Template:M intro work Large Learning Model: Difference between revisions

{{d|LLM|/ɛl ɛl ɛm/ (''also “[[large language model]]”'')|n|}}

Once upon a time, an [[LLM]] was a “Master of Laws”: the postgraduate mark of the ''sensei'' in the society of legal service providers — either of that, or of the indolence of one not prepared to strike out and put what she has learned into practice — but still: it spoke to perseverance, depth, comprehension and mastery, however pigeon-hearted its motivation.

''Who benefits'', primarily, from this emergent technology? Experience should tell us that the first  — and often the ''last'' — to benefit from legal productivity tools are the ''lawyers''. Should we expect [[this time is different|this time be different]]?

And bear in mind it will be the lawyers who deploy [[LLM]]s as a tool, not their clients. Why? ''Because of that [[ineffability]]''. An [[LLM]] is a pattern-matching device. It understands nothing. It cannot provide unmediated legal advice. It can only ever be a “back-breaker”: the “last mile” needs a human who knows what she is doing, understands the context and complicated human psychology at play in the cauldron of 1 [[negotiation]]. An [[LLM]] can draw pretty, impressive-at-a-distance doodles, but it cannot do that. Nor can it write [[legal opinion]]s — well, not meaningful ones — and nor, unmediated by a law firm, does it have the insurance policy or deep, suable pockets for which a client is paying when it seeks legal advice in the first place.

“But, [[JC]], come on. Be realistic. It is dog-eat-dog out there. Any lawyer keeps the bounty of the [[LLM]] from her clients will soon have her lunch eaten by others who won’t. You cannot fight the invisible hand. We are in a race to the bottom.”

Nor should we underestimate the overwhelming power of the lawyer’s intuition that ''what has gone before is sacrosanct''.

Fittingly, the first chatbot was a designed as a parlour trick. In 1966 Joseph Weizenbaum, a computer scientist at MIT created “[[ELIZA]]” to explore communication between humans and machines. [[ELIZA]] used pattern matching and substitution techniques to generate realistic conversations. By today’s standards, [[ELIZA]] was rudimentary, simply regurgitating whatever was typed into it, reformatted as an open-ended statement or question, thereby inviting further input. As a session continued, the user’s answers became more specific and elaborate, allowing [[ELIZA]] to seem ever more perceptive in its responses.  

Even though [[ELIZA]] was a basic “keepy uppy” machine, it proved surprisingly addictive — even to those who knew how it worked. Weizenbaum was famously shocked how easily people, including his own secretary, were prepared to believe [[ELIZA]] “understood” them and contributed meaningfully to the interaction.  

This is, of course, how all “mind-reading” works: with the right kind of clever questions, the conjurer extracts from the subject herself all the information she needs to create the illusion.  

[[LLM]]s work the same way. Like all good conjuring tricks, [[generative AI]] relies on misdirection: its singular genius is that it lets us misdirect ''ourselves'', into wilfully suspending disbelief, never noticing who is doing the creative heavy lifting needed to turn machine-made screed into magic: ''we are''. We are neuro-linguistically programming ''ourselves'' to be wowed by LLMs.

Yet again, we are subordinating ourselves to easy convenience of the machines. We need to have some self-respect and kick this habit.

By writing prompts, we create our own expectation of what we will see. When the pattern-matching machine produces something roughly like it, we use our own imaginations to frame, filter, boost, sharpen and polish the output into what we want to see. We render that output as commensurate we can with our original instructions.  

When we say, “fetch me a tennis racquet”, and the machine comes back with something more like a lacrosse stick, we are far more impressed than we would be had a human done the same thing. We would think the human a bit dim. But with [[generative AI]] we don’t, at first, even notice we are not getting what we asked for. We might think, “oh, that will do,” or perhaps, “ok, computer: try again, but make the basket bigger, the handle shorter, and tighten up the net.” We can iterate this way until we have what we want — or we could just use a conventional photo of a tennis racquet.

AI-generated image generation famously struggles with hands, eyes and logical three-dimensional architecture. First impressions can be stunning, but the second look reveals an absurdist symphony.  It is just as true of text prompts: on close inspection we can see the countless minute logical ''cul-de-sacs'' and two-footed hacks. (Many humans write in logical ''cul-de-sacs'' and two-footed hacks, but that is another story.)

