Pure Telecom, Ireland’s high-speed broadband and telecoms provider, today announces the results of its annual Connected Lives survey which reveals 370,000 adults in Ireland have engaged in a romantic relationship with an AI chatbot within the last 12 months. The research indicates a growing interest in AI for romantic connections, with a further 12% of adults – almost half-a-million people – revealing they wouldn’t rule it out in the future.
The nationally representative survey was conducted by Censuswide on behalf of Pure Telecom, surveying 1,001 adults in Ireland. The research examines the evolving relationship between humans and AI, in particular their personal relationships with chatbots such as ChatGPT or Gemini. As the lines between humans and technology become increasingly blurred, chatbots have opened the door for people to explore a new form of emotional connection.
Much like with human partners, AI ‘relationships’ are formed when an individual develops a sense of attachment to an AI chatbot through the sharing of daily experiences and engaging in deep conversations. The bot reciprocates this affection and can recall previous conversations, thus building a rapport and reinforcing the recipient’s perception of a relationship.
Pure Telecom’s research revealed that in the last 12 months, 13% of men have conducted a romantic relationship with an AI chatbot. This is almost double the 7% reported by women. Across age groups, the figure was highest among 25-34-year-olds at 16%.
The survey also indicated an openness among many to conduct a romantic relationship with an AI chatbot at some point in time. Almost half-a-million (12% of adults) wouldn’t rule out a relationship with an AI bot in the future. In keeping with this outlook, 20% believe romantic relationships with AI would be less complicated than humans ones. A further 10% maintain romantic relationships with AI are a good way to practice real life relationships. This belief is higher among men at 16% in comparison to women at just 6%.
The humanisation of chatbots doesn’t end there. Almost one in five (19%) of adults speak to AI like it’s a friend, while 36% admit to saying ‘please’ and ‘thank you’ to their AI chatbot. Furthermore, 18% use it to research health symptoms and treatments, while one-in-10 use it as a form of therapy and to get life advice. The same proportion use it to prepare for difficult conversations, like a break-up or confrontation.
Paul Connell, CEO, Pure Telecom, said:“This research highlights the pivotal role that AI is beginning to play not just in our professional lives, but our personal ones also. As people and technology become increasingly integrated, and AI becomes progressively more advanced, adults in Ireland have found it to be an unexpected source of emotional connection. The recent AI boom means that these bots are now readily accessible to all of us – and there’s no agony of waiting around for a reply!
“While relationships with bots may seem unconventional, it underscores the remarkable capacity of artificial intelligence to foster connections as it becomes increasingly responsive to our needs. The use of these large language models (LLMs) requires fast, reliable internet access and as a provider of this, we at Pure Telecom are excited by the possibilities that AI unlocks. However, it is no replacement for the rewarding spontaneity and depth of human connection.”
Every morning, I read a handful of AI newsletters over coffee. Not for the hype or the hardware updates, but for something far more practical. Prompts. Tucked between the headlines and think pieces, you’ll often find “Prompt of the Day” sections, promising to unlock the full potential of generative AI. Some are short, snappy, and direct. Others read like onboarding manuals for a new employee. And that got me thinking. When it comes to prompting AI, what really works best, and when?
Does Prompting Really Matter?
Yes it does. Short and simple. Prompting is not something you should ignore. It’s still worth putting thought into how you communicate with AI, especially reasoning-focused ones, which tend to respond better to well-structured prompts. And if you’re writing code that integrates LLMs, rather than just chatting with them, crafting a strong prompt becomes even more important. Obviously, for fooling around or chatting with AI companions, it doesn’t matter how you approach prompting. But if you’re using language models for learning, casual exploration, or answering questions, your energy is often better spent understanding what they’re good at, and where they fall short, rather than obsessing over the perfect phrasing.
With that being said, here’s when you should use short or long prompts.
The Short Prompt Advantage and Why I’m a Short Prompt Guy
There’s something to be said for brevity. Short prompts are like tossing out an idea in a group brainstorm. You don’t know exactly what will come back, and that’s half the fun. They give you room to respond, adapt, and riff. It’s not a rigid structure. For example, if you start with:
“Give me 3 creative ways to thank a customer.”
You’ll come with a short list: handwritten note, social media shoutout, personalized gift. Nothing groundbreaking, but enough to spark a direction if you’re not sure which one to go with. Then, if you follow up with something like:
“Make the gift idea feel more unique. The customer is a record enthusiast.”
