Speaking on a recent episode of The Corporate Counsel Show, produced in partnership with TransPerfect, he discussed the litigation landscape for legal departments moving forward and the importance of good information governance, as well as artificial intelligence and its place in preparing businesses for litigation.

“Everyone” is talking about AI, according to Balmer, who said he gets asked about it in almost every meeting he has.

“In the last two or three years since generative AI came around, I get asked about it a lot more. We’re dealing with huge amounts of data – one of the easiest examples is technology-assisted review, where we’re using AI to optimise document review one by one. And it was Judge Vickery who had the landmark case back in 2016, I think they had six and a half million documents involved. And the judge just said, this is completely unreasonable. A human could not get this done within the time period that’s required. So, I’m forcing you all to use AI,” he said.

“And that was great at the time and kind of set a precedent. But we now see that technology used on pretty much every case that’s over a thousand documents because it’s just adding value. So why is it important? Because it’s going to mean that as litigation increases, lawyers are going to be able to deal with the volume of data increasing as well within disputes specifically. But I think there are some areas of concern when we’re talking about generative AI particularly. And I think there are actually quite a lot of smoke and mirrors out there right now as to what’s achievable.”

In addition to technology-assisted review, AI is also being used within a variety of other areas, including translation transcription, of which TransPerfect is the world’s largest translation provider, according to Balmer.

“Our teams have been working on translation tech for a really long time. So, translation and transcription are two areas. We’re seeing AI used a lot and the quality increased a lot. And then we look at things that just take a lot of manual human time, things like redaction. Redactions now can be automated through AI, which is where we scrub out personal information or privileged information so it’s not disclosed in litigation. And then we’re seeing a lot more images in disputes. And image recognition, where it can automatically understand and tag the image with something it seeds, as well as to duplicate against another image.

“And then also in video as well. So, we see a lot of stuff for CCTV footage, working with police forces around the world, and they’re looking for or just general CCTV footage from various different locations, and trying to look through that content and find what you need is really hard. And so, there’s AI that has object detection within it, so it will see when things come on and off the screen and tell you. So, you can focus on just that part of the video, which I found really fascinating,” he said.

VIEW ALL

“Then if we look at generative AI as a box of tricks, it’s just crazy how fast it’s accelerating. It’s very, very good at summarising content, about taking in content, and you can ask questions about that content or data. It’s going to be useful for creating chronologies, these manual tasks, which it makes sense that it would be able to do that. But there are limitations, and we’ve already seen some things come up in the news where those limitations have led to lawyers being a little embarrassed. Regardless, it’s a fascinating area that’s only going to get much better.”

However, there are still a variety of key concerns around the use of AI – namely around practitioners being trained to use the tech before jumping in.

“A lot of people think that you can just pull this stuff out of the box and it’s going to work brilliantly. Or the technologies, even the technologies [that] have a quite specific use case, they still need to be trained. And that’s the other big one outside of training is the hallucinations that come from these systems. So, hallucinations being when it basically lies to you, which is lovely. Not that humans ever do that, of course, but when it doesn’t know an answer, or it will basically just tell you something that it thinks might be right because it works on logic,” Balmer said.

“The final one, which I don’t think a lot of people know about, is the cost. Running these tools takes a lot of computer power, and a lot of this is done in the cloud. And the cost of storage in the cloud is relatively cheap, everyone’s seen that. But the cost of computing in the cloud is extremely expensive. And [there are] also the ESG considerations, because with this much computing power, you need big warehouses of cloud computing technology to be able to deliver it. And those things are very hot and take a lot of energy. So [there are] lots of concerns from our side.”

The transcript of this podcast episode was slightly edited for publishing purposes. To listen to the full conversation with Tom Balmer, click below: