Skip to main content

Small language models: a big idea for the public sector?

It's likely there are hundreds of solutions that utilise generative artificial intelligence (GenAI) already in use across the public sector. Where I work we're already using AI for a variety of tasks, from drafting reports using i.AI's pilot of Minute, or writing job applications which we developed ourselves.

Under the surface of these solutions you'll find many have at least one thing in common; They are essentially user interfaces for large language models (LLMs) owned by OpenAI, Amazon, Google, or Meta.

We've seen a lot about scaling up the use of AI in the public sector recently, and this letter from the Department of Science, Innovation, and Technology states that the Government calculated their efficiency targets using the assumption that 100% of routine tasks could be automated.

Whether that's achievable is a question for another time, but even if half of that target was reached, that's 50% of UK public sector tasks essentially outsourced to US tech giants.

A Small Language Model (SLM) is a compact, efficient version of a LLM, designed to perform well with fewer computational resources.

Unlike LLM  such as OpenAI's GPT-4 (with trillion-scale parameters), SLMs typically range from a few million to around 10 billion parameters, making them faster, cheaper, and easier to deploy especially for specialised tasks.

Most interaction with the public sector is for a single simple task; renewing a passport, ordering a new bin, making a doctor's appointment, and so on.  I know this is true in local government having looked at councils' website analytics. There will always be users who need help from multiple teams or organisations, but that sort of requirement wouldn't be a routine task.

So here's my notion: small language models could save the public sector.

Lets take planning for example. Generative AI for planning doesn't need a LLM that's been trained on the collective works of Shakespeare, the laws of thermodynamics, and the synopsis of every episode of Friends, it just needs a SLM that's been trained on UK planning law, planning application formats, planning report formats and so on.

Using a LLM is like trying to do your weekly shop in a monster truck. Sure if you can make it road legal (or in AI terms, get it through a governance process) it's probably possible, but you'll take up much more space and use far more fuel that you actually need.

SLMs could be built, or commissioned, hosted, and owned by the UK Government, the NHS, or local government, therefore retaining control here, and could be updated quickly when UK legislation changes.

I would love to know if this approach is already being thought of, or even adopted in the UK, or anywhere else in the world, and as always, I welcome your comments below.


Comments

Popular posts from this blog

Digital best practice checklist

This week I finished the draft of a digital best practice check-list. It's not digital strategy, in fact I'm increasingly thinking organisations don't need a digital strategy, they need a delivery strategy. My draft has check-list of seven questions and recommendations, with one overall recommendation regarding best practice for delivering digital. Ideally it would be incorporated into a wider service and information delivery strategy. Below I've omitted the bulk of the content, the reasoning behind arriving at the recommendation from the question because it's still in draft, but here are the seven questions and eight recommendations: 1. Is the council properly promoting its digital services and content, to reduce avoidable contact? Recommendation: Establish a “digital first” ethos to the promotion of services and better targeting what, when and where they're promoted. 2. Are the digital services the council offers, especially where the design and...

Carl's Conundrum of Internal Influence

I'm writing this partly as a reply to an excellent piece that Carl Haggerty published about the disconnect between internal and external influence and partly due to various conversations over the past month about how to make using tools like collaboration platform  Pipeline common practice. This isn't really about Carl though, or Devon County Council, or any other council specifically, it's more a comment on the influence of digital teams in local governments, or lack of, and how to resolve this. So here's the question that prompted this piece. How can someone who's been recognised nationally for their work, first by winning the Guardian's Leadership Excellent Award and who has more recently been placed in the top 100 of the Local Government Chronicle's most influential people in local government , "sometimes feel rather isolated and disconnected to the power and influence internally". First, let's consider whether is this a problem to...

Pipeline Alpha

In September 2014, officers from 25 councils met in Guildford to discuss a platform to enable collaboration across Local Government. A "Kickstarter for local government" is the missing part to Makers Project Teams , a concept to enable collaborative working across different organisations put forward by LGMakers the design and development strand of LocalGov Digital . Based on the user needs captured at the event, LGMakers created collaboration platform Pipeline and by October people from over 50 councils had signed up . Pipeline is an Alpha, a prototype set up to evaluate how a Kickstarter for councils might work. It is a working site though, and is being used as the platform it is eventually intended to be, at present without some of finer features a live offer might have. So what have I've learnt in the eight months since we launched Pipeline? There's a strong desire to collaborate  LocalGov Digital isn't a funded programme. I wrote about how much it ...

Superfast highways

You may have seen this slide I put together to help explain digital transformation This week we launched a new beta service to report speeding traffic. It looks fairly simple but to give you an idea of what's happening in the background I thought it might be useful to show you the before and after. So here's the before and as you can see it's completely a manual process. Stuff might be recorded electronically but it takes someone to do something seven time to make the process work and send it to the parish or the district. Here's the after What this doesn't tell you is that it's basing whether the request is for the parish or district on three questions. It's also doing a spatial look up to find the parish and returning the parish clerk details using the Modern.Gov API. Because these are already part of our platform this is data that we currently maintain, so there's no additional work to keep this up to date and we've reduced the h...

Defining transformation to a wider audience

For the past month I've been putting together a paper on the next steps of digital transformation, for the organisation I work for. I'm proposing we look at two capabilities and two business areas, and if approved I'll be writing more about it. It's been a great exercise in gathering my thoughts and helping me to define digital transformation to a wider audience and how it fits into the bigger picture of service improvement. Here's some of the stuff I've learnt or had affirmed: Transformation, digital or not, starts with understanding the needs of the user through research. This should be obvious, but in local government too often I've seen "build it and they will come" approach applied. It's unlikely a commercial operation would launch a new product without first researching the market, so why would a digital service be any difference? A couple of years ago I wrote how the phrase "digital transformation" was hindering digit...