You’ll be looking at suggestions like limited-edition vinyl pressings or custom playlist curation. Short prompts work great here. There’s also a practical issue to consider when it comes to long prompts. Even the most advanced AI models can struggle when overloaded with too many instructions at once. The result? They’re more likely to produce inaccurate or confusing answers, sometimes called hallucinations, or just ignore parts of the prompt entirely. Shorter, more focused prompts make it easier to review the AI’s output and refine it step by step, keeping you in control and aligned with your original goals.
For that very reason, I prefer short prompts. For me and the things I use AI for, short prompts keep things fast and flexible. I can toss an idea at the AI, see what comes back, and tweak it in real time.
It’s not that I don’t like AI taking over. It’s more about steering, giving it just enough to start moving, then guiding it where I want it to go. I stay in control, adjusting as needed until it gives me something I can actually use. That’s the power of short prompts. It’s like creative ping-pong. You’re actively shaping the result rather than just watching it unfold. Of course, not everyone works this way. Some people prefer a detailed prompt to set the scene, and that’s fine.
But Don’t Count Long Prompts Out
Now, let’s be fair. There are plenty of people who live and die by the detailed prompt. You know the type. They write prompts that read like creative briefs, with personas, goals, formatting instructions, and disclaimers. And to be honest, sometimes those deliver scarily precise outputs. Long prompt may look something like this:
“You’re a creative marketing manager for a new energy drink startup. You’re prepping a press release for a product that promises energy but doesn’t contain caffeine. The goal is to sound confident but not pretentious. Mention sustainability. Keep it under 300 words. Use Gen Z humor.”
That’s a mouthful that may or may not get you there. You’re shaping the character, tone, purpose, all up front. If you know exactly what you want, a long prompt may get you there in fewer steps. The issue, though, is what happens when things go wrong. When your output is off, and with long prompts this happens more often than you’d expect, it’s harder to pinpoint what part of the prompt broke the logic. Did the tone fall apart because you mentioned Gen Z humor? Did the AI hallucinate a fictional product launch because your setup was too specific? Like mentioned earlier, shorter prompts will make things easier for AI to follow.
The Flexibility Factor
Long prompts are like formal contracts. They’re structured. Specific. There’s less room for improvisation. But when you’re experimenting, or when you’re not entirely sure what you want yet, short prompts leave breathing room.
Let’s say you’re writing a product description. You could type:
“Write a product description for a smartwatch.”
Or you could go further:
“You’re a copywriter at a high-end tech brand releasing a minimalist smartwatch designed for busy professionals. Write a short product description for the landing page. Highlight features like sleek design, long battery life, and smart notifications. Keep the tone confident, modern, and clean. Avoid buzzwords. Use short, punchy sentences.”
You see the difference. The first one gives you something to react to. The second is great if you’ve already figured out the vision. But if you’re still deciding on the tone, audience, or even the event’s name, then a shorter start gets the job done faster or get some online tutoring.
At the heart of this debate is a simple question: Do you want to start perfect, or get there step by step?
Short prompts are for iterators. Builders. People who enjoy watching the idea evolve in real time. Long prompts are for architects. People who know what they’re building before they break ground.
Neither is wrong. But knowing which style fits you best can save you a lot of time, and a lot of weird AI misunderstandings. My suggestion is to take a free prompt engineering course like this one and learn best practices and how to apply them to your work.
Find Your Own Style
So what’s the verdict? Are short prompts better? Are long prompts overrated? This article is not “good vs bad prompts”, so there’s no definitive answer here. For me, short prompts all the way. But that’s the thing. That’s me. That works for me personally. The best prompt is the one that fits how you think. Some people are outliners. Others are improvisers. AI doesn’t care either way. It’s not grading your prompt. If you’re someone who thrives on structure, you might find that longer, more detailed prompts hit the mark. It all comes down to what helps you think clearly and get the results you want.
Need a quick spark of inspiration? Keep it short. Looking for a more guided, in-depth response? Go long.
Here’s something to try: next time you’re working with AI, run an experiment. Ask it to “brainstorm vegan dinner ideas” in ten words. Then do the same thing with a detailed, 100-word setup full of context and specifics. Compare the answers.
You can also take a long prompt, maybe one you found online, and break it into pieces. Feed them in step by step, only moving forward when you’re satisfied with each part. You might find that layering your input this way gives you more control over the outcome. Which version feels better in your hands? Which one gives you answers that actually move the needle? There’s no single rule here. No prompt formula that works for everyone. It’s just about finding your rhythm, and once you do, the whole process gets a lot smoother.
Logos have long been the visual shorthand for brand identity, but in 2025, the process behind creating them is shifting rapidly—and it’s being driven by artificial intelligence. From bold minimalism to dynamic responsiveness, new trends are emerging not just from designer studios, but from the neural networks powering AI design platforms.
So what do these tools reveal about where branding is heading?
The Minimalist Revival Continues
Flat design and minimalism are nothing new, but AI has pushed these aesthetics even further. Algorithms trained on thousands of successful logos are favoring clean geometry, sans-serif typography, and restrained palettes. These styles are scalable, versatile, and align perfectly with cross-device demands.
Expect to see more logos in 2025 that ditch heavy gradients and intricate patterns in favour of simplicity and impact.
Logos That Adapt in Real Time
Dynamic logos—those that change depending on context—are becoming mainstream thanks to AI. Brands like Spotify and Google have experimented with responsive logos that subtly animate or shift based on screen size, user mood, or data inputs. Now, smaller companies are gaining access to similar capabilities through AI-powered platforms.
This “living identity” not only captures attention—it builds emotional resonance and boosts recall.
AI-Assisted Personalization
One of the most significant shifts in branding is the democratization of design. Brands no longer need to hire expensive agencies for professional logos—AI logo makers are turning that around.
Even the free version of an AI logo maker can generate multiple unique designs from just a few inputs—like business name, color preferences, and domain—while allowing robust customization afterward. And we are just at the start, while they keep updating and expanding on their capabilities.
Data-Driven Logo Optimization
In 2025, logos aren’t just designed—they’re refined continuously based on performance. AI tools now allow brands to A/B test logo variations across platforms and measure metrics like engagement, recognition, and emotional response. Real-world feedback loops, powered by machine learning, are helping brands tweak colors, spacing, and even icon shapes to maximize visual impact and user connection.
This iterative, analytics-backed approach transforms logos into flexible assets that grow with the brand. By integrating logo performance data into branding strategy, businesses can ensure their visual identity isn’t just trendy—it’s effective.
Emotionally Intelligent Design Gains Traction
AI is also enabling logos to tap into emotional intelligence. Through sentiment analysis and behavioral data, design platforms can suggest color palettes and shapes that evoke specific feelings—trust, excitement, calm, or urgency. This trend is especially powerful in industries like wellness, fintech, and education, where emotional connection drives loyalty.
A recent report backs this up, showing a surge in “emotion-first” design strategies where logos adapt to mood, time of day, or user journey. Brands are increasingly using flexible design systems that shift subtly based on real-time engagement—making the logo feel more alive and relatable.
Ethics and IP: Complications Ahead
However, the rapid rise in AI-generated logos brings forward ethical and copyright concerns. If multiple businesses use the same AI tool and prompts, how original are the outcomes? If AI-generated logos infringe on something else, whose responsibility is it?
The challenge lies in the way most AI systems are trained. These models analyze and learn from vast datasets of existing logos—some of which may be protected by copyright or trademarks. While AI doesn’t replicate logos verbatim, it can generate designs that resemble or echo elements of copyrighted work, sparking potential IP conflicts.
Legal experts advise caution: businesses should treat AI-generated logos as drafts—not final, exclusive assets—and refine them to ensure true distinctiveness and compliance. This often involves manual adjustments to layout, typography, or symbolism, backed by professional review and trademark vetting.
There’s also a growing concern over accountability. If a logo violates another brand’s rights, is the liability on the business, the AI provider, or the original training dataset? Currently, legal precedent places the burden on the end user—meaning brands should take extra care before launching an AI-designed identity publicly.
Ethical branding goes beyond legality. When dozens of businesses rely on similar tools and templates, the risk of visual homogenization increases. Brands may unintentionally dilute their uniqueness by deploying generic AI-generated logos, undermining the very purpose of visual identity: to stand out.
What This Means for Branding
AI isn’t replacing creativity—it’s amplifying it. Hybrid approaches, where designers use AI-generated concepts as foundations and then apply their own artistry, are emerging as the gold standard. Logos in 2025 are evolving fast—and not just visually. AI is transforming how logos are conceived, who can create them, and what they represent. The trends show a future where branding becomes more accessible, responsive, and data-informed.
Across EMEA, Artificial Intelligence (AI) is redefining industries, inspiring innovation, improving operations, and driving, growth. Government and Irish businesses are embracing and capitalising on AI’s potential to enhance customer experiences and gain a competitive advantage. But as adoption accelerates, new security challenges arise, demanding vigilant attention to protect these investments Ivor Buckley, Field CTO at Dell Technologies Ireland explains more
Forecasts indicate that AI could contribute trillions to the global economy by 2030, with Ireland well-positioned to capture a significant share of this value. According to Dell Technologies’ Innovation Catalyst Study, 76% say AI and Generative AI (GenAI) is a key part of their organisation’s business strategy while 66% of organisations are already in early-to mid-stages of their AI and GenAI journey.
As AI becomes more embedded in everything from customer management to critical infrastructure, safeguarding these investments and tackling the evolving cyber threat landscape must be a priority. To that end the success of integrating AI in the region depends on addressing three critical security imperatives: managing risks associated with AI usage, proactively defend against AI-enhanced attacks, and employing AI to enhance their overall security posture.
Managing the Risks of AI Usage
Ireland as a digital hub within the EU, must navigate the complex regulatory environment like the Digital Operational Resilience Act (DORA), NIS2 Directive, the Cyber Resilience Act and the recently launched EU AI Act. These frameworks introduce stringent cybersecurity requirements that businesses leveraging AI must meet to ensure resilience and compliance.
AI’s reliance on vast amounts of data presents unique challenges. AI models are built, trained, and fine-tuned with data sets, making protection paramount.
To meet these challenges, Irish organisations must embed cybersecurity principles such as least privilege access, robust authentication controls, and real-time monitoring into every stage of the AI lifecycle. However, technology and implementing these measures effectively isn’t enough. The Innovation Catalyst Study highlighted that a lack of skills and expertise ranks as one of the top three challenges faced by organisations looking to modernize their defenses. Bridging this skills gap is vital to delivering secure and scalable AI solutions because only with the right talent, governance, and security-first mindset can Ireland unlock the full potential of AI innovation in a resilient and responsible way.
A further step that Irish businesses can take to address AI risks, is to integrate risk considerations across ethical, safety, and cultural domains. A multidisciplinary approach can help ensure that AI is deployed responsibly. Establishing comprehensive AI governance frameworks is essential. These frameworks should include perspectives from experts across the organisation to balance security, compliance, and innovation within a single, cohesive risk management strategy.
Countering AI-Powered Threats
While AI has enormous potential, bad actors are leveraging AI to enhance the speed, scale, and sophistication of attacks. Social engineering schemes, advanced fraud tactics, and AI-generated phishing emails are becoming more difficult to detect, with some leading to significant financial losses. Deepfakes, for instance, are finding their way into targeted scams aimed at compromising organisations. A 2024 ENISA report highlighted that AI-enhanced phishing attacks have surged by 35% in the past year, underscoring the need for stronger cybersecurity measures.
To stay ahead organisations must prepare for an era where cyberattacks operate at machines’ speed. Transitioning to a defensive approach anchored in automation is key to responding swiftly and effectively, minimizing the impact of advanced attacks. The future of AI agents in the cybersecurity domain may not be far off.
This means deploying AI-powered security tools that can detect anomalies in real time, automate incident response and adapt evolving threats. Equally important is that business across Ireland need to start fostering a culture of cyber awareness across the workforce, which is supported by AI-driven training tailored to individual risk profiles to counteract evolving threats.
Leveraging AI to Strengthen Security
AI’s capabilities offer organisations powerful tools to fortify their defenses. With its ability to detect vulnerabilities, predict risk, and accelerate response times, AI is emerging as a critical asset in the fight against cyber threats. It can help Irish organisations move from reactive to proactive security postures. The Innovation Catalyst Study found 75% of business and IT leaders say AI/GenAI is a key part of their organisation’s business strategy, with many already seeing tangible results in their cybersecurity strategies.
Here’s how organisations in Ireland can leverage AI to enhance security:
Secure Software Development: AI can improve coding processes by detecting weaknesses early, helping teams reduce vulnerabilities in the development phase.
Advanced Threat Prediction: AI’s algorithms can identify patterns and anticipate potential attack paths, aiding teams in proactive risk allocation.
Enhanced Threat Detection: By processing vast datasets in real time, AI can discern genuine threats from noise with unprecedented accuracy.
Automated Incident Responses: AI tools can significantly accelerate containment and mitigation following an intrusion, reducing response timelines.
User Awareness Programs: AI-powered systems can deliver tailored security training to employees, fostering vigilance and reducing human errors that often lead to breaches.
For many businesses, the adoption of these advanced AI-driven tools will rely on partnerships with technology providers. It’s critical to ensure internal processes and data are structured and simplified to fully support the power of AI-enabled cybersecurity solutions. An automation-first approach ensures that businesses can adapt to a future where autonomous threats are the norm.
Building a Resilient Future
Ireland’s digital future depends on our ability to innovate with confidence and as we know AI has now moved beyond emerging technology status and now plays a central role in digital transformation. That means embedding security into every AI initiative, aligning with evolving regulations and investing in skills, talent and right technology/technology partners is needed to stay ahead of threats.
Companies that approach AI security with robust protections and innovative strategies will not only mitigate risks but position themselves as industry leaders. By addressing the three imperatives of managing risks, countering threats, and leveraging AI for security, businesses here in Ireland can unlock AI’s full potential.
Secured properly, the innovation AI enables will drive sustainable growth for businesses across EMEA, setting them up to thrive in an increasingly digital and data-centric world. The future belongs to those who innovate securely, balancing progress with responsibility.
Logicalis UK&I, a global technology service provider, today announces that it has set the foundation for John Paul Construction’s international future with AI services.
John Paul Construction (JPC) is one of Ireland’s leading contractors with over 75 years’ experience in delivering projects across a wide array of sectors. Currently the business operates in Ireland, the UK and mainland Europe with a reputation for integrity, professionalism, innovation and excellence.
As a result of its business growth and international expansion, JPC had to address language barriers and translation issues within and across its multinational teams, client base and international supply chain. As part of this, it needed to be able to share and interpret critical information and documentation to support operational efficiency, staff productivity and project coordination.
To ease the pressure on and enable seamless communication for JPC teams and external partners, Logicalis deployed an AI solution leveraging Microsoft Azure AI services and Copilot to deliver consistent and reliable translations for the business. This included the creation of a WebApp to allow employees to easily upload, manage, translate and distribute documents.
Able to process 6 million characters per minute, comprehend more than 30 languages, and handle thousands of requests simultaneously, the solution is enabling the company to automate translation and overcome language barriers.
In turn, this streamlines workflows and supports productivity across the 600-plus JPC team, allowing staff to prioritise more business-critical tasks. In terms of efficiency, these technologies prevent the need for manual translation, delivering more accurate translations and processing documents within 5 seconds on average.
As well as productivity and efficiency, they also provide enhanced security and scalability for JPC. As a result, the company ensures compliance with strict data protection requirements. Furthermore, the solutions deployed by Logicalis optimise daily costs and resources, reducing subscription fees and providing AI translation services to more users across the organisation.
Jim McDonnell, IT Manager, JPC, said: “The Logicalis team showcased how to seamlessly introduce AI into real-world applications, transforming complex integrations into straightforward and user-friendly processes. From the start, we were impressed with how the Logicalis team devised a very succinct plan, executed it flawlessly, and collaborated effectively with our team throughout the journey, resulting in a fantastic product and successful implementation. We look forward to engaging on future projects – underpinned by our shared principles of excellence, respect, and teamwork.”
Mairead Malone, Country Leader for Ireland, Logicalis UK& I, added: “John Paul Construction prides itself on going the extra mile and delivering projects of the highest quality. At Logicalis, we share that vision. For JPC, we are delivering technologies to eliminate language barriers, enable seamless communication and enhance collaboration among international teams. We also tailored these to support its overall business objectives, including the continued growth of the organisation. By simplifying the complex, we are helping to drive change and success in the construction sector.”
Galaxy AI is built to understand what you need before you even ask whether that’s suggesting a change in your routine or pulling up just the right information at the right time.
This level of personalisation can be incredibly helpful, but the more your phone knows, the more there is to protect. So, what’s keeping all that personal data secure?
Samsung believes there is no privacy without strong security. That’s why every Galaxy device is protected from the chip up by a multi-layered approach, which includes on-device personalisation, user-controlled cloud processing, and ecosystem-wide protection through Samsung Knox Matrix.
At the core of this system is Samsung Knox Vault, the company’s hardware-based solution for safeguarding your most sensitive information.
Secured at the Hardware Level
Most mobile devices rely solely on software to protect sensitive data. Galaxy devices go further.
Knox Vault is a hardware-level security solution that creates a physical barrier between your most private information and everything else. It works like a locked room inside your phone, with its own processor and memory to encrypt sensitive data, with Knox Vault securing the keys. It pairs a secure processor with dedicated memory, isolating your passwords, PINs, biometrics, as well as financial information and cryptographic keys. These are the kinds of details you don’t want anyone else to access, and Knox Vault is built to make sure they stay private. You don’t need to activate or manage it, as it’s always on, working silently in the background, keeping your data safe while you get on with your day.
