alansaber

18 ms·

31.
by alansaber 27d ago
Even with absolutely minimal dev effort, this product will break even/become profitable.
32.
by alansaber 27d ago
We'll see. If they harshly sanction slop/hallucinations there will be incentives to use the tech honestly.
33.
by alansaber 27d ago
Could be worse, you should see the facebook law comments
34.
by alansaber 27d ago
OpenAI would never misbehave with user data, perish the thought.
35.
by alansaber 27d ago
Precedent is very heavily weighted so yes- flagging insertion points for the re-use of past clauses with AI is a thing.
36.
by alansaber 27d ago
SWE can push to production without going to jail.
37.
by alansaber 27d ago
They'll be operating under a ZDR. Labs will still get some data but I wouldn't go as far as sucked dry of USP. Firms are very aware of the value of their USP.
38.
by alansaber 27d ago
Unlike Google, OpenAI/Anthropic's core business has bad unit economics. So they have to fuck around and find out.
39.
by alansaber 27d ago
Can't wait to benchmaxx on my judges model weights.
40.
by alansaber 27d ago
"Roman law? What other kind of law is there?"
41.
by alansaber 27d ago
Yes, it is. If they had built a specialised tool for a practice area, they could have improved it significantly- but they don't want to be THAT involved.
42.
by alansaber 27d ago
At the end of the day they're a VC-backed company that has to cash in on their brand reputation. Legal is even higher margin than coding with less discerning buyers.
43.
by alansaber 27d ago
Why fix an unsolvable problem, when you can demo DOCX generation and make just as much money?
44.
by alansaber 27d ago
They don't have a choice. If you're a firm and you work with IE a bank, the bank uses AI, their whole supply chain uses AI- your counterparty uses AI- it's a classic arms race with even more social pressure involved (since da
45.
by alansaber 27d ago
Multi-year delays for sentencing in the UK. AI judges are terrible, but when the system is so broken, people start to do impulsive things.
46.
by alansaber 27d ago
Legal search needs its own Arxiv. Nothing comes close.
47.
by alansaber 27d ago
They are cutting into their market. They can dress it up however they want, they might not be competing for enterprise contracts yet (hence the fluff statement), but they will.
48.
by alansaber 27d ago
This is exactly how I would explain inference to a medieval peasant.
49.
by alansaber 27d ago
I mean the law is somewhat deterministic, an amazing lawyer can't just win any case.
50.
by alansaber 27d ago
A lot of paralegal work has already been eliminated by intake automation. OCR and text-to-speech kicked things off.
51.
by alansaber 27d ago
Biglaw salary will probably keep going up for at least a decade. Some high street lawyers could be in trouble. Same story as before.
52.
by alansaber 27d ago
(1) already exists. A more banal example is rich people get their kids private tutors. Poor kids lean more on public resources, wiki, the youtube and the internet. Economic forces will always dictate this kind of change.
53.
by alansaber 27d ago
That's exactly it. You basically can't set up deterministic regression tests, which makes development of (legal capabilities) even more abstract and messy.
54.
by alansaber 27d ago
Friction is inherent to a low-trust transaction. An agreement between two business entities is kind of the epitome of that.
55.
by alansaber 27d ago
But emphasising "tech" advancements, that really only covers OCR.
56.
by alansaber 27d ago
I've heard this argument (basically, you can just have good legal skills and adapt your practice to more areas of law) but we're still pretty far from a layperson confidently navigating court (you could say the same about code).
57.
by alansaber 27d ago
I think the argument goes, even if 90% of a lawyers markups aren't useful, the 10% that are, are basically what you're paying them for.
58.
by alansaber 27d ago
I think you're onto something, but it'll be more of a reshaping of practitioner distribution to match the mode of work (perhaps fewer employment lawyers, many more litigators/ip litigators, etc...)
59.
by alansaber 27d ago
Yes the probabilistic smudging machine does not make many redlines with confidence, unless A) directed or B) it's a painfully out-of-distribution argument.
60.
by alansaber 27d ago
"the hierarchy of knowledge", right. A bad or insecure speaker will ramble about insubstantial minutae whilst a good one will get straight to the key points that form an argument. Basically the confidence to prioritise information

More ›