AITA - Am I The AI you are looking for?
Toward a Bespoke Future (AI + LAW = OS/P2P)
Legal AI :
Small model,
Big model,
Multimodal;
Blah blah blah.
Closed source,
Open source,
Multi-source;
Check.
Basement brew,
Mostly true,
AI like magic glue;
Down.
It isn’t a match yet for Law + AI.
We have been, and are trying,
The giant giga model model, retailed DTC direct-to-consumer (LLM), with all the ingredients included (unlabeled) and then an infrastructure of prompt engineering and a cottage industry of (essentially) ‘how to write questions’ and (related) ‘how to write the questions for the questions.’
‘We are like SAAS Law, but, what, wait, the large number of data makes unreliable — (for Law) it cannot be a FAAB — Feature As A Bug — it can’t, it really just can’t — I went and said 9T!’
‘Just put another quarter in it, maybe it will work better, kid.’ (Tokens) + 8b customers, so LFG.’
We have been and are trying the
Elite select, reserve, secretive capital-intense ‘private’ Law wrapper model over the above in 1 but ‘hi-brand’ for Law. A ‘Prompt Privilege’ play.
‘Because of our partnership (undisclosed for 18 months) with BigCo1 + BigCo2 we can sell our wares (including through internal ops, existing sale relationships, and from the technology footprint already there) to the other BigCos - the people who have funds available and looking to “get into AI.”
We are targeting large law firms, because yeah, we are the “leading legal AI company” (we guess) but ‘shit, why don’t they want us?, and they are building their own, and so let’s end up selling to PWC instead, and their friends, so that’s cool.’
Accounting not Legal seems biggest buyer of Legal AI. It’s Cannibal-tech.We have been and are trying the (RW) “Redditor Wrappers,” — that certain widely proliferating, basement-created, homebrewed (Python-putty) always look hokey-jokey somehow, retail to retail one-off, and you know but we can pretend — wrappers which deliver for 20 bucks only a slower shielded window around an LLM of 1; a Walmart-in-a-Walmart — but hey ‘We‘re The New AI/Legal™’ ‘and did I mention’ ‘Hey We Are AI™‘?!
AITA - Am I The AI you are looking for?
No. None of these three.
This is not the AI we are looking for, at least not for Law + AI.
This is the AI we have right now.
‘And so what gives Mr. Non-Code Writing “AI Counsel” - ha! You think you have it clocked? Dummy. You once told me you didn’t have an interest in C because it was once B for Bell Labs then must have missed a quiz and became C, so its only Satisfactory?? That’s so stupid.’ Nobody likes your cologne. And you wear cologne.’
In all seriousness, I believe the successful future of AI + Law, because of the fact that it is a highly and tightly regulated space (for good and bad reasons), because it is exclusive nomenclature-dependent and outside of the ken of most laypersons, because with ai in law you can’t get it 90% right - the successful future of AI + Law is open source P2P. (AI/LAW/OS/P2P)
The successful future of AI + Law is OS/P2P. AI, Law, Open Source, Peer-to-Peer.
Arguments:
Like medicine, you will never be able to truly have laypersons do legal on their own because they don’t know the language of the inputs nor the outputs. There will be many mistakes with high upside risk (unknown).
The one-size-fits-all LLM methodology uses plain language which is not the rubric/tool for legal thought.
Lay people do not understand their discrete circumstances (fact pattern) nor they risks they inhabit.
Lawyers, and lawyers’ regulators, will not go quietly into the night and let the computers have it. (Legal + Technology War)
Local is data, because local are people. Accuracy is ensured when a tool speaks among the data of a person, not when the person takes a slice of their data and puts it into a machine, especially in Law.
It will take some time for Law to integrate AI and the current multiple choice [Walmart, Hi-Price, Walmart Wrapper] is not good enough to serve real legal needs.
We have to admit that our technology is not magic but an extension of our common humanity. It is a tool. Serving the human element, which the profession of Law has shown is collaborative, and neither isolated nor singular, is in the interest of mankind and of the technology itself. (HCA) Human-Centered-AI.
‘If you want to go fast go alone. If you want to go far, go together.’ We are in the “flash in the pan” era/stage of AI and as we settle in and the dross burns off and people burn their fingers it will give way to more specific community (Bespoke!) models. Building close to the ground across 10k groups is faster and more enduring.
AI is P2P. This is what we are talking about anyhow. The model for someCo to crunch all of humankind’s data into a product has been done already. We call it the Internet. If we see AI as a data comms tool instead of a data product tool it will work better because that is what it is.
Open Source is how we got AI in the first place. From Alan Turing tinkering and trying (in between really long distance runs from Cambridge to London!) to make talking computers when there were not computers, collabing at this school or that, this group or that, tapping into biology (?), calling math imagination to this (now unknown to history) student with a will to try, to the early days of OAI. We didn’t get AI by knowing precisely/exactly what we were building in the first place. And so why don’t we continue to let the market and the creative minds lead instead of an entrenched AI-rchy?
For AI + Law for me it is bespoke lawbots, in every city in every place, made possible by those lawyers who know their discipline best and those developers who can put up with lawyers. Oil and vinegar. But also match and gunpowder.
******BREAKING*************** As I wrote this early on (7/17/24) various computer systems completely failed worldwide impacting medical, airlines, computer systems, banking, casinos, energy, everything. SO DIDN’T MENTION ABOUT ONE VERY BIG ADVANTAGED OF HAVING OPEN SOURCE DISTRIBUTED P2P SYSTEMS AND THAT IS THIS VERY EVENT TAKING DOWN CENTRALIZED SYSTEMS……………WHY NOT DISTRIBUTE THESE SYSTEMS INTO NODES AND LOCALS…? LOCAL. DailyMail “What will you do when the LLMs go down?” (-AI Counsel April 2023)
