One of my pet peeves is the conflation AI, machine learning and Document Automation. AI was the 2017 buzz word, and now seems to be used as a blanket term for any kind of automation ever. The AI effect is all very well and good, but defining AI as “whatever hasn’t been done yet” is not very helpful. Just because its technological and the concept isn’t fully understood it doesn’t make it artificial intelligence.
Conflating these terms is a dead giveaway that you haven’t done your research. Most people don’t need an in depth understanding, just the basics. So, here’s the crib sheet for anyone looking to explore a career in Legal Knowledge Engineering, or even just wanting to engage in informed conversation on the subject of legal tech.
Artificial Intelligence (AI)
Artificial intelligence can be understood very broadly as the simulation of human intelligence processes by machines. What this means in practice has been open to some interpretation.
True AI was theorised way back in the 40s and 50s to be a sentient machine. There’s so much academic discussion on the subject that it has made its way into pop culture: if pushed, most people will recognise the Turing test, Asimov’s principles etc. The Eugene Goostman program has supposedly passed the Turing test, but sentient machines just don’t exist yet, at least not in the public sphere.
More recently, this has been reinterpreted in more limited ways. If you isolate the processes that require intelligence, AI becomes more achievable. So, the program might not be true AI, in the sense that it cannot simulate a human, but it may be able to simulate human level intelligence in a specific area. With this lower bar for AI, digital assistants and information gathering interfacing could be classed as AI.
Machine learning is a sub category behind AI that enables the program to improve the output it gives based on the response it receives. This means that user feedback can be processed and incorporated organically to improve the output for the next user. This is increasingly pervasive; its key to all of the ‘AI’ with which we are familiar, but its also a big part of less ‘intelligent’ solutions – autocorrect and google search both use machine learning algorithms to improve the results for a specific user.
Document Automation (DocAuto)
Document Automation is a means of facilitating speedy drafting of documents. Using some specialised software (see my post on software), LKEs design interviews that guide users through the key drafting points of a document to give something between a ‘good first draft’ and ‘signature ready’ version as the output. This is not an automated process; LKEs need to read the documents, pull out the key legal points, and design and ‘code’ (I put ‘code’ in inverted commas as different software has different interfaces and requirements – see this post about software) the automation.
There are many possibilities on the horizon for applying machine learning and AI principles to the DocAuto process, but they are not yet common place.