Having already reflected on my predictions for 2024, it feels that the past year has been somewhat chaotic in the land of AI. With constant model releases, battles on benchmark leaderboards, and a trial and error approach to developing solutions, it was clear that the world needed some time to find its feet and really think about the implications of AI technologies on society and the ways we work. But, if 2024 was the year of noise, then 2025 is the year of order. Or at least, the year we start to put things into order. Now the 2024 dust has settled, we find a lot of organisations reflecting on what the future really looks like for them, and how they can use a promising technology in a way that makes sense for their business. It will be an interesting year ahead, and below I've pulled out 5 key areas that I'm paying attention to in 2025...let's check in again at the end of the year to see what really happened.

‍

01. Agents, agents, everywhere!

Let's get this one out the way, AI agents became the big genAI topic of the last part of 2024, and 2025 is likely to see that popularity surge as OpenAI continue to push their o1 model, and focus on how they can chain model outputs together to tackle more complicated workflows. However, that will come with the same challenges that came with LLMs and AI assistants, as developers and users trial and error to find the right use cases. Questions will also arise around the cost-effectiveness of these solutions, as prices to deliver these workflows remain higher than automations (and in some cases higher than for a human to complete them). Whilst it is likely that agent based systems will stick around and become more sophisticated in the future, companies pushing agents will focus largely on language use cases and automations for now. Think: marketing/CRM, web scraping, and collaborative tools.

‍

02. OpenAI struggles to lead

The gap between OpenAI and the chasers closed significantly in 2024, and 2025 could be a challenging a year for the company as it struggles to build a moat around its product. Questions had already started to emerge about the company's ability to maintain its lead as the chasing pack of LLMs significantly closed the gap, and OpenAI shifted focus to chain-of-thought "reasoning" and additional features such as search. Since penning this prediction, OpenAI has already faced a significant challenge from DeepSeek, the Chinese competitor that has matched its models in benchmarks andintroduced a new method for training models. The challenge is exasperated further by DeepSeek's decision to publish a paper detailing their training and technical approach, as well as open sourcing their model for anyone to use. So far, OpenAI's response has been to release further features built on top of their model such as the newly announced "Research".

‍

03. Year of the AI builders

As the initial wave of fear, excitement, and noise has calmed down, and buyers/developers are beginning to understand more clearly what AI can do, conversations are turning more towards "What can AI do for me?". Organisations are beginning to learn more about the data and processes that allow AI models to be built, and explore how they could tackle challenging workflows within their organisations with this technology. We can expect a major focus on building AI solutions in 2025, with 64% of CEOs listing AI as a top investment priority. This will show in two ways: organisations will begin to demonstrate how they have built their own internal AI workflows to empower their businesses, and we will begin to see more comprehensive AI powered products gaining traction with a clear value proposition.

‍

04. Embedded Language Models

The drive to create smaller models is not just about ecology and efficienct, it's also enabling language models to be embedded into hardware. By growing the capabilities of smaller models, we are enabling engineers to explore how they can run models on devices with limited storage, memory, and compute power, which means: we may begin to see a wave of language features appearing on other devices. We're already familiar with smart speakers such as Alexa, but we may now start to see them deployed locally within our phones (allowing access to the models without internet connection), within robots to allow you to give spoken instructions, or perhaps even within appliances as people re-explore the smart home dream. With these products just around the corner, it's likely that we will see early version of these hit the market in 2025. It will be interesting to see how this advancement influences the design of products.

‍

05. Increased strain on AI Skills

As the AI demand refuses to ease, and as organisations are realising the potential of AI, and facing pressure to show that they are doing something about it, we will see a lot of businesses committing to building skills within their teams. The 2025 World Economic Forum Jobs Report published at the start of January highlighted that 70% of respondents intended to hire for AI skills. This creates a major challenge, as the market for these skills is already limited. There is a high likelihood that organisations will struggle to fill roles and retain staff with AI skills as companies jostle to build their own internal capabilities. It will take time for universities, apprenticeships and other training routes to deliver the talent needed to relieve this demand. One route some organisations may consider is upskilling/reskilling internal staff, however 63% of survey respondents considered this a major barrier, with it likely to be costly to invest in and a need for training organisations to step up and provide support.

