Google just launched the most powerful large language model (LLM) the world has ever seen – Gemini. Gemini now powers Google’s publicly accessible genAI service: Bard.

Last week the UK government published a report claiming that the finance, public, legal and education sectors would be the sectors most severely impacted by genAI.

So we asked Bard (Gemini) how those sectors will be impacted by AI, and we then asked our sector experts and our AI director, Luke Budka, for their opinions.

How will the legal sector be impacted by AI?

Bard says:

GenAI is revolutionising the legal industry by automating tasks, providing legal research assistance, and supporting legal decision-making. It can be used to:

  • Process legal documents: genAI can extract information from legal documents, such as contracts, pleadings, and case law, and summarise it in a concise and understandable format. This can save lawyers time and effort and improve the accuracy of legal research.
  • Analyse legal precedent: genAI can analyse vast amounts of legal precedent to identify relevant cases, predict outcomes, and provide insights into legal trends. This can help lawyers make informed decisions and strengthen their arguments in court.
  • Draft legal documents: genAI can assist lawyers in drafting legal documents, such as contracts, leases, and wills, by suggesting language, identifying potential issues, and ensuring compliance with relevant laws. This can free up lawyers to focus on more complex legal matters.

Alex Goldstein, our creative director and legal sector expert (having worked with Hunters, Irwin Mitchell and more) adds:

“Let’s not forget translating legalese. Recent research has shown that lawyers find plain English contracts easier to read and remember. Crucially, they also found them equally enforceable.

“With genAI they can produce faster friendlier versions – pleasing clients who are increasingly demanding legal paperwork that it doesn’t take a degree to read.” 

What should the legal sector do about it?

Luke Budka, our AI director says:

“This sector is ripe for disruption. In a lot of ways, it will level the playing field and help smaller legal firms compete. You will no longer need teams of lawyers pouring over documentation when you can augment a large language model (LLM) with the latest case law.

“This increase in speed was most notable in a recent research report published by the University of Minnesota Law School. They gave 60 random students access to GPT-4 and asked them to complete four specific tasks. Most notably they found small improvements to the quality of work produced, but large and consistent improvements in speed.

“They also discovered that while AI provided a universal boost to quality across the test group, there was a substantial improvement in the worst performers. Minnesota concludes: “These findings suggest that AI could substantially transform the legal profession, streamlining tasks, improving lawyer satisfaction, and reducing inequality between lawyers.”

“The prompt engineering piece will also be important; Microsoft recently demonstrated how their researchers can replicate the output of a specialist medical AI using GPT-4 plus a specific blend of prompting techniques.

“The ‘Medprompt’ model outperformed specialist models such as Med-PaLM 2. You could argue therefore, that steering GPT-4 with an equivalent ‘Lawprompt’ could result in a comparable reduction in error rate on relevant datasets. This research suggests if the prompt is good enough, then the latest LLMs like GPT-4 Turbo will likely outperform dedicated legal models. It’s why Microsoft has started to take prompting much more seriously with a new prompt resource on Github.”

As Microsoft says:

We’re seeing exciting capabilities of frontier foundation models, including intriguing powers of abstraction, generalization, and composition across numerous areas of knowledge and expertise.Even seasoned AI researchers have been impressed with the ability to steer the models with straightforward, zero-shot prompts.Beyond basic, out-of-the-box prompting, we’ve been exploring new prompting strategies, showcased in our Medprompt work, to evoke the powers of specialists.”

Luke concludes:

“Finally, it’s also imperative that firms are aware of the risks involved in leaking confidential documentation via unregulated genAI use – they’re only going to get that wrong once – every firm will need its own private genAI environment, this goes without saying.”

 

Drop us a line if you’re interested in having chat about our genAI services, from prompt engineering to private environments.