Rov_Scam
18hr ago
And LLMs are horrendously incapable of returning all matches for a query, even a super simple one.
Good God, that's a frustration that I totally forgot about in my various AI criticisms. I have, on multiple occasions specifically asked it to "Provide a comprehensive list of X", which list would include dozens of items, and it instead provides 5 examples, complete with shitty summaries I didn't ask for, which examples are the most obvious and well-known examples (obviously pulled from the Wikipedia page) that I wouldn't need a trillion-dollar technology to find.
Rov_Scam
1d ago
I had my first "AI at work" experience the other day when I sat through a luncheon meeting presented by a rep for one of the big legal research companies. It was billed as a continuing education event but was really just a sales pitch for their AI products. The guy was able to cite two uses for AI in the legal field:
- Legal research
- Pulling information from documents (in this case up to 5000 pages)
That's all well and dandy, but I don't do either of those things very often. This wasn't presented as "the technology is quickly changing and you'll be able to do more in the future" as much as "this is all you can do within the bounds of ethics and without exposing yourself to a malpractice suit". The idea that law firms will consist of a few partners handling a suspiciously large number of cases by prompting AI to generate outputs is pure fantasy. The people who think that AI will take over everything do so on the assumption that all work boils down to a set of deliverables that simply need to be generated, when that's not the case. If I'm looking to generate deliverables, I can already have a paralegal do all the drafting and research and just put my name on it, because there's nothing that says you need a law license to do legal research or draft documents.
What the client is paying for is for someone to take responsibility for the case, and it would be irresponsible of me to "handle" a case about which I knew nothing. Most of my time is spent reviewing and analyzing facts. Sure, an AI may theoretically be possible that can determine what's relevant and formulate a strategy better than I can, but the AI is not going to be responsible for its output. I'm never going to trust AI with tasks I wouldn't trust to support staff, no matter how much I trust my support staff (and they're great, btw), because the client doesn't want to hear about how it's the paralegal's fault. If I allow AI to do all my work for me, and I go into negotiations missing something, that's a pretty big matzo ball hanging out there. It's not that I'm perfect, or even necessarily better than AI theoretically could be, but the client is ultimately trusting me to make the relevant decisions, and I can't make them without a thorough knowledge of the case. It's the same problem with autonomous vehicles. I said a decade ago that they would never catch on, not because of any technical limitation, but because auto manufacturers aren't going to take responsibility for them. We've already seen this with Tesla being very aggresive in their defense of lawsuits stemming from autopilot. I don't necessarily disagree with Tesla's stance on this as things stand now, but if a vehicle is truly autonomous then an accident isn't caused by negligence on the part of the driver but on products liability on the part of the seller and manufacturer. As long as auto makers take the stance that the owners of vehicles are ultimately responsible for them, true AVs will never exist.
The other big issue is data security. You can tell me all day long about how great Claude, or ChatGTP, or Gemini are, but in the legal world using any of these is a complete nonstarter. Any lawsuit is going to deal with confidential data, and some suits are going to deal with little but confidential data. At the very minimum, we need to use settlement histories to evaluate potential settlement value of a case. Google literally built its business around data harvesting, and the tech sector as a whole doesn't have a stellar reputation for protecting client data. Regardless of whatever "opt out" provisions are allegedly in place, no law firm in their right mind would take the risk of feeding reams of data into a chatbot if there's any risk whatsoever that that information will show up later in a chatbot response. And no, this isn't the same as companies feeding their proprietary code bases into chatbots; the code's confidential status is subject to the discretion of management. An attorney does not have the discretion to reveal confidential information, especially if that information will be harmful to the client in the wrong hands.
This is before you even get to the fact that the current technology is underwhelming even for legal research. It looks good in demos but as soon as you try to use it for anything it proves its inadequacy. For document summarizing, 5,000 pages sounds generous, but it's rare that I'm concerned about finding information in a single document. I said this in a comment last week, but the utility would be more like "search all the depositions we have on file and pull all the ones where a witness testified about X". Well, we have tens of thousands of depositions on file, most in PDF but some in a special format used for court transcriptions. Conservatively assuming 100 pages per depo and 140 words per page, that comes out to something like half a billion tokens of context required, before we even consider that PDFs take more tokens than plain text, and a lot more if they haven't been OCR'd (which most of these haven't). Even the document functions described by the sales rep weren't that good; the example he gave was that if you were searching medical records for mentions of cancer it could broaden the search to include mentions of specific cancers.
Rov_Scam
3d ago
I can't find any evidence that this was a honeypot operation. She had some role in his campaign and was photographed with him, but I haven't seen any evidence of a romantic relationship.
Rov_Scam
4d ago
Well, that's the problem, since it's only worth doing if I can feed in all the transcripts at once. If I knew which transcripts I needed, I wouldn't need an LLM to tell me! We're talking literally tens of thousands of transcripts here, most of them well over 100 pages. If I have to curate them to fit inside a reasonable context window, then I've already done 99% of the work, since it only takes a few minutes for me to look in the index and see if there's any relevant testimony. Even by my conservative calculations, at $5/token it would be prohibitively expensive to do anything, even if a large enough context window existed, and with that much context the LLM's accuracy would start to break down pretty quickly.
