Generative AI – hype or game-changer?

A brief history 

Gen AI works by using neural networks to identify patterns from large sets of data, then generates new and original data and content. This AI revolution was given a kickstart in 2017 by Google scientists looking to improve the accuracy of Google Translate. At the time, the tool translated each word in a sentence sequentially.  

The scientists were working on a concept they called ‘self-attention’ that could speed up and augment how computers understand language. This involved considering the context of language, and when applied to the translator meant taking an entire sentence and analysing all its parts in order to produce better context and more accurate outputs. This led to the development of an architecture for processing language known as the ‘transformer’ which is embedded in Google Search and Translate but also powers all large language models including ChatGPT and Google’s own large language model, Bard. In addition to driving autocomplete and speech recognition by smart speakers (eg Amazon Alexa, Siri, Google Assistant) it also works in areas beyond language, generating things with repeated patterns such as images, with tools like Midjourney and Dall-E, or computer code generators such as OpenAI Codex or GitHub Copilot. 

Is Generative AI a game-changer?

RJ: It’s true there IS a lot of hype about Gen AI, but it IS a game changer. It’s one of those transformative moments in history. We’re now looking at a place where we can have copilots for every single human endeavour and essentially have an assistive technology that increases productivity, helps with efficiency, even for creativity or ideation. We’re seeing that happening across a lot of different sectors right now. And it’s heartening how many organisations in our space, are embracing this technology. 

However, Martec’s Law talks about the principle that while changes in technology occur very rapidly, changes in organisations do not. This tends to apply to the regulated industries that we work with, which are usually quite hesitant to adopt new technology. We saw that with Cloud technology adoption, so it’s been interesting to see just how many organisations have openly embraced experimentation and are seeking to understand how they can best use Gen AI 

Q: What value does Gen AI offer to businesses in general and technology in particular? 

RJ: Businesses and technology in particular have an opportunity to reduce joyless work by using Gen AI. There is work that gives us joy and there’s work that does not give us joy, and this joyless work is ripe for automation. In technology, for example, this can include things like supply chain management, modelling or creating articles, content generation, in fact anything that has an element of pattern to it.  

Gen AI can assist here because it’s a probabilistic pattern-matching tool, which means that the answers it gives you may not be true, but there’s a probability that they may be close enough to something that humans might produce using the datasets that it knows about. There are obviously pros and cons to that approach, but it’s now approaching the point of ‘good enough’ for businesses to start exploring. 

Q How are organisations currently using Gen AI? 

AI-generated pair programming is a fantastic use case and endeavour in the software engineering space. It’s something that every single developer grows up doing, and high-performing teams do as well. To have an assistive technology copilot your work, which can comment on your code and say that these are the things that require attention is very useful. 

This builds upon some of the tools that already exist. For example, SonarQube, which has been around for a while, which runs through a series of rules and is deterministic. The assisted version is a bit smarter – so rather than running a series of checks against rules, you now have Gen AI helping as a human might do to offer an opinion on the code. It goes beyond just rules.  

Of course, in software engineering, a lot of documentation is produced, and this is also ripe for automation. 

Q: What do you predict the impact will be on productivity? 

RJ: You know, I think there’s productivity and there’s valutivity. I know that word is frowned upon sometimes but it’s the opportunity to do higher value work that Gen AI offers, rather than straight productivity, which is about utilisation and uptime.  

It’s probably going to be used in some sectors, like call centres where there can be large-scale impacts across the industry. The reality is, it’s going to free up time for other endeavours, and it will change the way we engage with work. So, it will change the things that we do, as opposed to making us more productive. It allows us to go up the value stream. For example, before we might have spent a lot of time on generating content, maybe now we can spend a bit more time with our clients, to really understand what the problem is that we’re trying to solve. Or in the case of generated content, we’re now more often in editor mode.  

Q: Do you think companies need to adopt it now in order to remain competitive? 

RJ: Right now, I’m seeing there’s a big FOMO out there!  

I think it’s a question on everyone’s mind is, ‘do we need this to be competitive?’ I don’t think generative AI is going to be the thing that organisations use to compete. But I do think it’s going to be table stakes for every organisation. 

Q: Is there a cost-saving element as well? 

It will be unique to every single context and the question you need to ask is, ‘is this simply a redistribution of budgets? Are we redistributing line items to processing power and large language model uptime? 

For some industries, however, buying a subscription will be a cost-saving. I’m thinking of a marketing agency that is able to reduce the number of writers they use to generate content, for example. 

Q: Is it beneficial to be an early adopter of Gen AI?  

This is true for any technology really; the early adopters take on the risk, but their risk appetite is naturally higher. Laggards, by their very nature, have a lower risk appetite. So, if something does fly, the early adopters get the benefits.  

Laggards take their time and can see everybody else making mistakes but also risk losing market share.  

It comes down to the risk appetite of the organisation itself. And one of the things we’ve seen, across sectors, across size of company and even across the enterprise, is that different companies have different risk appetites. Even for something like a regulatory requirement, for example, MIFID, that came in and was large scale and disruptive across the entire industry. Some people were implementing it to its full extent, going beyond the regulation and applying everything in the spirit of the regulation. Others were doing the bare minimum. There are different scales of risk – some organisations would never want to pay a fine reputationally; others would roll with it. So, when it’s a technology to adopt. They’ll either embrace it wholeheartedly and think in the spirit of how we maximise the value of this or not. 

Q: Has Gen AI crossed the chasm? 

RJ: In terms of being fully adopted, we’re not seeing full-scale, mass adoption just yet. 

We’re seeing some usage. Many organisations already use assisted Chatbots. But on a large scale, Gen AI hasn’t crossed the chasm yet. 

We’ll be talking much more about Gen AI and its uses in software engineering over the coming months. If you would like to find out more, please get in touch [email protected]