This is particularly crucial in the age of AI as user concerns are expanding from traditional cybersecurity threats, like viruses and malware, to worries over leaking personal data, such as conversations with your AI assistant. As AI becomes part of more everyday tasks, the types of data that need protection are also expanding.
For example, metadata from your most personal photos not only details the resolution and file format but also shows the exact location where the image was taken. This personal metadata is more than just files — it’s information that is deeply connected to your daily life, and in the era of AI, these types of data used to provide personalised suggestions need to be kept private.
Knox Vault helps mitigate these growing concerns by safely storing personal information in a secure, hardware-isolated environment designed to block both physical tampering and remote attacks, ensuring your data can’t be accessed without approval.
Personalised AI, Protected at the Core
Knox Vault not only provides protection for today’s threats, but it also ensures your privacy as mobile experiences continue to evolve.
As Galaxy AI becomes more useful, it also becomes more personal, learning how you use your device and adapting to your needs. These highly tailored AI experiences rely on deeply personal data; Knox Vault plays a crucial role in keeping that information private and secured.
Galaxy AI ensures privacy by processing tasks directly on-device where possible, keeping data in your hands and off online servers. For example, Audio Eraser, removes background noise from videos or voice recordings without the need for any cloud-based processing — so your personal information stays private. Call Transcript[2] operates in the same way, keeping your calls organised while ensuring personal conversations stay private by remaining on-device.
Knox Vault ensures your data is protected, confidential, and secure. Building on its role in Galaxy AI as the trusted foundation for security and privacy, Knox Vault will expand across Samsung’s growing AI ecosystem as AI becomes more deeply integrated into the user experience.
Knox Vault is more than a security feature, it’s Galaxy’s promise that no matter how advanced your devices become, or how much AI evolves, your privacy is secured.
To find out more about Galaxy AI and Knox Vault, please visit: Samsung.com/ie
Data drives every layer of effective multimodal AI systems, making it essential for connecting information from text, images, audio, and beyond. These systems thrive on high-quality, well-annotated, and diverse datasets that enable more accurate understanding and integration across multiple data types. With AI-powered computer vision development, visual data can be transformed into actionable insights, broadening the reach and impact of multimodal AI functions.
As businesses look to innovate, the ability of multimodal AI to process varied data inputs is vital for real-world applications. Proper data strategy shapes not only how AI perceives information but also determines the quality and reliability of outputs in practical environments.
Key Takeaways
Data quality and diversity are critical for multimodal AI.
Cross-modal data integration enables sophisticated analysis.
Real-world performance depends on robust data-driven strategies.
The Foundation of Data in Multimodal AI Systems
Data is essential in training multimodal AI, as it allows systems to integrate language, visual, and audible information. By leveraging diverse and high-quality data, these systems can achieve greater accuracy and adaptability in real-world applications such as medical imaging, sentiment analysis, and image captioning.
Types of Data Used in Multimodal AI
Multimodal AI systems utilize a mix of data from different sources. Common data types include images, text, audio, and video. For example, computer vision leverages visual data, while natural language processing handles textual information. Speech recognition and sentiment analysis benefit from both audio and text.
This integration helps machines learn relationships between modalities. In generative AI and deep learning, handling multi-modal data such as audio-visual clips or paired text-image samples is crucial. Large language models often use a combination of structured and unstructured data to enhance their capabilities. Popular multimodal datasets include the Flickr30K and COCO datasets, which offer paired images and captions for robust model training.
Importance of Data Quality and Diversity
Effective multimodal learning depends on both the quality and diversity of the datasets. High-quality data minimizes errors and ambiguities, ensuring that multimodal models perform consistently across tasks like image captioning and medical imaging. Diverse data, including different languages, accents, visual contexts, and environmental noises, supports the model’s resilience and adaptability.
If one data channel is noisy or missing, a multimodal system can rely on another for context. Well-curated, balanced datasets reduce biases and improve reliability in applications such as AI healthcare and generative AI multimodal AI systems are also more robust when drawing from varied and representative sources.
Multimodal Datasets and Benchmarking
Benchmarking multimodal AI requires comprehensive datasets that cover multiple types of input. Widely used resources like the COCO dataset and Flickr30K dataset mix paired images and text, supporting advanced tasks in image captioning and visual question answering.