‍

The AI world is moving fast, and it will be interesting to see how organisations adapt to the technology in 2025, I look forward to returning to these predictions at the end of the year to see what actually played out.

What is Federated Learning?

In today's world, the importance of data privacy and security cannot be overstated. With the exponential growth of connected devices and the ever-increasing volumes of data they generate, traditional centralised machine learning approaches face significant challenges. Enter federated learning—a revolutionary paradigm that promises to reshape the landscape of artificial intelligence (AI) by enabling collaborative model training without compromising data privacy.

Federated learning is a decentralised approach where multiple devices or servers collaboratively train a shared model while keeping the data localised. This innovative technique allows organisations and individuals to harness the collective intelligence of distributed data sources without the need to transfer sensitive information to a central server. Imagine hospitals around the world collaboratively training a medical AI model on patient data without ever sharing the sensitive information. The potential for breakthroughs in healthcare, finance, and beyond is immense.

But how exactly does federated learning work, and what makes it such a game-changer in the realm of AI? We can delve into the fundamentals of federated learning, explore its numerous benefits whilst addressing the challenges that come with this groundbreaking technology.

Definition

Federated learning is a decentralised machine learning approach where multiple devices or servers collaboratively train a shared model without exchanging their local data. Instead of sending raw data to a central server, each device computes model updates (gradients) based on its local data and then sends these updates to a central server. The server aggregates these updates to improve the global model, which is then shared back with the devices. This process continues iteratively, allowing the model to learn from distributed data while preserving privacy and reducing data transfer.

In essence, federated learning enables collaborative learning while keeping data localised, ensuring data privacy and security.

How it works

Data Stays Local

In FL, data remains on individual devices instead of being centralised. This means your personal data never leaves your device, maintaining privacy and security.

Local Training

Each device trains a copy of the global model using its local data. For instance, your smartphone might improve its predictive text model based on your messages, without sharing any content.

 Sharing Model Updates

Devices send only model updates (gradients) back to a central server, not the raw data. This ensures privacy while still contributing to the model's improvement.

Aggregation

The central server aggregates updates from all devices to refine the global model. Techniques like Federated Averaging combine these updates to enhance the model.

Iteration

The improved model is sent back to devices, and the process repeats. This iterative cycle allows the model to get better while keeping data private. 

Benefits

Data Protection

Federated learning offers significant advantages in terms of data protection and privacy. By keeping data on local devices, FL ensures that sensitive information never leaves its source. This decentralised approach means that personal and proprietary data remains secure, minimising the risk of data breaches and unauthorised access. It complies with stringent data protection regulations, such as GDPR, by ensuring that raw data does not leave the network, thereby preserving user privacy.

Limiting Data Transfer

Another key benefit of federated learning is the reduction in data transfer. Traditional machine learning methods require transferring large amounts of data to a central server, which can be costly and time-consuming. In contrast, FL only sends model updates, significantly reducing the bandwidth and computational resources needed for data transmission. This efficiency makes it particularly suitable for edge devices and IoT applications, where network connectivity may be limited or expensive.

Improved models and collaboration

Federated learning enables the creation of improved models through collaborative efforts. By leveraging the diverse data distributed across various devices, FL can capture a wider range of patterns and behaviours than a single centralised dataset could. This leads to more robust and generalised models. Moreover, FL fosters collaboration between different organisations and entities, allowing them to collectively train AI models without sharing sensitive data, therefore accelerating innovation and advancements across industries.

Use cases

SAFERAI

At Raiven we are using federated learning to transform the manufacturing industry, particularly in health and safety risk prediction, through initiatives like SAFER AI (Safety Advancing Federated Estimation of Risk using Artificial Intelligence - SAFER AI). We collaborate with P&G to train a shared AI model that predicts risk in a manufacturing environment, thus improving workplace safety without sharing sensitive operational data. Other companies can train the model on their own local safety data, then sends model updates to a central server. These updates are aggregated to enhance the global model, ensuring robust predictions across diverse environments. This approach enables companies to benefit from collective insights, enhancing safety standards while maintaining data privacy and protecting proprietary information. SAFERAI exemplifies how federated learning fosters a safer and more secure industrial landscape.

