Generative AI has exploded into the mainstream. It is changing the way people work and disrupting entire industries. It also comes with challenges and risks that the business world will need to grapple with.
To unpack some of these key considerations, UKTN hosted a webinar discussion with a panel of sector experts, in partnership with transatlantic law firm Womble Bond Dickinson.
The session explored how different sectors are embracing generative AI, whether the hype is justified and some of the technology’s risks and limitations.
Here are three key generative AI takeaways from the webinar that businesses should know.
The hype is justified – but there will be peaks and troughs
At the heart of the rapid adoption of generative AI tools is the “incredibly human-like, and in some cases, astonishingly accurate” responses, said Katie Simmonds, managing associate and technology and data privacy lawyer at Womble Bond Dickinson.
The most popular generative AI tools – such as ChatGPT, Bard and DALL-E – all have low barriers to entry. This accessibility means adoption has been led by public sentiment “rather than a tech fad for techies”, said Steve Harris, head of technology, business and commercial banking at Lloyds Bank.
“There is definitely that user case, there is definitely that commercial aspect of it,” Harris explained. “And therefore because we are only scratching the surface of the user cases and the capabilities that might be there, I think it is going to continue. I think there’s plenty of runway for new products to be built using the models.”
Advances in AI have been underpinned by the decreasing cost of powerful hardware over time and the availability of large training datasets. However, there are two areas that “may dampen the hype for a little while”, said Hannah Stothard, chief operations officer at SeerBI, a data startup for the maritime and logistics sectors.
The first is bottlenecks from the physical components, with the US and China both tightening semiconductor supply.
“It’s all actually dependant on very physical things, even though it’s kind of just out there on the cloud,” Stothard explained.
The second is the risk and governance side. Governments are currently playing catch-up on regulation. In the meantime, businesses must be mindful of best practices and existing frameworks, such as data protection laws, when adopting the technology.
While there has been a drop-off in ChatGPT users, the panel expects generative AI adoption among businesses to continue.
“I think it might be one of the peaks, but I think there will be more to come,” said Harris.
It’s already impacting sectors
The effects of generative AI is already being felt by companies large and small, and across a wide range of sectors. Stothard says her company has not needed to hire as many people due to AI increasing productivity, such as creating templates for job descriptions using ChatGPT.
“For us as a startup, it really does enhance the effectiveness and efficiency of our business plan,” Stothard said, adding that it is leading to a more efficient use of capital.
On the legal front, Womble Bond Dickinson is looking at ways to use generative AI to summarise judgements for seminars, and an internal knowledge hub for staff to ask queries about company policies.
“That’s quite a narrow niche use, but you can see that scaling up the internal time that we’ll spend making queries, and responding to queries could be quite streamlined,” Simmonds said.
Financial service companies are also exploring ways to incorporate generative AI tools. Harris gives the examples of better-connecting data for knowledge management, improving search for large sums of data, conversational chatbots, and AI “co-pilots” working alongside humans.
“Because we are highly regulated, we do have to venture into it relatively carefully and there’s obviously lots of data and confidentiality aspects,” Harris said. “But we do genuinely believe that there can be a real client benefit and also colleague benefit of implementing the technology.”
Generative AI risk is multifaceted
All the hype surrounding generative AI means it can be easy to lose sight of the risks. ChatGPT is “an excellent web scraper” but it “cannot challenge the accuracy of the underlying sources that it scrapes the data from”, said Simmonds. This means it can perpetuate historic biases and discrimination in its responses, Simmonds added.
“We’ve seen ChatGPT blend fact with fiction, and reference journals that just don’t exist,” Simmonds said. “These flaws, coupled with the privacy and discrimination risks around using these technologies, have really emphasised the gap in regulation in this space.”
Keeping a human in the loop is one way that businesses can reduce risks.
“I think people need to understand what the capabilities are currently. I think a lot of people think it can do everything and it can’t – there’s still got to be a human element to it,” said Stothard.
There’s also the risk of what Stothard calls an “AI data winter”. ChatGPT, for example, is trained on data from the internet up to September 2021. Future models trained on online data will be feeding in AI-produced content – and it’s unclear what the impact of that will be.
“Now you’re getting models that are hooked up to the internet,” Stothard explained. “So there’s going to be very little information that is produced that is human data, because in some way shape or form it would have been impacted by AI.”
It’s not clear how this will influence future AI models, but it could cause “stagnation”, Stothard said.
Businesses must also be wary of generative AI tools being adopted by criminals to commit fraud.
“Some of this technology in the wrong hands could be misused,” Harris said, pointing to audio deepfakes to obtain confidential information as one example.
The hype surrounding generative AI tools can also mean companies lose sight of solving a specific problem.
“Generative AI is one tool, so it’s important not to have tunnel vision and think that generative AI is always the answer,” Simmonds said. “It’s part of a wider suite of tools that are available including automation or rule and template-based systems, which are powerful tools and also have a wide application.”
To find out more about generative AI and what it means for businesses, read UKTN’s guide on the topic here.