Is the anecdote about the lawyer who used hallucinated precedent in his argument broadly known in the industry, would you guys say?
At this point it feels like there have been hundreds of well-publicized instances. Weāre all on alert for that (at least those of us who do our jobs). The penalties have been getting more severe but still arenāt nearly severe enough. To me, first instance should get your license suspended for a while, second gets you disbarred (a court canāt do those on its own but can make a referral to the state bar). For our own outside counsel, my position is that if you send me a draft (or god forbid file a brief) with fake cites, we never work with you or anyone from your firm again. Hasnāt happened and likely will never happen, but thatās my position. Lying to the court should have consequences.
ETA: I truly donāt mind if lawyers use AI for research & drafting as long as the arguments are good and the briefs are cite-checked. Quality work with fewer billable hours is good news for me.
I encourage everyone to read (or re-read) Snowcrash, because thatās where weāre headed. With the extra excitement of religious zealotry.
It sure is in the OGCās I deal with.
I donāt know that. Iāll take a look.
AIs, machine learning, ChatGPT, and the Singularity are four things, not one, big thing.
In the final analysis, itās 30% bad, 20% good, and 50% ugly.
(I was CTO, VP Data Eng, Chief Architect, and more. Iām retired now.)
I am neither a developer nor a web designer, but in 2017 I drew the short straw to build an internal web app for my employer. (Iām the only one in our two-man IT department to have ever āwritten codeā because I took some CS classes in high school and college.) I spent over two weeks working on it and it was a minimally viable product in every sense of the term, using coding sensibilities and technologies that I largely hadnāt used since the early/mid 2000s and the most rudimentary HTML/CSS. More importantly, though, it got the job done, and continued to do so largely uninterrupted for the next eight years.
Because the technologies I used were going to lose support, the current iteration of this app gained an expiration date of 8/1/2026 and earlier this month I finally had to find a way to modernize it. So I got a Claude subscription and turned Opus loose on my codebase, telling it that I wanted a completely new app on a modern tech stack, with all the same features and a few new ones that I never got around to implementing, an attractive and responsive design built on modern coding and security principles, and resolving some specific corners that I cut. Within five minutes it fully understood everything my app did. Within an hour I had the new appās proof of concept up and running, and by the end of the next day it was feature complete and ready for production. It even discovered some of the workarounds Iād come up with for those cut corners without me pointing them out yet, and implemented a solution for migrating them into the new architecture.
So, in the end I 1) accomplished a critically needed task, 2) did it in a small fraction of the time it wouldāve taken to do it on my own, 3) got a far better product, and 4) all it cost was a $100/month subscription. I already have a list of multiple other scripts and apps that we have needed for years and just havenāt had the resources to create. I wonāt lose sleep over using AI for this because it didnāt cost any humans any work: we will never be big enough to have someone on staff capable of doing the work, and we will never be able to afford to pay a developer worth a shit a fair fee to make bespoke solutions for us. If AI doesnāt do the dev work for us, the dev work will simply go undone and weāll continue to have a bunch of tedious, error-prone processes that we do ourselves.
It is undeniable that AI has a lot of bad use cases, whether they be harmful (impersonating people) or just lazy (low-effort text/image slop). But it is also a genuinely useful tool, and there are plenty of useful tools - guns, telephones, cars, drugs, the internet - that have a lot of bad use cases. Anyone that says itās universally good or bad is eschewing a lot of valuable nuance.
Everything that is dangerous is dangerous because it adds something to our lives. Otherwise it would pose no threat. Nuclear energy is staggeringly beneficial and staggeringly destructive. Social media is fun and convenient and good for business while being psychological megadeath for reams of human young. The more sophisticated our tech becomes, the more and deeper and faster we can lay waste to our ecosystems before any regulating authority has the chance to catch up. Additionally, the more sophisticated our tech becomes, the more permanently irrelevant regulating authorities become. Embracing nuance while the collison unfolds before you is a terrific way to accomplish exactly nothing. The only legal action any of us can pursue that could possibly make any difference is just to refuse to use the stuff and refuse to politically condone its expansion.
Sire, our spies report that Waldo has completed the Great Wall of Text!
Stupid Mongorians Break Down My Shity Wall!
If I could add infinite hearts I would.
Who among us shall channel the spirit of Noe?
juneberno
I just want to know how DAS makes his refried beans.
I grab a can out of the pantry, grab a can opener, put the can opener on the can and turn the handle until the lid pops off, I stab a hole in the bottom of the can so the beans slide out into a bowl maintaining the shape of the can, I put the bowl with the beans in the microwave and I heat for two minutes and stir. Voila!
Can of HEB borracho beans + HEB pico heated up in a pot, then stick blender
Can of HEB borracho beans fried and mashed in butter (if no manteca is available), with some extra chili powder and cumin.
Iām happy to learn that Iām not the only person who likes HEBās borracho beans. Maybe that means there are enough of us out there that they wonāt go away.
Iām seriously considering paying one of you jerks folding money to send me some of these:
Whatās the Austin bandā¦Soup or something?