I recently attended an insightful webinar hosted by Iga Kurowska of the LegalTech Academy, with a presentation from guest speaker, Roman Koch (a lawyer based in Poland).  The presentation was focussed on emergence of ‘prompt engineering’ as a specialist skillset for lawyers in the context of the hot topic of ChatGPT. 

I would recommend anyone interested to check out the podcast recording from the session, which will be available shortly - https://www.igakurowska.com/legaltechacademy and from Spotify - but below is a summary of some of my key takeaways from the discussion:

1. Get ready...!

For many of us, AI felt like something that would be confined to sci fi movies but the latest releases of GPT3 and GPT4 by Open AI (as well as others) have gained mainstream notoriety and are here to stay, whether the public and business are ready for them or not.

2. Friend not foe

Large language models (LLMs) such as ChatGPT will undoubtedly change the way lawyers (and, indeed, virtually any industry that relies on written text) do their job. 

Lawyers can either view this as an existential threat or as an opportunity.  Lawyers ought to be well placed to extract the most out of the technology given their existing experience in using language with a keen focus on accuracy and precision – these skills in drafting detailed ‘prompts’ intended to elicit particular outputs from ChatGPT will be absolutely key.

With any LLM, the quality of output will always directly reflect the quality and accuracy of the question one asks it – you get out what you put in.

Hence, ‘prompt engineering’ is likely to become a necessary skillset for those using AI.  Indeed, law firm Mishcon de Reya are already recruiting for a ‘GPT Prompt Engineer’ to work with their legal and transformation team to see how AI might best be able to support those teams.

3. Proceed with Caution

The technology is already incredibly advanced – and is progressing at a spectacular rate - but it is not perfect and is unlikely to ever be so.  The limitations of ChatGPT as a work-tool are well known and unlikely to go away.  For instance:

  • Accuracy – AI is known to ‘hallucinate’ (i.e. presenting as fact information that it has deduced based on inferences or which it has drawn from inaccurate or unreliable sources);
  • Cross-jurisdictional issues – At present, ChatGPT can have difficulty distinguishing legal sources applicable in some specific jurisdictions/scenarios but not in others;
  • Outdated data – ChatGPT was ‘trained’ on data from 2021 and has limited knowledge of events and information after that date, which can clearly pose issues in rapidly moving sectors/industries;
  • Privacy/Confidentiality – Technology which relies on the collection and storage of millions of datapoints poses an obvious threat from a data protection perspective and, as such, ChatGPT has been banned in Italy, with Germany potentially set to follow suit;
  • Copyright – Information is presented by ChatGPT without sources and generally without regard for any IP/copyright implications of that source data.

Consequently, lawyers are absolutely right to view the technology with suspicion.  Governments and private industry alike will need to ensure that safeguards are in place to prevent misuse and, for law firms, to eliminate risks to clients’ personal and commercial interests.

Government regulation is already being discussed at length by data regulators within the EU and, just this week, the Chief Exec of OpenAI (the creator of ChatGPT) appeared at a US Senate committee hearing and agreed with Senators that the industry will need regulation.  It appears, therefore, that there is a consensus that AI should not become a runaway train.

4. The time is now!

Despite the limitations/risks of ChatGPT, the existing technology is already capable of providing tangible benefits to industries such as the legal profession.  The guest speaker showed a number of examples of tasks he uses ChatGPT for on a regular basis.  These were primarily administrative in nature, for instance:

  • Providing high level, ‘at a glance’ summaries of cases and legislation;
  • Converting/reformatting written text (for instance, particulars of claim) into tabular form for ease of review;
  • Bulk document analysis (AI has already been marketed as a solution as part of 'e-discovery' and disclosure  exercises for some time)
  • Producing generic first drafts of written text (e.g. LinkedIn posts, covering letters and internal communications);
  • Correcting grammar and changing the length, tone or intended recipient of written text (particularly useful for those working in a second language).

In the legal profession, embracing ChatGPT presents a world of opportunities for efficiency and client engagement. However, it also comes with inherent risks, such as bias and accuracy concerns. To navigate this landscape effectively, lawyers must cultivate prompt engineering skills, ensuring they harness the full potential of AI while maintaining ethical and reliable legal services.

[This closing paragraph was generated by ChatGPT]