Judge decides case based on AI-hallucinated case law

[Dannyalcatraz]

What's stopping them from making a new law that says that all current rules are null and void and just start over with 'make sure people don't get killed by defective buildings' and a short list of valid objections and all made public to both officials and the people in general? This is something I find odd and frustrating with common law, the need to pay attention to thousand years old precedent to the point now people need AI bots to check them all.

Ok, here in my country we still had a Visigothic code of commerce originally made in the seventh century a hundred years ago. But it has been long gone and nobody can use it to make up a case of law today.

Most countries that use common law don’t have significant laws that old that are actually still valid.

And with modern research tools, it’s pretty easy to find it when when which laws were voided, overturned, revised, etc.

 

[Jfdlsjfd]

It's not specific to common law to have complicated exemptions. Because oversimplifying texts will either lead to unwanted side effect -- for example, if you forgot to add an earthquake exemption to the "don't build houses that kill you", then a lot of people will go to jail when an earthquake happens -- and also, "building houses" isn't just a matter of physical injury. You might want rules to protect the environment (don't use pollutant in the building), the historical value of another building (don't build a jarring gaudy building just in front of a historically significant building), the rights of neighours (if you buy a house in a quiet residential district, you might be able to convince politicians that you're harmed if a heavy industry is built next to your house)...

The most recent rewrite of the building code in France, a purely civil law country, dates back from 2018 and it isn't simpler than common law country regulations. I'd guess it's even more stringent than most common law countries.

WRT to old regulations, it can happen in civil law countries, too: I still find very funny to pass judgements mentionning a 1539 royal edict -- even in civil law countries, you can get some surviving pieces of legislation. Basically, it is unwise to say "everything until now is void and here is the new system": you might mistakenly make legal some awful things. When the French authorities in 1945 wanted to undo the legislations passed during WWII by both the occupying German forces and the collaborationist Vichy government, they cancelled a few specific texts but dared not undo everything without prior individual review: saying "everything done by unlawful authorities is void" was considered, but unlawful authorities did a lot of harmless and even useful things, like deciding on local speed limit, and they didn't want to just cancel everything.

 

[môrgondír]

The law is a bachelor.

The law is a dagger that is gripped by the blade.

In practice, the law favors class, wealth, privilege and other factors besides. Without those, you're down to pure luck or a sympathetic judge, and hopefully some nice people in the jury box to gamble with your fate when you show up at the casino in your suit and tie, hoping that you look sincere.

One may choose or not to interact with AI chatbots - no one except litigants, lawyers, and judges chooses to interact with the law - everyone else would rather be doing anything else.

 

[Cergorach]

Yes.
And note how folks don't mistake a hammer for an all-purpose tool? That's the point.

The whole quote: "it is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail."
It's not for nothing that there's a 150+ year old term for it, a "Birmingham screwdriver". More info: Law of the instrument - Wikipedia
Just because people (should) know better, doesn't mean they actually act better...

But, really, you don't only have a hammer. You have a human brain.
Right?
RIGHT??!??!

Sure, people have had human brains for thousands of years, but they still kept using that hammer for everything. And that's also the case with LLM and image generation, people get one tool and want to do everything with it, not asking if that's the best tool for the job, not asking if the output is actually desirable or correct, it just seems correct. These tools give power to people, real and imagined power... It gives people that don't have the skills, the willingness or ability to learn, a tool that would seemingly allow them to do things they normally couldn't.

Compare it to cars becoming popular (in the US), sure people could still walk, bike or use a horse, but going long distances made cars more desirable. Cars were dangerous to people, unclean, unhealthy, expensive, required a lot of expensive infrastructure, etc. People aren't even that skilled at driving... Looking specifically at the US, it's still a very car centric country, many countries aren't anymore, even though they initially followed the US example after WW2... I expect similar things to happen all over the world, countries/cultures that will rely heavily on LLM and image generation, while others won't...

And what people call 'AI' is just the next step in the industrialization, automation trend of the last 250+ years. And you could go even further back where people were replaced by animals, animals replaced by primitive machines (windmills for example), etc. in the two decades before LLM and image generation became common, I was already hard at work replacing people with computer systems/processes, heck often even replacing myself. But now suddenly LLM and image generation are 'problematic' amongst a certain subset of the population... But I remember protests in my youth when factories got automated, the world didn't end, those people got new jobs. Heck, around here unemployment is now significantly lower then in my youth.