Rov_Scam
5d ago
I haven't used AI for work and I don't know of anyone who does. I honestly don't know what I would even use it for. I guess I could theoretically load deposition transcripts in case I needed to see if there was one taken in the past where a witness said something I could use, but that would literally require millions of tokens of input context, assuming it was even capable of handling such a request, and the utility of that would be limited, i.e. I'd do it if it were cheap enough but there's no way it would be cheap enough. People bring up research a lot and it might be useful there, but I do research like twice a year.
Rov_Scam
5d ago
To illustrate the point from my own retail experiences from years and years ago: We had to do customer service training every year. I worked the service desk a lot, and my first year there I was always baffled by the manager's willingness to give refunds for stupid shit. For example, it was a grocery store, and we sold deli pizzas that you took home and made yourself. Someone tried to return one that 1. Was already cooked and 2. Had two pieces left. The couple's stated reason for the return was that it "wasn't as good as we remembered it being". I had to call the manager because I wasn't allowed to refuse refunds (this wasn't normally an issue since most refunds were pretty routine), and I was incredulous when he gave them store credit.
It wasn't until they started the customer service trainings that I realized that $3 was a small price to pay to keep from pissing these people off. They shopped there every week and weren't constantly returning items, and it would probably cost the store a lot more in the long run if they decided to go somewhere else. We had already disappointed them with the pizza, after all. Add to it the fact that stores will spend huge amounts on advertising without even thinking about it and then try to nickle and dime the customers as soon as they get into the store. I was told that we needed to provide an absolutely flawless experience to the extent possible. If someone asked where an item was we weren't allowed to tell them; we had to walk them to the location. The thing is, it's not like it was that great of a store or anything. Good service is just a customer expectation, and if you can't provide it, and can't make up for it in other ways (like having rock bottom prices), people will take their business elsewhere.
Rov_Scam
5d ago
Oh, you can find out from the address which schools your kid would go to, there just isn't a way to make a map that makes any sense. I just checked it out, and if you live in the trendier parts of the East End like Friendship you will not only be going to high school with kids from the Hill District but the school is actually in the Hill District. Honestly, though, it's not that far from where Schenley was, so it does make historical geographic sense.
Rov_Scam
5d ago
Right now the churn is more in recent grads leaving for what they perceive to be greener pastures. The problem right now isn't so much a shortage of lawyers as it is a shortage of experienced lawyers. I work at a smaller firm, and just a couple months ago a younger guy who clerked for a judge after law school and whose wife works across the street from us quit to take a different job. I don't even know if the pay is any better, but it seems like everyone under the age of 35, and several people who are older, think that whatever job they're doing is unsatisfying and wants to do something closer to what they imagined the practice of law would look like. In the meantime, we can't find anyone to replace these people. Hiring out of law school only makes the situation worse because it takes several months to get an attorney to the point where they're actually making money for the firm, and they're unwilling to do that for a guy who is going to bolt in six months.
Rov_Scam
5d ago
What are the schools for each of these neighborhoods?
Nobody knows. I mean that literally; the district doesn't publish feeder maps and an independent effort to produce them 15 years ago resulted in a complicated patchwork of school zones. The district proposed a controversial consolidation plan a couple years ago that would close a bunch of schools, but they had no idea what effect it would have on the feeder patterns. So they hired a consulting firm from Boston that put professional demographers on the case and even they couldn't make heads or tails of the situation and couldn't draw a new feeder map that would make sense based on the schools they wanted to keep open. The problem arises from two issues. The first is that Pittsburgh half-assed their integration plan compared to other cities. White parents were opposed for the usual reasons, but the preferred solution of black parents was to bus white students in to their schools so their own kids wouldn't have to leave the neighborhood. The school board decided that only black students would be bussed, a solution that satisfied nobody, and they tried to avoid bussing as much as possible by implementing magnet schools, which took a school building "out of circulation", so to speak. Then when financial difficulties and declining enrollment forced them to close schools, they closed the ones that were the most expensive to keep open. So now the schools aren't distributed in any way that's close to even, and the feeder patterns reflect 50 years of gradual tweaks.
The interesting thing about your comment is that Friendship, Shadyside, and the Hill District all fed into Schenley HS until the 1980s, when they started bussing kids from the Hill to Brashear in the South Hills as part of their integration attempt. Shadyside Academy is a prep school for the rich and famous that has nothing to do with the PPS system.
Rov_Scam
6d ago
The only dog I ever had died when I was about three and I remember when he died and it not really affecting me much. That being said, I'm sure there are books available to help young kids deal with this sort of thing. Maybe check some of them out?
If that were the end of the story it wouldn't be an issue. It's that it evidently uses significantly more computing power than the performance improvement would suggest, raising the spectre of rapidly diminishing returns.
More options
Context Copy link