Raising the Bar in Law Practice and Beyond
Back in 2009, I graduated in Law writing a thesis on artificial intelligence applied to law practice (yes, I am indeed a J.D. although my career evolved in a totally different direction!). Back then, the term "artificial intelligence" mostly meant expert systems - databases of expert knowledge with elaborate query mechanisms - which were of help to law practitioners, especially in Common Law systems where knowledge of precedents (as opposed to written law) is paramount and very hard to master. Fast forward to 2023 and we now witness the evolution of such technologies, to the point they are poised to revolutionise one of the most traditional industries. Let's dig in.
Featured Insight: AI as co-pilot for lawyers and consumers
You might have heard of Microsoft GitHub Copilot, an AI-powered tool that developers can use to autocomplete code. It's powerful enough that, in some instances, it can compose half a developer's code.
Similar tools are now emerging in the legal world. Spellbook, for instance, is a clever Word extension that helps legal professionals write and review contracts. Its accuracy is startling, trained as it is on both generic (like GPT-4) and specialist models.
The utility of large language models (LLMs) in absorbing extensive information is not surprising; they have already shown exceptional competence in challenging exams, such as the bar exam, where students are assessed on their knowledge of an immensely vast quantity of information.
It is also interesting to notice the difference in performance between GPT 3.5 and GPT 4, released just a few months later, which really underlines the exponential growth phase we are currently in.
These technologies aren't limited to legal professionals; they also empower consumers and small businesses.
Genie AI, for instance, allows pretty much everyone to access more than 3000 legal templates, customise them to fit their need and rest assured they are compliant with UK and US law. More to that, it can also help review an existing legal document offering advice and alerts on what to watch out for in plain English.
Does this signify the end of expert legal advice? Absolutely not. But it will expedite legal workflows and make the law more accessible to non-professionals.
Bringing this closer to home, it's a great example of how AI will level the playing field between large corporates and smaller players if the former are not careful.
Just like Midjourney and ChatGPT can enable a small business to compete with a Fortune 500 brand on content marketing, these "legal AI" tools can help smaller teams to become competitive against larger firms, for instance by allowing them to compete for bigger contracts which require a lot of discovery work.
So here you go. There are two diverging priorities to manage. On the one hand, Fortune 500 companies need to be prudent in applying these new technologies (we touched a couple of weeks ago on ethical sourcing and copywriter issues) but on the other hand, smaller companies and individuals who are less risk-averse can leverage them to gain an unfair competitive advantage against big players and erode parts of their revenues.
This is where large organisations need to be decisive in their implementation of AI. Starting from low-risk areas of opportunity, building in-house AI models trained on high-quality, proprietary data in a segregated, secure environment, can be an effective way to manage risk whilst protecting if not even recasting, their position in the market.
Spotlight on: Databerry.ai
In this week's special edition, I reported from the Generative AI Summit and touched on the emerging trend of training smaller, vertical models that excel at specific use cases, versus trying to build an artificial general intelligence.
Databerry.ai is a very interesting tool that, starting from ChatGPT, allows businesses to build their own, vertical AI with zero coding involved. You simply provide them with your own data and voila - you'll have your very own AI. For instance, you could give this tool your knowledge base for new hires. In a matter of days, you'll be ready to offer all new recruits a conversational solution where they can ask questions in natural language and receive answers in real-time and in multiple languages.
Unfortunately, this tool relies on ChatGPT and you have to give them your data, which is not ideal and which is why I already recommended Hugging Face as the largest repository of open source models and data sets you can download and play with on your own secure environment. Nevertheless, Databerry.ai is impressive as it shows, again, how easy it is for a newcomer to offer value to the market. Take a look.
Follow me
That's all for this week. To keep up with the latest in generative AI and its relevance to your digital transformation programs, follow me on LinkedIn or subscribe to this newsletter.