 

[Jfdlsjfd]

The law is a bachelor.

The law is a dagger that is gripped by the blade.

In practice, the law favors class, wealth, privilege and other factors besides.

And yet, “Between the strong and the weak, between the rich and the poor, between the master and the servant, it is freedom that oppresses and the law that liberates.” (Lacordaire)

 

[Jfdlsjfd]

The whole quote: "it is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail."
It's not for nothing that there's a 150+ year old term for it, a "Birmingham screwdriver". More info: Law of the instrument - Wikipedia

Thank you for pointing out this link (and for allowing me to put a source on the quote).

And what people call 'AI' is just the next step in the industrialization, automation trend of the last 250+ years. And you could go even further back where people were replaced by animals, animals replaced by primitive machines (windmills for example), etc. in the two decades before LLM and image generation became common, I was already hard at work replacing people with computer systems/processes, heck often even replacing myself. But now suddenly LLM and image generation are 'problematic' amongst a certain subset of the population... But I remember protests in my youth when factories got automated, the world didn't end, those people got new jobs. Heck, around here unemployment is now significantly lower then in my youth.

The question that remains is, "do we still have unfilled jobs to fill?". When we replaced our hunters-gatherer with peasants, it freed manpower to occupy jobs as shaman and soothsayers that we needed greatly. When we had enough of them and they invented a lot of new technology with their free time, we had a hard time turning enough peasants into craftsmen, until agricultural progress (and demographic increase) allowed for more peasants to turn into workers. With automation, we freed workers to work into a needed service industry (we could easily use more servants) without losing the manufactured goods. When we'll free people from doing service jobs, the question is "what will we make them do?" The answer can be "nothing", "jobs aptly described by David Graeber that are forbidden to name here by the anti-anthropology filter", or "a new wave of yet-to-be-invented useful purpose". All three have been explored in science-fiction.

While the emergence of new needs has always worked so far and is certainly something that should assuage the fears of our time, it is not possible to rule out the other possibilities. I for one would very much like to live long enough to see humanity freed from the need to work.

 

[Umbran]

But now suddenly LLM and image generation are 'problematic' amongst a certain subset of the population...

The objections to generative AI are recent, not "sudden", and are so because the concerted pressure for its widespread corporate adoption is similarly recent. OpenAI only started trying to monetize ChatGPT in February 2023, for example.

And, objection to technology taking away people's jobs is hardly "sudden" - it is as old as technology taking people's jobs.

But I remember protests in my youth when factories got automated, the world didn't end, those people got new jobs. Heck, around here unemployment is now significantly lower then in my youth.

First off, you are comparing now to your youth. I don't know how old you are, but that's probably... 20+ years, yes? So, you are saying a generation later things are good. That skips over the ones directly impacted by the change, which is very convenient.

And it is common - it is easy to refer to the long-term, macro-economic situation ("the world didn't end") and elide over the detailed human costs.

I don't know precisely which events you are speaking of, so I won't try to address them. But, broadly and historically speaking, major job dislocation isn't just, "They find another job." Jobs that they are trained for go away, and don't exist. For many, if a next job exists, it is a lower-paying role. Job dislocation will cause many to spend their savings, so families lose generational wealth, or ability to retire. Folks lose their healthcare, their housing. Substance addiction rates among those impacted rise.

The model to look at here is Detroit. Yes, today, Detroit is doing pretty well. But, after the automobile industry crashed in the 70s, it took decades for the city to recover, which implies tens to hundreds of thousands of lives were impacted for a generation and more.

Or, we can look at Appalachian coal country, which has not really recovered from the loss of the coal and steel industries.

By no means am I saying that the world should not adopt technological change. I am saying that wonton adoption causes harm that we can mitigate if we actually approach things thoughtfully. You would see much less resistance to change if you actually offered those who will be impacted something to help them through that change, or you targeted the change to boost, rather than replace, the people impacted.

Until you actually think in terms of managing change, you generate, and earn, resistance.

 

[The Firebird]

My impression is that we're past the peak of the AI hype cycle. I don't see as many people trying to use it everywhere and there is a better understanding of the limits and the use cases.

 

[Jfdlsjfd]

I don't know precisely which events you are speaking of, so I won't try to address them. But, broadly and historically speaking, major job dislocation isn't just, "They find another job." Jobs that they are trained for go away, and don't exist. For many, if a next job exists, it is a lower-paying role. Job dislocation will cause many to spend their savings, so families lose generational wealth, or ability to retire. Folks lose their healthcare, their housing. Substance addiction rates among those impacted rise.

The model to look at here is Detroit. Yes, today, Detroit is doing pretty well. But, after the automobile industry crashed in the 70s, it took decades for the city to recover, which implies tens to hundreds of thousands of lives were impacted for a generation and more.

Thing is, those individual situations are not showing at a more macro level, so for every family that lived worse due to the change, another family lived much better than before, so it is more than compensating overall. It is of course of no comfort to the family that is living worse, but that's why societies, in general, have developped public policies to mitigate the widening of inqualities.

And despite what it seems, it is somewhat working. Here is a graphical representation of the part of the income of the 1% wealthiest in the total revenue by country, over time:

Inequalities between the wealthy class and the average population decreased worldwide from 1870 (before the graph, but trust me) and roughly 1975. Then it rebounded slightly, more in the US (dark blue line) than elsewhere, though.

Focusing on average households, the same observation can be made: even excluding the top 10% of wages, the average, inflation-adjusted wage increased by 43,7% between 1979 and now in the US according to the economic policy institute. Being a measurement of wages and not overall revenue, it isn't affect by the capital gains that explains a lot of the rebound of the dark blue line in the above graphic.

It doesn't mean nobody in Detroit suffered, it means that on average, workers got wealthier, so any observation of a single family being negatively affected is more than compensated, overall, by another family getting richer at the same time (ie, for one affected family in Detroit, there is a family that is living much better in the Silicon Valley).

What makes the situation difficult, and which colors one's perception, is that this situation might not be mitigated by social safety nets to ensure that the mass of wealth created by technology is shared equally. A problem to which different countries have responded with different intensity, and different means:

This graphs shows that social spending as percentage of GDP has the US in the second highest position when it comes to net social security spending, yet ranks at 23rd position in public welfare spending. Basically, it shows that while there is a lot of collective wealth used for social security, it has the widest gap between public spending and net spending, the difference being tax breaks with a social purpose. Of course, the latter doesn't help the poorest (who pays very little taxes in the first place) but helps first and foremost those who benefit from the evolution in the job market (not only are you getting richer, but public policies helps you save money for your retirement).

Basically, the technological progress since the 1950s didn't make anyone suffer in Detroit. It made people richer overall, and the focus of redistributive policies allowed some people in Detroit to suffer. Cum hoc, send non propter hoc. The pie got bigger, but some families in Detroit were denied a part of the pie.

(There is even a possibility that even the suffering families got better, but since poverty relies on a comparison to the mean, the feeling of living worse got widespread even if the situation objectively stayed close to what it was before. If everyone around gets richer, staying at the same level will make one feel disclassed, for no longer being part of the middle class.)

Or, we can look at Appalachian coal country, which has not really recovered from the loss of the coal and steel industries.

By no means am I saying that the world should not adopt technological change. I am saying that wonton adoption causes harm that we can mitigate if we actually approach things thoughtfully.

Sure, and public welfare policies is the way to adopt wealth-creating technologies while protecting those who will be affected. Approaching things thoughtfully might require a political approach and has no link with technology adoption (you can get large inequalities without a lot of tech, though you need obviously some tech to allow for inequalities to happen -- hunter-gatherers were notoriously equal).

You would see much less resistance to change if you actually offered those who will be impacted something to help them through that change, or you targeted the change to boost, rather than replace, the people impacted.

Until you actually think in terms of managing change, you generate, and earn, resistance.

But being a proponent of technology (or an enemy) has no bearing on whether individual countries will choose to mitigate and manage the change. The rational approach is to boost wealth-increasing techs and mitigating policies so the pie gets bigger and nobody's share of the pie gets smaller. The opposite stance sounds like fighting the cure for cancer because only a few will be able to afford it -- the "right" answer to me is not to stop cancer research, but to invent the NHS.

The pie-sharing outcome can range from Gini-ideal situations of 1 (everything belong to a single citizen) or 0 (everyone gets the exact same share) or any situation in-between. It is disconnected to technological progress and dependant only on our collective choices on how to cut the pie (in democratic countries).

 

[môrgondír]

I think big tech wants AI out in the world, semi-wild and interacting with people, partially to train AI, but also specifically for the purpose of better identifying particular use-cases, optimally ones that can be monetized with little or no backlash because AI becomes both very routine to use, and because it provides desired utility.

Big tech will gather data based on all these interactions, and then they'll decide how best to proceed.