Mike Lynch, the billionaire tech pioneer behind legal AI company Luminance and investment fund Invoke Capital has given his view of the AI sector in the UK at an industry evidence hearing organised by the British Parliament, which included sharing his insights into the legal AI sector.
Lynch raised many points about AI and of central importance was the underlining the need to think about data as having strategic value for AI businesses and also for those that may then use the applications and insight gained from that data by machine learning.
He added that from his personal experience as an investor in machine learning companies that the difference between wannabe AI applications and those that are truly robust is significant, with only around 5% of the ‘AI companies’ he sees that pitch to him qualifying as ‘real’ in his view.
He also noted that the legal market, which had initially been viewed as a sector where resistance to AI was expected, had in fact seen rapid adoption of the technology.
The All-Party Parliamentary Group on Artificial Intelligence (APPG AI) that heard Lynch’s, as well as other experts’ views, was set up in January 2017 with the aim to explore the impact and implications of artificial intelligence across a range of issues, including trade, employment and the law.
In this session the experts from around the UK’s AI sector were invited to present their views to the APPG AI Chairman, Conservative MP Stephen Metcalfe, primarily with a focus on how AI would impact the overall economy and what part AI companies would play in it.
Lynch began by giving a short history of his involvement in AI, noting that he had been working with the technology since the 1990s. Following the sale of Autonomy to HP he created Invoke Capital and is now working largely as an investor, if not a catalyst, for new AI companies.
At present, other than Luminance, Lynch and Invoke have backed a number of other AI companies: Dark Trace, which handles cyber security, and SophiaGenetics which is a machine learning genetic information company. A fourth company where Lynch is involved is machine learning image/pattern recognition system, Neurence.
Lynch then gave his personal view as to why AI is such a hot area for investment today.
‘AI is still hard to do. The number of people who can do it is low. There is also a very strong first mover advantage in investing in AI now. Whoever gets going first has an advantage [because of their ability to learn from data].’
‘There is a positive feedback loop. For example, the more people talk to Siri, the more it will learn,’ he added as an example.
Other speakers later echoed his point about the number of people that ‘can do AI’, noting that salaries are rising rapidly for people with machine learning skills. This in turn raised some fears with the Chair as to whether the UK was doing enough to train and keep AI talent, (which is a topic that may need to be covered in more depth at a later stage.)
However, from a legal tech perspective, the following point that Lynch made about data was of more immediate importance.
‘The issue now is ‘strategic data’. [i.e. data sets that allow AI to learn],’ said Lynch. ‘An example is data held by the NHS (the UK’s national health service).’
That is to say, having a large and deep data set allows those companies that can access it to develop very useful AI applications via machine learning. Using the NHS example, nowhere else in the UK would an AI company be able to find such a huge store of current and longitudinal data on medical issues and their diagnosis and cures. Therefore, NHS data becomes of strategic importance to any AI company in the medical field.
The same could be said for all the legal data across the world, whether that is in the courts, in each individual law firm, or held by the clients. This is, one might say, ‘strategic legal data’, and it’s of great value to all stakeholders with access to AI technology.
That said, the AI companies Lynch sees were not all equal in his view, at least among those that came looking for investment.
Lynch said: ‘Most proposals I see [for investment] involve using machine learning. But, 90% of these don’t really have AI. The other 5% have not tried their solution in the real world yet and the last 5% are real.’
‘Most investors cannot separate the 90% from the rest. Most companies will demo very well, but not work in the real world. The issue is being robust. So there will be some disillusionment [among investors] about AI,’ he added.
That’s not to say there are not dozens of very robust AI companies out there, even just in the UK and many more globally, however, it does suggest that too many businesses are claiming to have AI capability when they don’t really do much more than have a bit of automation or some very rudimentary responsive capability. Or, in other cases, the systems are of a high quality, but simply break down in more complex and unpredictable real world situations were users can be guaranteed not to behave in an easy and linear way.
That said, from market feedback that Artificial Lawyer has received, the legal AI market is – so far – generally stocked with credible and robust AI companies, such as Luminance, iManage/RAVN, Kira, Seal and many others. Although there are a couple of ‘mislabelled’ applications still doing the rounds (mostly in the document completion space). But, these seem to be the minority, at least in the legal sector.
Lynch also noted the challenges ahead in relation to the insurance of driverless cars, the potential for machine learning to extract information that may appear to be biased and the need for explainability of how an AI system had come to a particular answer.
One of the most interesting points Lynch made here was that as financial markets become increasingly automated, e.g. the use of algorithmic trading and high speed trading, then ‘we may need markets to be regulated by AI [because the markets will be using AI also to trade and only other AI systems will be able to help there.]’
This is especially interesting from a legal and regulatory point of view, as rather than seeing AI as a threat to the legal landscape, it would instead be seen as an ally in compliance and regulatory oversight, in part because nothing else will be able to keep up with how trading is taking place.
When the committee was opened to questions, Artificial Lawyer asked Lynch how he viewed his venture into the legal world with Luminance. He replied that initially he had not expected lawyers to be so receptive to AI.
‘However, they are falling like dominos. This might have something to do with how their business model is changing,’ he added.
Lynch concluded by saying that the focus on legal due diligence came about in part because ‘M&A due diligence was a good area for AI because you had all that data [for machine learning],’ which related back to his earlier point about the value of strategic data.
Interestingly, the founder of another legal AI company, Tim Pullan, of Thought River, was also in the audience and asked the committee if they could define what AI was. This created a moment of nervousness among the room, before Lynch took up the challenge.
‘AI is two technologies. There are the expert systems that were developed in the 1980s and the new wave of applications using machine learning and neural networks.’
‘It is the second that is having an impact and AI is just shorthand for machine learning. Expert systems are not shaking the world up,’ he summed up.
Artificial Lawyer’s written content is Copyright (c), unless otherwise stated that it has been produced by another party. Do not reuse, repost, or otherwise host Artificial Lawyer articles on other sites. Naturally I am very happy for you to link to articles, retweet etc.