Top AI Trends to Watch in 2024

  • Author Richard McIntyre, BJSS

  • 17.01.2024

Chief Engineer at BJSS, Richard McIntyre, shares a glimpse into the top AI trends that are set to dominate this year.

Just over a year ago, many of us had never heard of Generative AI or Large Language Models (LLMs). However, 2023 was a big year for Artificial Intelligence and I do not see this trend going away in 2024.

We’ll continue to see further maturing of the landscape, including new tooling that I am sure will blow our minds, and help us do our jobs more enjoyably and efficiently.

I must admit, I am a bit of an AI fan, I enjoy being surprised by how an AI model or tool can create content better than what I had imagined. As I continue looking for tools that help me on a day-to-day basis, let me share the types of tooling that currently exist, how they differ from each other, and offer some interesting examples.

4 Types of AI Tooling

1. Specialist task tools

Many large language models and tools such as ChatGPT are incredible at doing a whole wide range of tasks, however, they are aimed for broad use and might not be tuned to specific tasks. Now, many tools, that may use LLMs under the hood, are designed to do one exercise and do it well.

This space is growing (massively), and the time is ripe for these vendors to create useful software-as-a-service platforms get better outcomes for domain specific tasks that go beyond the output of a generic Generative AI service.

Here are some examples:

Grammarly – AI authoring tools are not completely new. Grammarly has been about for a few years. It is brilliant to help you improve any text that you might be writing, helping you with style and delivery effectiveness. I am likely to check this prose on Grammarly before I post it. Grammarly now has Generative AI to enable you to set your voice, help generate content, and identify gaps in your writing.

Copy.ai – A service that gives you a rich toolchain and workflow for creating copy for a multitude of purposes. I even got it to help me write an outline for this blog post – I did abandon it however, as I didn’t want to feel like I was cheating. You can feed it a ‘brand voice’ so that it sounds just like you.

There is also a whole suite of useful tools such as competitor analysis, a plagiarism checker, short LinkedIn connection notes, and more. If you create a lot of content, I would definitely sign up for this service.

Midjourney AI – You may have heard of Dall.E, which is now built into ChatGPT 4, should you have a ChatGPT Plus subscription. Midjourney is a great contender and is being used by many to not just create individual images but even to do design layouts for websites.

But Why would you use Midjourney over DALL.E?

Midjourney gives you far more control over the image, including adding weighting to various parts of the image, eliminating elements from the image using –not and even camera and lens types.

2. Helpers Built Into Existing Tools

More and more of the software we are using day to day, now have AI assistants built in, here are few stand-outs.

Microsoft 365 – At BJSS I have been enjoying having Copilot functionality injected into Word, PowerPoint, Outlook, etc. I will be honest with you, it does feel like a bit of a Beta product right now, more than not I completely rewrite the content it has suggested. It is, however, amazing at summarising documents, helping you with a writing style, or checking you have the right understanding.

An screenshot of Copilot functionality

Adobe Photoshop AI Photo Editor & Adobe Firefly – I have seen demos of how Adobe has enabled these tools to enable you to enrich photos and images very easily with AI-created content with incredible precision. My mind has been blown! I would love to hear more from content creators as to how they have found these tools work in the real world. Are they just a gimmick? Do you often need to start again, such as with ChatGPT-created content? One thing is for sure, this image voodoo is going to become more powerful, accurate, and widely adopted over time.

An image using ChatGPT to create a website for basketballer, Michael Jordan.

GitHub Copilot – Now GitHub Copilot also has Copilot chat. Developers use their text editors for many hours a day. Copilot has an amazing ability to autocomplete code, comments, and tests. However much of software engineering is not just about writing code, but problem solving. Now with a chat assistant, it means that there is a better way to find solutions and debug errors.

With Copilot Enterprise announced last November, Copilot will also write code in the style of your whole codebase and help you to solve problems in the context of your project/s. Other coding assistants are available!

Apple has not yet turned up in this space, however, I am fully expecting them to turn up and wow us. Check out these current predictions for late 24/25.

3. D.I.Y Tools

Just as the software vendors are creating assistants and software-as-a-service AI based tools, it has never been easier to start building your own tools and products. It might be to make you and your teams more productive or to perhaps, come up with your own business or product idea. AI growth is certainly set to continue in the next few years, this could be your chance to get involved early and be successful.

Foundation models and services – To create your tools, there are several foundational models and services that you can sign up for and get working with quickly.

Microsoft Azure – I have been using Azure AI studio, you have all the ChatGPT models at your fingertips that you can pull into your projects and initiatives. The power you get immediately, and the ease of use are amazing. The one thing I have learned is how unpredictable these LLMs are, however – so be careful!

Google – Was the next contender at the heels of OpenAI, whose Bard project immediately showed this was not a one-horse race. Google Generative AI Studio is a great place to dig deeper

AWS – Although AWS was late to the game, Bedrock has recently been released. It is brilliant Generative AI creation suite, building on top of many popular foundational models to make your own more powerful algorithms that can work around your context and data.

Meta – have a great set of open-source models and libraries that you can get going with. Some can detect hate speech or reinforcement learning. Meta also created PyTorch which is one of the de facto machine learning frameworks for building and running models in your software.

Integration with your documents – as pointed out above, these libraries are gaining much more capability to be context-aware by supplying them with your documents and data. I believe over the next few months; it will be far easier to feed models with your company-specific data to power what appears to be your bespoke custom models.

4. Proprietary Data Preparation and Enriching Tools

Data, data, data‘ is what we have been hearing to make a success with AI, some of us will want to start collecting more data and some will want to better prepare the data we already have. Platforms such as Snowflake are incredibly powerful, but they are also a big commitment and come at a high price.

At BJSS we have been getting quite excited about Microsoft Fabric which helps your business to start thinking in a data-centric way and provides the tooling to do this. It also can lay models on top of this data and then visualize those insights through PowerBI and other analytics tools. One to watch out for!

AI Improvements

Chat As An Input Device – AI assistants sitting next to you in your applications is how many software vendors are betting you might see better productivity whilst using their tools. Throughout 2024 we will see them appearing in more of the software we use.

They will also get faster, accurate and more mature. Currently, they feel a little clunky, and cumbersome and often create content for you that you could create better the first time itself (depending on the task).

This will improve; however, they are a ‘Copilot’ and not designed to take over from you completely.

Integration With Your Documents – Many of these assistants have until now, just been guessing context from the documents that you are currently in.

Being able to feed other business documents in to the software, point it at your wiki or other Excel data for example, will mean that the results of prompts will be far more in line with your business domain and expertise, making them far more powerful.

Closing Thoughts

The power that is available to us in these tools is incredible. All of these were almost unthinkable just over a year ago. They are still new; some provide immediate value and some need better coaching and ‘prompting‘ from us to produce meaningful results. They are often also a little slow, and I am sometimes thinking to myself ‘Once this response comes back, will it even be helpful?’.

The tools will get more useful and powerful, and we will also get better at using them.

I have not even scratched the surface of what other tools are out there in these categories. I encourage you to go out and find the tools for the problems you are trying to solve; I am sure there is much out there to help you work smarter not harder!

All events

All sponsors