In an era where safeguarding data privacy and security are paramount, the convergence of privacy preserving machine learning and IoT technology presents both immense opportunities and daunting challenges. The need to prioritise trustworthy AI through development that is robust, explainable and transparent is an urgent priority.

Recognising this pivotal moment, OctaiPipe, a pioneer in federated edge AI innovation, has joined forces with industry leaders P&G, SME partners Site Assist, Raiven and Digital Catapult and InnovateUK to deliver a groundbreaking SaferAI concept. groundbreaking concept for safer AI

“Bringing together SMEs, large organisations and research organisations, these novel solutions will demonstrate how trusted AI and machine learning technologies can aid and be incorporated into many of the UK’s industries and sectors.”

Dr Kedar Pandya, UKRI Technology Missions Fund Senior Responsible Owner

“Federated Learning with OctaiPipe is unlocking potential not yet realized in industry. It’s enabling competing business to collaborate on the use of technology and AI for good. Specifically, within SaferAI, I can see significant value in utilizing federated learning to share model results for predicting safety incidents and risks. I can see this changing approaches in industrial safety. Before, we could only cooperate on best practices, technology or processes, but now we can actually prevent incidents with the use of AI and shared data.”

Christoph Wagner-Gillen, P&G Product Supply HSE Governance

SAFER AI

SaferAI, or as the projects catchier title suggests, Safety advancing federated estimation of risk using AI, is a 12-month, £1.64m project focused on the development of a highly comprehensive federated learning model for predicting incident occurrences and risk.

SaferAI is an industry first, enabling world leading consumer goods manufacturer, P&G, to collaborate with other interested Consumer Goods Companies to forecast safety incident rates to reduce Health & Safety (H&S) incidents.

Until now, with low event occurrence in any single setting, there has been insufficient data availability to predict future incident events or near-misses. However, using federated learning with OctaiPipe, multiple large-scale parties with similar Health, Safety and Environment (HSE) situations can now collaborate to pool intelligence to achieve scale.

‍

HSE in the workplace

Many organisations already collect high-level operational and HSE incident data intelligence through various Industrial IoT devices and cameras to successfully predict safety incidents. However, analytics based on this is not meaningfully actionable to drive changes that preventatively reduce risks.

For predictions to be meaningfully actionable, they must be made at a sufficient level of granularity within a factory or workgroup.

Whilst we’re fortunate that HSE critical events are rare within single sites or even organisations, it means insufficient data exists to employ ML models capable of predicting when and why H&S incidents might occur so they can be prevented.

To date, progress towards an AI-enabled solution has been impeded by:

    a) a requirement for more observations than one organisation can generate alone, so it is imperative to share data, and

     b) barriers to sharing data across organisations that, until now, have not been overcome.

Federated learning solves this. SaferAI will enable organisations to combine data to facilitate actionable incident predictions for small work groups.

‍

‍Federated Learning

Federated Learning is a privacy preserving machine learning technique that allows for machine learning models to be trained across multiple distributed edge devices without the need to see or move the data.

It is a key technology in the convergence of AI, IOT and connectivity as it enables edge intelligence whilst safeguarding sensitive or private data.

Federated Learning will be used the project to establish a viable on-device (Edge) AI system that pairs privacy-focused telemetry and computer vision (CV) systems. It will also enable continual collaborative learning in edge environments, combining model results provided by collaborating organisations to yield a highly comprehensive HSE model.

During SaferAI, OctaiPipe are further developing and validating as a high-trustworthiness FL-for-IoT system along with a high-accuracy model that uses observed use case data generated by AIoT camera systems. This solves the challenge of predicting safety critical incidents, near misses and hazardous/non-compliant conditions in industrial settings, enabling organisations to pre-emptively take remedial/mitigating actions to reduce H&S risks.

‍