Either way, the novelty soon palls: as we persevere we begin to see the magician’s wires.We get a sense of how the model goes about what it does. It has its familiar tropes and tics and persistent ways of doing things which aren’t quite what you have in mind. The piquant surprise at what it produces dampens at each go-round, eventually settling into an [[Entropy|entropic]] and vaguely dissatisfying quotidian.  

In this way the appeal of iterating a targeted work product with a random pattern-matcher loses its lustre. The first couple of passes are great: they get from zero to 0.5. But the marginal improvement in each following round diminishes, as the machine reaches asymptotically towards its upper capability in producing what you had in mind, which we estimate unscientifically as about 75% of it.

And many improvements we will see will largely be in the [[meatware]]: as we refine and elaborate our queries; we learn how better to frame our queries, and “prompt-engineering” becomes the skill, rather than the dumb, parallel pattern-matching process that responds to it. Ours is the skill going in, and ours is the skill construing the output. What the machine does is the boring bit.  

In all kinds of literature ''bar one'', construal is where the real magic happens: it is the [[Emergent|emergent]] creative act that renders ''King Lear'' an timeless cultural leviathan and {{br|Dracula: The Undead}} forgettable pap<ref>Maybe not ''that'' forgettable, come to think of it: it has stayed with me 15 years, after all.</ref>. A literary work may start with the text, but it barely stays there for a moment. The “meaning” of literature is personal: it lives between our ears, and within the cultural milieu that interconnects the reading population.  

“Construal” and “construction” are interchangeable in this sense: over time that cultural milieu takes the received corpus of literature and, literally, ''constructs'' it into edifices its authors can have scarce have imagined. ''Hamlet'' speaks, still, to the social and human dilemmas of the twenty-first century in ways Shakespeare cannot possibly have contemplated.<ref>A bit ironic that Microsoft should call its chatbot “Bard”, of all things.</ref>

Now there is one kind of “literature” where the last thing the writer wants is for the reader use her imagination to fill in holes in the meaning. Where clarity of authorial intention is paramount; where communicating and understanding ''purpose'' is the sole priority: ''legal'' literature.  

The ''last'' thing a legal drafter wants is to cede interpretative control to the reader. Rather, she seeks to squash all opportunities presented by creative ambiguity. Just as there are no atheists in foxholes, [[there are no metaphors in a trust deed|there are no metaphors in a Trust Deed]].  

Legal drafting seeks to do to readers what code does to computer hardware: it reduces the reader to a machine, a mere symbol processor. It leaves as little room as possible for interpretation.  

This is one reason why [[legalese]] tends to be so laboured. It is designed to chase down and prescribe outcomes for all logical possibilities, remove all ambiguity and render the text mechanical, precise and reliable. Where normal literature favours possibility over certainty, legal language bestows [[certainty]] at the cost of [[possibility]], and to hell with literary style and elegance.

Legal language is ''finite''. Literature is ''[[Finite and Infinite Games|infinite]]''.

Now: the punchline. Given how important the reader and her cultural baggage are to the creative act in normal literature, we can see how a large learning model is a feasible model in that domain: to move from a model where ''most'' of the creative work is done by the reader to one where ''all'' of it is is no great step. Indeed, there is enough mediocre literature out there that meets this description now, only written by humans, but to formula and slavishly aping hackneyed archetypes that is is no great stretch to do without the writer altogether. In that case, what does it matter what the text says, as long as it is coherent enough for an enterprising reader to make something out of it?

''But that does not work at all at all for legal language''. The language must say exactly what the parties require: nothing more or less, and it must do it in a way that leaves nothing open to a later creative act of interpretation. We should regard legal drafting as closer to computer code than literature: a form of symbol processing where the meaning resides wholly within and is fully limited by the text. But unlike computer code, you can’t run it in a sandbox to see if it works.

We don’t doubt that LLM is coming, nor that the legal industry will find a use for it: just that there is a ''useful'', sustained use for it. It feels more like a parlour trick: surprising at first, diverting after a while, but then the novelty wears off, and the appeal of persevering with what is basically a gabby but unfocussed child wears pales.

==== Coda: is ChatGPT getting worse? ====

In other news, scientists are concerned that [[ChatGPT]] might be getting worse. Studies indicate that its accuracy at tasks requiring computational accuracy, like playing noughts and crosses or calculating prime numbers, is rapidly diminishing.