The Semantics of Machine Learning & Artificial Intelligence in eDiscovery

The Semantics of Machine Learning & Artificial Intelligence in eDiscovery

Precision Of Language Is Important – Or Is It?

I recently was asked to speak at the SoCal Women in eDiscovery Technology Symposium on the Future of AI + eDiscovery with Cat Casey from Disco (and previously Gibson Dunn) and Christine Milliron at Robins Geller. The panel talked about numerous predictive and automation technologies being adopted in legal services. But we spent the vast majority of our time discussing variations on Technology Assisted Review and Machine Learning in eDiscovery review applications.

Spirited debate on the differences between TAR, Machine Learning, and AI is a favored pastime of slightly tipsy conference veterans [1]. After the panel, an attendee posed a thought-provoking question to me, that I’ve continued to revisit over the past 3 weeks:


“Does drawing a distinction between Machine Learning and Artificial Intelligence really, ultimately, matter?”


I had very mixed feelings on the topic.

  • On the one hand, I work in an industry fundamentally based on the notion that precision of language is important. Linguistic shortcuts conflate concepts and prevent clarity of thought. Using the terms interchangeably today creates potential for miscommunications tomorrow.
  • Moreover, as a company laser-focused on value delivery, precision in our language is important towards our everyday commitment to setting clear, actionable expectations as to the services that we provide. At Acorn, we mean what we say -- we have to; our clients are skilled (professionally trained!) at looking past pretense and fuzzy promises.
  • On the other hand, if sexy terminology drives market exposure to emerging technologies, then who cares about semantics? The pragmatist in me only cares about whether we, as an industry, are collectively raising the bar on efficiency and quality.

After three weeks of ruminating on the topic, I have somewhat ironically settled at a lawyer’s answer: It depends. And what it ultimately depends on is the audience. When I am talking to those who hold themselves out to be sophisticated actors in the space, I will use precision of language on topics like this as a measure of their actual expertise. Experts should have a point-of-view on this. However, when I am encouraging people to explore new technology solutions to old problems, I’ll stay more focused on their excitement than their diction.

Ultimately, eDiscovery is a very tactical field. So, in my day-to-day life, differences in AI semantics matter much less than being specific about technologies, workflows, budgets, timelines, and risks.

~~~~~~~~~~~~~~~~~~~~~~~

[1] I don’t even think there’s consensus on the definitions of either, other than that they’re vaguely related but still different. Try googling “difference between machine learning and AI” and see if you can make heads-or-tails of the results. 



Meena Sandhu

Growth Leader | eDiscovery service professional, Chapter Membership Director at Women in eDiscovery - Delhi Chapter

4y

That's perfectly answered Lia Majid. There may be difference for those who really understand it to the core and for others they may be synonymous.....

Safina Lalani

Relativity Master ; E-Discovery Lawyer

4y

I'm always up for stuff like this. Thanks Lia Majid

Khurram Khwaja

Information Technology Consultant

4y

My understanding is that machine learning is the process whereby a machine acquires artificial intelligence. Machine learning algorithms are tailored towards specific 'intelligence' requiring tasks such as speech and facial recognition, etc. The major definitional challenge is actually the lack of a concrete definition of intelligence itself which appears to be predicated on the human ability being replicated rather than consideration of better ideals to base AI design on. Considering human 'intelligence' is the product of evolution and not a well thought out design, this may be a poor starting point to build AI on but I guess it makes sense since AI exists solely to solve human problems. Not sure if this particular take is useful in the eDiscovery domain, though

Like
Reply
Lia Majid

Looking for a Chicago-based business to buy and operate

4y
Like
Reply

To view or add a comment, sign in

Insights from the community

Explore topics