|The JC pontificates about technology |
An occasional series.
“Any sufficiently advanced technology is indistinguishable from magic.”
- —Arthur C. Clarke’s third law
“Any sufficiently primitive middle manager will be unable to distinguish a basic chatbot from magic.”
It is a lot easier to write a thought piece about how robots can do “360,000 hours of legal work in seconds” because you read that in a book somewhere than it is to get a robot to actually understand a legal agreement.
- —The JC
1. (Vogue usage) Information technology that assists in the provision of or, more likely, delivery of, legal services. In these pages used interchangeably with “reg tech”. Both quite disappointing.
2. (Contrarian usage) Not chat-bots, AI, metadata extraction, fuzzy logic or reg tech or blockchain. Instead, the code that lawyers are trained in how to programme from their first day at law school: words.
We have mapped out the legal tech landscape, at least insofar as it maps onto the commercial contracting process, here.
Legaltech and the problem of rent-seeking
Legaltech addresses inefficiencies which manifest themselves as negative annuities: ongoing costs and resource drains for quotidian tasks with minimal value. Its business model is therefore predicated on the vendor earning not just a fee, but an annuity. The rationale is this:
If customers have an ongoing cost of ten, they will be prepared to pay me an ongoing cost of two to remove it.
Mathematically, unimpeachable logic.
But there is a paradox here: If your legaltech solution itself generates ongoing labour, soaking up costs and resources to keep working, such that that two represents an honest margin on that ongoing work, then itis not legaltech but something else. This is more like process-reengineering coupled with outsourcing. That is not legaltech. That is management consultancy.
If your solution really is legaltech: if everything needed to remove the customer’s ongoing cost of ten is done upon implementation then, once the customer has paid for it, why should it pay any more to operate the machine? Why should there be an ongoing marginal cost per unit?
Here there is no longer an ongoing cost of ten: the customer’s problem is solved. The machine costs nothing. The customer’s question is now: what on earth am I paying this ongoing running cost for?
What is new about technology
Ray Kurzweil will tell you we are at an inflection point where our technology is so good, and developing so quickly, it is about to become self-aware. Not only that, the universe itself is about to wake up and become self aware. Now that particular cup of Kool-Aid hasn’t made it to the JC yet — it seems to be going the other way around the circle as a matter of fact — so let him hold forth on what he knows, and that is this: the startling developments in technology in the last forty years hail from three interconnected places:
- The analog/digital transformation: The discovery that information can be abstracted from the substrate in which it is usually embedded, so that data can be transferred from place to place without being buried in an analog medium of some kind. A letter, as an informational construct, can exist without ink, paper or an envelope.
- Moore’s law: Now we have liberated data from its substrate, we need the kit to process it. This finally came good when the vacuum tube — still a thing of great beauty, especially when over-driven with the signal from a stratocaster in a Fender amplifier — gave way to the transistor. Transistors suddenly got smaller, and cheaper. The smaller and cheaper they got, the more you could pack on a chip, and the faster they got. Moore’s law documents the exponential increase in processing power through that decrease in size and cost of processors themselves. The information in a letter can be automatically, quickly and cheaply copied, augmented, processed, changed. This is revolutionary, has burnished the careers of many a legal eagle, and ensured our world is bedecked, from the farthest reaches of the alluvial fans to the highest peaks, with flannel.
- The network effect: The exponential increase in our own digital inter-connectivity across the globe. The abstracted flannel in your average business communication can thus be (i) extracted from the earthly shackles of its substrate (ii) decomposed into addressed packets of data; (iii) routed across a network and (iv) reassembled at the destination address and (v) injected (if need be) back into a substrate (e.g., printed).
Any one of these developments is powerful, but when the three work together the results are revolutionary. The analog/digital transformation commenced as long ago as the Jacquard loom in 1804. Moore’s law has been a thing since before Gordon Moore first noticed it in 1965. The internet — a global network of interconnected computers, used mainly by the military industrial complex — became a public thing when Sir Tim Berners-Lee invented the world wide web in 1989. Suddenly, we were cooking with gas. Well — we were using a webcam to know whether the coffee pot needed topping up, at any rate.
But note: all of these things are hardware developments. They make software possible and worthwhile, but this is all about the kit. Self-awareness, and intelligence, you would think, is all about the software. The thing about the kit is that it is there. It is cheap. It is fungible. These three effects are costed in — they come with any software solution free of charge.
And as for software: it is, we are told, about to eat the world. All the world is a coder. You can find cheap coders in rented rooms in Bratislava, Bogota and Bangalore. Coders are mainly fungible too. So if one guy can do it over a weekend, so can ten thousand others.
What makes a killer app
The effect of technology is to lower barriers to entry. To publish a book you used to need a printing press and a distribution network. Now all you need is an iMac. A killer app needs to create its own barriers to entry. In the networked world there are two means to that, and you need both:
- Killer software — you write inspired, clever, imaginative code that does stuff that no-one else thought of, and it took a lot of effort to put together. Think Industrial Light and Magic, Pixar, IK Multimedia or Microsoft Office — here the barrier to entry is the sheer complexity and cleverness of what you have done with your software. It won’t last forever: killer software developers can charge a few people a lot of money, but they have to keep improving their software to stay ahead of the competition, unless they get so far ahead they can generate some kind of network exploitation effect too. For example, Microsoft Office.
- a natural monopoly — since we are all about the network, an application that exploits the network effect — think eBay, Facebook, Amazon or Google — can come to dominate it and generate its own natural barriers to entry. Here the first mover has an (assailable) advantage — but a late-comer (like Google) can overhaul first movers with killer software, and first movers can defend against latecomers with killer software that it can afford to develop while its natural monopoly persists. Network exploiters won’t generally be able to charge users anything so will make their money off their interactions with users, by serving ads to them or selling data they have collected about users.
Next:Why is reg tech so disappointing?
- Reg tech creates waste
- Algorithm, heuristic and algorithm vs. heuristic
- Legal tech landscape
- A faster horse
- Code: Version 2.0
- Plain English
- Why is reg tech so disappointing?
- Software as a service
- ↑ Outsourcing has its own hidden costs and shortcomings, of course.
- ↑ As good as nothing. Electricity, processing power. But the customer does not need to rent these from you.
- ↑ See The Singularity is Near. Now there is a view that the universe is already self aware, only it operates at level of abstraction so far above our own mortal plane that we can’t see it — we are to its consciousness as our brain’s individual neurons are to our consciousness — and this idea has force (even if it ios a shade unfalsifiable). But that is not what Kurzweil is saying.
- ↑ They didn't distort harmonically, which meant they made for lousy guitar amplifiers however.
- ↑ Though it may now, after 60 years, be approaching its logical limit.
- ↑ did I say flannel? I mean data.
- ↑ See Wikipedia for more.