediately reached to "make it illegal" and setting up protections for their trade like they always done. Roberts is commenting that they are going to be disappointed if they think they can get legal protections that X or Y must always be done by a human.
this is one of the huge issues for the unemployed lawyer problem, as those jobs were the entry level for straight of school lawyers. I expect similar problems with dev roles coming soon.
The problem is that we’re going to have people litigating court cases and stupid bureaucratic nonsense within companies by lobbing nonsense that nobody has read at each other. Everyone will wink and nod, but these organizations and institutions will be exponentially more clueless.
As technologists, it’s easy to shrug. What do you do when a judge takes your kid away after using an LLM to read and interpret bullshit generated case notes that a CPS worker generated with another LLM?
Instead what we got instead are surveillance capitalism and now this LLM/AI stuff that becomes much better are emulating human behaviour and languages. UX in software has NOT improved, and I dare to say it has gotten worse. Just feel the performance of iOS 6 on iPod touch vs the lates iOS. Tech over quarter century has just tech getting more complex, and made more easily to centralised and consolidate control. Is this what we really want?
Replace LLM with "clerk" and "generated case notes" with "copy pasted from my standard legally approved phrasing" playbook. Government workers who fill out standard reports already have standard forms pre filled with the results they know they are landing out, with certain sections that differ ready to be edited while the rest stays the same, look at things like warrants as an example. When I worked a government job i was literally handed templates by my supervisor of "pre approved ways to phrase things", that they percieved would help avoid lawsuits or any contest.
With the LLM, everyone will clutch their pearls and be shocked.
At an abstract level, this sounds like a precursor to war. Everyone has competing interests and the peaceful way to resolve them is communication. When that stops force starts.
Maybe the out is that there is some automated sense of listening and agreement.
The legal field does not have a strong cartel at the moment.
In 1940 and in 1970, the population of lawyers was about .13-.16% of people in the US. Currently, it's 3x higher at .39%... after 3 years of declining numbers.
The market is saturated. Large numbers of lawyers can't find enough work, or have moved to other fields.
Edit: Link to report: https://www.supremecourt.gov/publicinfo/year-end/2023year-en...
There is a plethora of things that are minor, don't require going to court, and can be handled via bog standard forms and documents that are just "replace the names" and that lawyers have their paralegals do for them in their entirety. But the paralegal cant go into business on their own can they?
What is the equivalent of the nurse who can get you antibiotics in the US legal system?
A paralegal. In about half the US, said nurse practicioner must be working under a supervising doctor, just like the paralegal works under a supervising lawyer.
So there isn't one.
NPs can independently practice - to some extent - in the other half.
(You'll find a lot of paralegals coming very close to practical indepdendent practice, though. Supervision can be quite theoretical.)
I get what you are trying to communicate - maybe we are over-regulated... yet any time your professional work has the potential to harm your customer if done incorrectly, I don't think having some educational requirements are unreasonable.
The medical field has splintered. Osteopaths kind of invented their own thing and broke into medical doctors' monopoly. Registered Nurses can do some things, but not others. BSNs. CNAs. Nurse Practitioners. Physicians Assistants. Chiropractors. etc. The strength of their cartel has led to attacks by new and novel qualifications, and the compelling need for more medical providers has allowed these attacks to successfully carve out roles.
Some can prescribe medications, others can't. Some are limited to specific medications. Others can perform surgery, others can't. Some can diagnose diseases, etc. Moreover, some qualifications are a step toward a higher qualification, others are traps that don't progress you to an MD/DO at all.
The legal field seems far more egalitarian in that any lawyer can practice any area of law (except patent law), and people can generally represent themselves and their minor children without any qualifications. Folk don't need a lawyer to contest tickets, or to draft contracts that they're a party to, or to negotiate their own settlements, etc.
However, appallingly, the general public is not allowed to prescribe medicine for themselves or their children. Folks cannot buy contacts / glasses on their own, etc. Folks shouldn't need a nurse to get antibiotics for ourselves, and yet we do.
That's news to me.
Edit: When buying for nearsightedness.
This is a massive boon to the training data set. GitHub is also massive, but legal has other systemic advantages as well (e.g. being similar to past work is a structural advantage rather than just a practical one).
I'm not saying machines would do better, but this really lays bare how ridiculous the justice system is.
You should not incentivise people to lie, or give a specific answer.
Nor should you incentivise innocent people to proclaim their guilt
It's like giving someone a longer jail sentence for having a blue aura instead of a pink one.
Because, for instance, accuracy in judgment of sincerity is a basic requirement of getting through any day for any human. In other words, we can certainly judge it accurately enough for many purposes.
Judgement of sincerity is woven into the entirety of court procedures. Why would you exclude this additional relevant data point regarding sincerity of allocution?
"Gee, I'm so sorry I drove drunk, please give me a lower sentence" is so transparently self-interested it should be discarded as evidence.
For millennia, systems of justice—including the one Chief Justice Roberts heads—have been based in no small part on human judgment rather than ironclad objective rules. In a criminal context this leads to leniency as often as harshness.
A purely mechanistic justice system would, I suspect, rapidly be seen to be not worthy of the name.
And that's the entire problem.
A priori certainty of legal outcomes isn't a problem to be "fixed", it's a basic requirement of any reliable legal system.
If purely mechanical solutions lead to bad outcomes, that's a good thing! It means those solutions can be acknowledged as bad and removed or reworked. If instead you arbitrarily make the pain go away because it's convenient you're stuck with a bad solution.
This, by the way, is the kind of non-mechanical, subjective analysis judges and juries are expected to perform all the time.
Suppose it were a voluntary option. What happens when Black people prefer AI judge over the white judge?
How often are we really asking about empathy for a situation vs empathy for social privilege?
It would be a parody of the name. Justice depends on context. The world is too complicated for any set of mechanistic rules to cover every possible context.
Attempts to remove context from justice end badly at best and are often intentionally abused to leverage the “justice” system as a weapon.
But people want to use AI to replace that lawyer, taking them out of the loop completely.
Every fake citations should be on the lawyer who submitted it. That'll end this real quick.
> “Judges, for example, measure the sincerity of a defendant’s allocution at sentencing,” Roberts wrote Sunday. “Nuance matters: Much can turn on a shaking hand, a quivering voice, a change of inflection, a bead of sweat, a moment’s hesitation, a fleeting break in eye contact. And most people still trust humans more than machines to perceive and draw the right inferences from these clues.”
Removing judges from this may well be a good thing. This allows a lot of bias to seep in to the process. Judges have been caught giving vastly different sentences to similar cases where the only real differences were race, gender and background (eg [1]). An extreme example is Brock Turner [2]. That judge should've been impeached.
[1]: https://www.opensocietyfoundations.org/publications/racial-d...
[2]: https://en.wikipedia.org/wiki/People_v._Turner#Sentencing
Inquiring minds want to know because there are certainly more pressing legal issues than AI that can have trouble with simple math.
Let everybody else be disaffected by AI—that’s fine.
I hate this species
Why am I not surprised?
I’m for “you have a job, you automatically get enrolled” medical coverage. But how he went about the ruling was surprising.