These multimodal datasets serve as standard benchmarks for comparison across different deep learning models. Organized benchmarking allows researchers to systematically evaluate performance across various AI applications, from sentiment analysis to computer vision, multimodal datasets have been especially valuable for medical imaging tasks and emerging large language models. Regular benchmarking encourages the development of more accurate and generalizable AI systems.
Data-Driven Strategies for Building Effective Multimodal AI Systems
Developing robust multimodal AI systems demands more than just collecting information. Quality, integration methods, learning strategies, and safeguards for privacy and security are fundamental for performance across real-world tasks like recommendation systems, object detection, and diagnosis.
Data Integration and Fusion Techniques
Effective multimodal AI relies on data integration and fusion to combine signals from diverse sources such as text, images, audio, and video. Early fusion merges input data at the raw stage, enabling neural networks like convolutional or recurrent neural networks to learn joint representations. This approach works well for closely related or synchronized data streams.
Late fusion processes each modality separately before merging high-level features, which is key when dealing with weakly correlated or asynchronous data. Stacking and random forests are often used for late fusion in classification tasks. Combining data using these techniques is critical in sectors like healthcare for integrated diagnosis, or in self-driving cars where visual and sensor data must be fused.
Learning Approaches for Multimodal AI
Multimodal AI systems benefit from flexible machine learning strategies tailored to diverse data. Supervised learning remains central, training neural networks such as convolutional and recurrent models on labeled modalities. However, self-supervised and contrastive learning approaches are growing, utilizing unlabeled data to learn robust latent representations. For example, contrastive loss forces systems to associate related data (like matching image and caption pairs), enhancing cross-modal retrieval and recommendation systems.
Probabilistic models can be used to handle uncertainty in input processing, especially when modalities might be noisy or incomplete. Diffusion models, another neural approach, help generate synthetic data to supplement limited training sets, improving object detection and action recognition tasks.
Conclusion
Data is essential for building robust multimodal AI systems. It enables the integration of varied input types—such as text, images, and sensor data—which leads to more capable and context-aware models. Effective use of data allows these systems to learn relationships across different modalities. This results in improved accuracy and adaptability in real-world applications.
Well-curated and diverse datasets are key for ensuring performance and reliability. The quality, completeness, and integration of data sources directly impact how well multimodal AI can function in practical scenarios.
Dell Technologies has today announced that it has teamed up with the Crann Centre, a Cork-based charity, to develop an AI-powered solution that enhances care for children, adults and families living with neuro-physical disabilities. This collaboration has resulted in the development of a bespoke AI-powered intake application that reduces the administrative burden on Crann staff, streamlines intake processes, and enhances organisational efficiencies and service delivery.
The collaboration began as a local volunteering connection, and it has since evolved into a relationship that harnesses the power of AI to support how care is delivered to families living with neuro-physical disabilities. Dell Technologies’ Global Presales team worked closely with Crann to streamline their intake process, reducing processing time by 33%, enhancing data capture, and improving the overall experience for families.
The solution has transformed how Crann performs its client intake appointments, improving the consistency and quality of data captured. The final step, currently in progress, will be the full integration of the app with Crann’s Salesforce system to ensure a single and reliable data source.
With fewer administrative burdens, the Crann team can now devote more time to delivering personalised support, strengthening relationships, and improving care outcomes. This collaboration is a testament to how AI can be used for societal good, increasing Crann’s capacity to serve more families without requiring additional resources.
Speaking about the collaborationDes O’Sullivan, Vice President, Dell Technologies Customer Solution Centres said “At Dell Technologies, we believe innovation truly matters when it drives meaningful change in people’s lives. Through a shared commitment of making a difference, our team at Dell Technologies collaborated with Crann to develop an AI-powered solution keeping in mind Crann’s deep-rooted commitment to family-centred wraparound care.
“With the AI-powered solution that our team helped to create, Crann has increased capacity, allowing them to serve more families than requiring additional resources.The benefits extend far beyond efficiency; Crann team members now have the time and space to focus on deep, meaningful interactions with clients and their families, strengthening emotional and practical support.
“As we look to the future, we’re proud of what has been achieved. Our Dell Technologies team in Ireland and our broader Global Presales team has been at the heart of this journey, bringing Dell’s AI innovation to life in a way that delivers real and lasting value to our community partners.”
The Dell-built solution is designed with future scalability in mind, offering a framework that can be adapted across sectors such as education and customer service.
Crann, which offers wraparound services focused on improving independence and wellbeing, now has increased capacity to deliver support that spans generations underpinned by a shared commitment to personalised care and innovation.
