Artificial Intelligence #28: Should all lawyers understand data science?

Last week, Nidhi Singh invited me to be a keynote at a new AI centre for law which she was launching.

Nidhi has formidable educational qualifications (Oxford, Harvard, Stanford) and is a counsel in the Indian Supreme court.

The talk was based on fostering discussion, new debate, and new ideas.

In this talk, I discussed some key ideas around explainable AI.

This year, at the #universityofoxford Artificial Intelligence: Cloud and Edge implementations, we have a session covering explainable AI by ?two legal professionals Tu??e Yal??n and Matthew Berrick

I think the impact of explainable AI is not fully appreciated in the wider community

So, I asked if all lawyers should be data scientists??

Today, the legal profession is obsessed about AI taking over legal jobs or in predicting court decisions (ex lex machina – a memorable name!)

But much more is at stake – and many miss the big picture of AI and its implications

I made two points

First. let’s consider the case of AI ethics

AI ethics is complex debate (and a superset of Explainable AI). Spinoza’s ethics was the first attempt to apply Euclidean thinking to philosophy to create a system of ethics from first principles. Could we apply the same first principles thinking to formulating AI ethics? The idea is not as far fetched as it sounds and may well be doable. I discussed this at Of Spinoza’s ethics and AI ethics: Why you should create an AI ethics framework for your own organization from first principles

?Image source Amazon

?And the second point was awareness of explainable AI

As it stands today, there are three big limitations to AI (as based on neural networks)

• Supervised learning requires too much labeled data and model-free reinforcement learning requires far too many trials. Humans seem to be able to generalize well with far less experience.
• Current systems are not as robust to changes in distribution as humans, who can quickly adapt to such changes with very few examples.
• Current deep learning is most successful at perception tasks and generally what are called system 1 tasks. Using deep learning for system 2 tasks that require a deliberate sequence of steps is an exciting area that is still in its infancy.

But explainable AI has a wider impact on institutions. Institutions matter to democracy. In the book, Why Nations Fail – the authors Daron Acemoglu and James A. Robinson propose that

?

????????Political and economic?institutions?define a set of rules and?enforcement mechanisms that exist in society

????????These institutions are the main reason for differences in the economic and social development of different states

????????The authors contrast two types of institutions: extractive — aimed at excluding the majority of society from the process of political decision-making and income distribution, and inclusive — aimed at including the widest possible strata of society in economic and political life.

?

Hence, the real question is: How will our institutions be shaped by AI?

• ?Which decisions will be taken by proxy
• If we trust algorithms and algorithms trust data and if there is no data
• How can we trust the decision?

?So, should all lawyers be data scientists? Yes – in the sense of understanding how AI makes decisions

• In law, the blindfold represents impartiality – scales represent weighing of evidence.
• As a lawyer, you can do neither if you do not fully understand the decisions taken by AI (by proxy on your behalf)
• The bad news is .. AI today is data driven and hence cannot be easily explainable
• The good news is .. The next decade of AI could solve these problems
• But .. It will come from more complex models that will allow us especially to apply AI to areas where we cannot currently
• And the rate of change for AI is very high – which calls for a need to understand AI even more

There are interesting papers in this space ex A Taxonomy of Explainable Bayesian Networks Iena Petronella Derks, Alta de Waal

I am reading Stephen Schwarzman’s biography What it takes. A very interesting book. Not since Barbarians at the gate – I have seen a book which talks of the inner workings of high finance

?

Finally, I am pleased to say that the Digital twins course is almost full

If you are interested, please sign up soon - Digital Twins: Enhancing Model-based Design with AR, VR and MR (Online)