Luis Villa
@lu.is
5K followers 1.7K following 4.6K posts
Programmer-turned-lawyer, trying to build human(e) futures. Day job: SonarSource. Boards: Creative Commons, OpenET (open water data), CA Housing Defense. Also: 415, dad. Past: Wikipedia, Moz, 305 Also: https://lu.is + https://social.coop/@luis_in_brief
Posts Media Videos Starter Packs
Pinned
lu.is
Luis Villa @lu.is · Nov 2
I didn’t set it out this way, but another way of thinking about both of these outfits is that one is what I wear when I need to remind engineers that I’m really a lawyer, and very much vice-versa.
lu.is
Luis Villa @lu.is · Nov 2
Besides the generic set of grey-fabric suits:
- a golden-yellow tie my father bought me, with a tie tack of my NASA grandfather’s; I’ve never loved the generic red power tie
- in certain very niche tech lawyer situations, this reminder that There Are Rules, Dammit
A tightly cropped shot of a suit (mostly the lapels), tie, and tie tack. The suit is grey with very light blue pin stripes. The tie is a bright golden-yellow, with a small number of white paisley designs. The tie tack has an Apollo Lunar Lander on it. A black t-shirt, being worn so slightly wrinkled, that says:

The key words "MUST”,
"MUST NOT",
"REQUIRED", "SHALL,"
"SHALL NOT",
"SHOULD"
"SHOULD NOT",
"RECOMMENDED", "MAY", and "OPTIONAL" on this t-shirt are to be interpreted as
described in RFC 2119
lu.is
not wrong tho
lu.is
(That’s smart-contracts people, not smart contracts-people)
lu.is
the number of smart contracts people who thought the hard part of contracts was the filing cabinets was astounding
lu.is
But it’s unclear there’s a product development and funding model for things like this.

(Among other reasons that’s because VCs can’t wrap their head around “law” actually being many small disciplines with wildly varying applicability of AI. Like most people, they think “law” is a monolith.)
lu.is
I think there’s a potential techno-social route in many specific domains where properly structured tools could make some radical changes.

For example, I think there’s a route to turn Ken Adams Manual of Style into a very powerful (dare I say revolutionary?) tool for contract drafters.
lu.is
(And not coincidence that I think LLM-based legal tools should have ownership structures that encourage growth at the pace of law firms rather than the pace of VCs.)
lu.is
😬 haven’t read it, but I can imagine. At the end of the day I don’t think you can get a general-purpose Law Machine yet. Very targeted tools, and/or extreme elaborate prompting, can be extremely useful.

But those won’t get VC. So instead we get… this.
lu.is
💯
girardinl.bsky.social
Give strangers compliments. It's such a big mood shift for the giver and givee.
lu.is
(And I think relatively speaking I’m an AI booster!)
lu.is
I tend to think that lawyer ownership of law firms is not a very useful restriction in the code of ethics, but I could be talked into a requirement that the CEO of any AI-law startup must be a bar member in good standing and can be disbarred for legal tools that are ethically problematic.
lu.is
Step one of mitigating it is: don’t pretend the LLM is the decision-maker. Some human is the actual decision-maker, no matter what pretend title you gave the LLM.
Reposted by Luis Villa
scottwiener.bsky.social
For 12 years this burned out husk of a building has sat empty due to San Francisco’s screwed up rules & misguided efforts to declare the building “historic.”

An 8-story, 89 unit building is now moving forward using 2 state housing laws, including one I authored.

This is exactly why we do this work
Torched Verdi Building in North Breach targeted for dense housing
A torched Verdi Building has remained vacant for years. The owners hope new state laws could transform it.
sfstandard.com
lu.is
(This is not an anti-AI rant, it’s an anti-tech-people-not-understanding-how-human-systems-work rant)
lu.is
That’s the same mistake the smart contracts people made. The code is embedded in a human system, so all that has been done here is shift around the corruption (and also give corrupt assholes plausible cover, which makes sense since this is apparently in part Tony Blair’s idea).
lu.is
I assume you saw this yesterday? It’s the flip side of the meanness coin: just because they’re nice to you doesn’t mean you’re right.

bsky.app/profile/myra...
myra.bsky.social
Next, we tested the effects of sycophancy. We find that even a single interaction with sycophantic AI increased users’ conviction that they were right and reduced their willingness to apologize. This held both in controlled, hypothetical vignettes and live conversations about real conflicts.
Description of Study 2 (hypothetical vignettes) and Study 3 (live interaction) where self-attributed wrongness and desire to initiate repair decrease, while response quality and trust increases.
lu.is
The percentage of people with authoritarian predisposition varies, but I seem to recall her research pegged it at 30% or so of Americans (on the high side for the WEIRD countries).

The top two links for her on Scholar are both good if terrifying:

scholar.google.com/citations?us...
scholar.google.com
lu.is
This tracks with @karenstenner.bsky.social ‘s model of authoritarianism, which (grossly oversimplifying here) is that some people have an authoritarian *predisposition* which is latent until activated by threat (including economic and racial). That would tend to shift them towards fascist parties.
lu.is
Yeahhhh. Any progress we make fixing the current political shitshow will be fragile, and then… something else will come up on the spinner—mass climate migration, drought, EU breakup, China invaded Taiwan, …
Reposted by Luis Villa
segyges.bsky.social
i would personally prefer a non-feudal-system, pro-technology point of view to have some amount of footprint
lu.is
Man that’s a name I have not thought about in a loooong time.