As smart technology and AI become increasingly integrated into our daily lives, new research finds that nearly 9 out of 10 Europeans (88%) express concerns about their privacy.
But, for the majority of consumers, these concerns stop at their phones – despite other smart devices in the home also collecting personal data and being just as vulnerable to threats.
This new research from Samsung Electronics, which surveyed over 8,000 Millennial and Gen Z respondents across Europe, reveals that more than a third (36%) of young Europeans never think about the security of their smart appliances. In contrast, 1 in 2 (49%) think about the privacy of their mobile phones every day — a worrying blind spot given how connected our smart devices have become, and with the number of smart homes in Europe expected to surpass 100 million by 2028.
The research highlights how managing privacy can feel overwhelming for many—with 1 in 5 (20%) of those surveyed saying it’s too complex, and 75% finding it stressful. Within Europe, Spain tops the list of countries where people find managing their privacy the most stressful (88%), followed by Greece (87%) and France (75%) and Italy (75%).
Privacy concerns are wide-ranging—from fears of financial theft (67%) to unease about metadata being used to identify them (56%).
The findings point to a deep education gap when it comes to privacy. Many want to take control of their data when it comes to their technology. In fact, 8 in 10 (80%) put such considerations front and centre at point of purchase, alongside their familiarity with the brand. Yet despite rising awareness, many still feel unequipped to take meaningful action:
Only 13% say they feel ‘very’ knowledgeable about privacy
Just half (48%) are aware of the data their apps and devices collect about them
3 in 10 (31%) accept default settings on apps without fully understanding them
Almost two-thirds (64%) feel they can never be fully in control of their privacy across their devices
Over two-thirds (67%) are calling for better education on data and privacy
Only 1 in 2 (50%) consumers say they trust brands to take the protection of their data seriously
These fears are holding consumers back from harnessing the full potential of the latest technology. Almost 1 in 5 (18%) haven’t shared data between smart devices in the past year due to security fears—showing that Europeans aren’t taking full advantage of connected living, and the enhanced lifestyle benefits it can bring.
Yet just over 6 in 10 (62%) say they would be more open to fully embracing AI and smart-home technology if they better understood the benefits to their lifestyle and felt confident their data was secure (64%). From personalised shopping discounts (17%) to smart home updates (15%) and tailored travel recommendations (15%), many recognise that they are missing out on everyday enhancements due to concerns around how their data is handled.
In light of these findings, Samsung is reinforcing its commitment to empowering users with strong safeguards and transparent privacy controls in an increasingly AI-driven world.
Dr. Seungwon Shin, Corporate EVP & Head of Security Team, Device eXperience Business at Samsung Electronics said: “At Samsung, we believe true innovation starts with people—which is why we put privacy at the core of everything we do. This research highlights a growing trend: while consumers are proactive about managing privacy on their smartphones, they’re often overlooking the broader ecosystem of connected devices. It also reflects a hesitation to fully embrace AI-powered experiences, largely driven by uncertainty around data use.
As advocates for privacy-first design, we’re committed to earning trust through transparency, choice, and built-in protections. Everyone should be able to explore new AI capabilities with confidence, knowing their data is protected and they remain in control. That’s why we’re focused on putting privacy in the hands of users—where it belongs.”
At the heart of this mission is Samsung Knox, the company’s government-grade security platform that safeguards its smart home appliances and Galaxy devices. With the rise of connected living, Samsung is expanding Knox security across its ecosystem so users can enjoy all smart appliances and AI-powered features with peace of mind[3].
Knox Matrix is Samsung’s long-term vision for connected security—where devices in the same ecosystem work together to safeguard one another. Built on private blockchain technology and backed by future-ready protections like post-quantum cryptography, Knox Matrix enables real-time, multi-device protection across smartphones, tablets, TVs, and more.
Complementing this is Knox Vault, which stores sensitive information like PINs, passwords and biometric data in a separate, hardware-based environment. This ensures that even if the main operating system is compromised, private information always stays protected.
With this layered security in place, users are empowered to manage their privacy across all their connected devices, making choices entirely on their own terms.
The research also shows that consumers want more than just promises—they’re seeking practical tools. 29% support clearer data usage policies, and the same amount value privacy information that’s simple and easy to understand.
Recognising this need, Samsung is expanding its efforts to educate users and equip them with tools that put privacy into their hands across all Samsung devices. For more information on privacy, see here. For more information on the Samsung Knox security principles, see here.
