Olivia Wang

CS PhD student @UCSC. Research interests in Human-Centered #XAI (#hcxai), Neurosymbolic AI & Law (#NeSy). Animal Whisperer. Views (and the dog) are my own, INFJ

[Olivia Wang @oliviawpy2023]

Please allow me to present you Olivia now as a PhD candidate 🥳🥳

[Olivia Wang @oliviawpy2023]

@ThomasWolgast A rule to live by for sure

[Olivia Wang @oliviawpy2023]

One of my pet peeves is that some reviewers would look at the paper with some pre-determined opinions. Like I am not saying your suggestions are not valid, and you have pointed out some real gap that can be future work. But that’s not what my paper is about.

[Olivia Wang @oliviawpy2023]

Like you could totally do those yourself and I would be very interested in reading them. But there are thousands of alternative directions that I can take and there’s no way I can do every single one of them just in one paper 🤣🤣😅😅

[Olivia Wang @oliviawpy2023]

Had to decline the invitation because of visa issues. God, it stings.

[Olivia Wang @oliviawpy2023]

I really want to attend Dagstuhl Seminar 26472 on Neurosymbolic Reasoning About Policies and Legal Texts; the only caveat is that it'll be in Germany, and I likely won't be able to travel overseas. It will be fun to see how far my kindergarten-level German can get me though. Ugh

[Olivia Wang @oliviawpy2023]

Fingers crossed that they have remote/zoom options. I would really hate to miss it 🥲, but I also cannot risk not being allowed to re-enter and finish my PhD with all the craziness related to F1 students going on.

[Olivia Wang @oliviawpy2023]

@ThomasWolgast we should do both😂

[Olivia Wang @oliviawpy2023]

I think we should do something about the reviewers who ghost after the initial review🙂

[Olivia Wang @oliviawpy2023]

@luislamb Thank you Luis!!

[Olivia Wang @oliviawpy2023]

I passed my advancement 😬 time to actually start working on the future works I proposed now as a PhD candidate 🤣🤣

[Rob Freund @RobertFreundLaw]

“Counsel’s lengthy and unfocused filings contain gibberish and abuse the litigation process in different ways—including by churning out convoluted, false and frivolous arguments,” said the court in rejecting the petition.

“Soroka denied using AI to draft her filings in a statement to https://t.co/IwFIknyQqT, which has additional coverage.”

“‘I can confirm that in my practice of law, in all matters, I do not use artificial intelligence to generate court filings,’” Soroka reportedly said.”

At that point I might just falsely confess to using AI…

[Olivia Wang @oliviawpy2023]

@notguiltyCHI @writeclimbrun I think lawyers are so expensive in the U.S. compared to other civil law systems because you are essentially paying for the judgment and experience of the lawyers you hire. If there is a way to commodify that, sure, but it will take at least a decade (likely more)

[Olivia Wang @oliviawpy2023]

@SMB_Attorney I said the same about working in big law for just one year doesn’t mean you know everything (or honestly, much) about how law firms works. I think the problem really is that in current times people get major FOMOs and everyone is trying to “make it” as fast as they can.

[Olivia Wang @oliviawpy2023]

@vdtmehra I mean... I'm gonna ask the question again.. What is AI native law firm anyways 😂😅? Like, what is the definition, and what does it mean exactly?

[Olivia Wang @oliviawpy2023]

@langer_han Awww thank you for the kind words 🥺

[Olivia Wang @oliviawpy2023]

The cost of legal services is not approaching $0 🤦🏻‍♀️🤦🏻‍♀️ as a matter of fact it may be approaching ♾️ now that you use your AI and screw it up 😬🤣

[Rohan Pavuluri @RPavuluri]

The cost of legal services is approaching $0. I've mostly seen two ways to monetize this theme: (1) selling tokens to law firms (@WeAreLegora & @harvey) who "resell those tokens" to their clients, and (2) providing legal services on your own to businesses or consumers at a much lower cost profile & taking on more clients.

One thing I haven't yet seen is how to express this theme in other industries on the assumption that legal services are trending to $0.

In a world where you have unlimited free lawyers, what are the opportunities to do things that you wouldn't otherwise be able to do? Some ideas:

One idea is flavors of investing that are heavy on interfacing with legal/regulatory challenges i.e. public activism. On an adjacent note, buying B2B receivables and then using AI to collect.

Interesting to think about the others. Especially with the advent of long running agents, opportunity set increases with ability to answer emails and engage with adversarial processes in a way that seemed harder before. Muse & Instinct feel like a first taste into what's possible.

[Olivia Wang @oliviawpy2023]

@langer_han Btw, I also made a similar point in one of my blog posts! Great minds think alike 😬😬
https://t.co/7YwnbwrchF

[Olivia Wang @oliviawpy2023]

@langer_han And fixing mistakes is so much more costly than doing it right the first time, plus there’s no guarantee that you will be able to fix it even with the money you put in. My ex used to joke about it saying “today is the day that I actually earned my paycheck” 🤷🏻‍♀️

[Olivia Wang @oliviawpy2023]

I feel seen

[boisselle 🇺🇸 @seanboisselle]

If you raw dogged a CS degree pre-LLM you can never be a true vibe coder. You’ve seen too much.

[Olivia Wang @oliviawpy2023]

When the evaluation results are published, they should also publish how the models/tools are used to achieve these results (aka the “right” way to use them in order to get the results we promise). YMMV but at least it’s more concrete than not

[Olivia Wang @oliviawpy2023]

When those legal AI startups publish their models/tools, evaluation of efficiency increases should also be reported. That will be a decent indicator of how much of that model improvement can actually translates to real efficiency increases. It’s useless otherwise

[Nick Berk @theNickBerk]

Most lawyers are just using AI for individual tasks. That isn't going to move the needle. It's going to make it slightly faster...but you lose a lot of that time checking its work.

For clients to really see the benefits, lawyers need to standardize and continually improve their processes.

That means (1) identifying the components of work that get repeated; (2) creating a standard prompt/context for that task; (3) improving that process over time.

You get some improvement for free...because models keep getting better. But if you're not creating repeatable, improving processes...clients aren't really going to benefit.