At Media Distillery, we live and breathe AI. From enhancing content discovery to improving media workflows, we constantly explore how AI can shape the future. But where is AI really headed? To get a clearer picture, we sat down with two of our own machine learning wizards, Ward Pennink and Redouane Dahmani, to discuss the state of AI today - sharing insights, a touch of humor, and some thought-provoking predictions. Here’s a sneak peek into their world, where cutting-edge innovation meets a healthy dose of reality.
“ChatGPT brought Generative AI into the public consciousness, but GenAI has now moved from text-to-text models to what we call ‘any-to-any’ models—systems that can handle multiple types of input, whether it’s text, video, images, or audio,” Ward explained. Previously, AI models were limited to handling a single type of input, requiring multiple systems to achieve tasks like video understanding. The latest generation of Gen AI models can process and generate across multiple modalities, opening up a vast new landscape of possibilities. For some years now, here at Media Distillery, we’ve been developing our own solutions to do this kind of multi-modal analysis of video. But with Generative AI we’re able to go much further, much faster. And analysis that was previously out of reach (such as generating real-time video previews) are now possible with GenAI advances.
But it’s not just about more advanced models; it’s about what they enable. As Ward put it, “One future scenario is AI being fully integrated into most everyday objects. Some examples of this already exist, such as AI-enhanced glasses that observe alongside the wearer, allowing them to look at something, ask a question, and receive an answer instantly.” Utopian or dystopian? That depends on your perspective. Either way, the possibilities are vast—especially for industries like media, where AI could completely change how content is created, personalized, and consumed.
Take Google DeepMind’s Gemini, for example. It can analyze video in real time, answering questions about what’s happening on screen. This advancement lays the groundwork for AI systems that truly understand the world around us.
Ward and Redouane are quick to remind us that while AI can be revolutionary, it also has its quirks—and that’s part of the fun. “One of my favourite examples,” Ward shared, “is AI generating a biblical verse about peanut butter in the style of the King James Bible.” AI’s creative applications have made it a viral sensation.
These playful use cases highlight how accessible AI has become. What started as a tool for tech experts is now part of everyday creativity, helping people write songs, learn languages, and even create niche humor. “It’s gimmicky, but it’s also genuinely useful,” Ward added, referencing how tools like ChatGPT can help users practice Spanish or brainstorm ideas.
Of course, the shift from novelty to real-world application is happening fast. Its true power lies in boosting productivity. Redouane noted, “Most of our developers now use AI assistants, and the difference is huge. We can write more code in less time, and the code is often higher quality because AI helps catch errors and suggest improvements.” He added, “This shift from information retrieval to instant answers is a game-changer. But it also comes with risks—if the AI provides incorrect information, users might not question it.”
As AI becomes more powerful, concerns about its impact are growing. Training large models requires immense amounts of energy, resulting in significant environmental costs. "Big companies love to end their blog posts with a sentence about sustainability," Ward noted, "but it’s often just greenwashing." The demand for computing power is increasing rapidly, but scaling indefinitely isn’t sustainable. Ward believes the industry must prioritize developing more energy-efficient models.
It’s not just the environmental impact that raises concerns. The dominance of big tech companies in AI development often means smaller players get swallowed up, limiting opportunities for open-source innovation. But there’s hope: Ward pointed out that open-source models are “closing the gap” and could democratize access to AI, putting power back into the hands of smaller companies and individuals.
One of the most significant recent examples is DeepSeek. While much of the hype around it has been about matching the performance of OpenAI’s GPT-4 (o1 model), what sets it apart is its efficiency and accessibility. “For some time, AI development has been an arms race for compute, where having the most powerful hardware often meant staying ahead,” Ward explained. “Scaling compute was seen as the best and most straightforward way to drive progress, leading to many of the biggest breakthroughs. While that still holds true in some ways, Deepseek has challenged the idea that more compute is the only path forward, proving that models can reach the same level of intelligence at a fraction of the cost.”
Beyond efficiency, DeepSeek has also pushed open-source AI forward. “Over the past year, we've seen open-source models stepping closer towards closed-source performance, with examples like Meta’s Llama 3 and Google’s Gemma,” Redouane noted. “But none of them has been as 'revolutionary' as the DeepSeek model family, which, in a short amount of time, has started competing at the top of the leaderboard.”
While debate continues over whether DeepSeek is truly open-source - since it provides model outputs but not the full code or training data - it still represents a major shift toward more accessible AI development that doesn't come at unsustainable costs.
So, what’s next for AI? According to Ward and Redouane, 2025 will be the year AI moves beyond providing intelligence to taking action. AI agents - systems capable of autonomously completing tasks - represent this next leap. Unlike traditional AI, which primarily assists by offering information, these agents will be able to plan, make decisions, and execute complex actions on behalf of users, whether in digital environments or, eventually, the physical world via robots.
“Imagine asking an AI to book a flight to Paris,” Ward said. “It’ll search for the best options, compare prices, and handle the booking process for you.” While we’re not quite there yet (and giving an AI your credit card might still feel risky), the groundwork is being laid. These agents will likely start by automating routine tasks in the digital space, such as organizing files or managing databases, before moving into more complex territory.
At Media Distillery, the team is keeping a close eye on these developments. “We’re not using agents in our products yet—it’s still too early—but we’re experimenting with ways to incorporate their capabilities in the future,” Redouane explained. For now, AI assistants are giving our developers a massive productivity boost.
AI’s potential is undeniable, but as Ward and Redouane emphasized, innovation must be balanced with responsibility. The environmental impact of AI, the need for ethical guidelines, and the dominance of big tech are challenges that can’t be ignored. Ward pointed to the EU AI Act as a step in the right direction, although regulations often lag behind technological advancements. “The challenge is finding a balance between fostering innovation and ensuring AI is used responsibly,” he said.
Open-source models, like DeepSeek, play an important role in this transformation. While big tech has driven AI breakthroughs, the rise of open-source alternatives is helping to level the playing field. “Endeavors using open-source models to at least stay up to speed with closed-source models are an important trend to continue,” Redouane noted. “This will diminish the dominance of big tech and ultimately benefit end users the most. So, I hope to see more developments in that sector in 2025!”
At Media Distillery, we’re committed to developing AI that benefits both businesses and consumers. Whether through Topic Distillery’s intelligent metadata generation or Preview Distillery’s AI-driven video previews, our focus is on practical solutions that improve the media experience without sacrificing reliability or transparency.
Ultimately, the future of AI isn’t just about what machines can do—it’s about what we, as humans, choose to do with them. So, what’s next for AI? We’ll let Ward have the final word: “The next phase is about turning information into action. AI agents are coming—and they’re going to change everything.”
-
If you’d like to learn more about AI—and how Media Distillery harnesses its power to transform content discovery and media workflows—get in touch today. We’d love to show you how our AI-powered solutions can help your platform unlock new opportunities.
March 11, 2025
