Artificial Intelligence is everywhere these days. It powers your phone’s autocomplete, curates your newsfeed, and probably recommends the next show you’ll binge watch. But where is AI really heading and what’s just hype?
Nedjalko Milenkov, Senior Engineering Manager at Scalefocus, tackled this head-on in his talk “AI-Focused Research and Software Development”. With nearly two decades in IT (and plenty of hands-on AI work), he pulled back the curtain on what AI can (and can’t) do right now.
Spoiler: it’s not quite Skynet… yet.
From Lab Curiosity to Everyday Tool
A decade ago, experimenting with neural networks required patience, serious math skills, and weeks of setup. Today, you can spin up a large language model on your laptop in just a couple of hours.
This leap came down to two breakthroughs:
- Hardware finally caught up. Modern GPUs made the heavy lifting feasible.
- The transformer architecture (“Attention Is All You Need”) unlocked the power behind today’s ChatGPT and open-source rivals.
But let’s not romanticize it – these models aren’t flawless super-brains. As Nedjalko quipped, they’re like someone trying to recall War and Peace: they can summarize the last few chapters in detail, but the earlier parts? Already forgotten.
Beyond ChatGPT: Where AI Is Headed
When people think about AI, chatbots usually come to mind. But that’s just one corner of the field. Nedjalko highlighted several areas where AI is already reshaping industries:
- Generative AI – not just text, but images, video, and even music.
- AI for security – building safer systems and detecting cyber threats.
- Advanced analytics – spotting patterns humans would miss.
- Custom AI models – tuned for specific businesses, not just generic tasks.
- MLOps & AIOps – the less glamorous, but essential, infrastructure that keeps AI running at scale.
In short: if your idea of AI stops at chatbots, you’re only scratching the surface.
Myths Busted
One of the most engaging parts of Nedjalko’s talk was his “Myths and Legends” segment.
A few crowd favorites:
“You need a supercomputer to train AI.”
Only if you’re starting from scratch. Fine-tuning existing models can be done in hours on a few GPUs.
“AI must run in the cloud.”
Not true. Many state-of-the-art models can run locally, giving you privacy and control.
“Bigger is always better.”
Small, fine-tuned models often outperform giant ones for specific use cases.
“It’s all about text LLMs.”
Wrong again. Speech-to-text, video analysis, and image recognition are exploding.
Open Source vs. Closed Models: The Big Trade-Off
Another hot topic: should you rely on closed commercial APIs like ChatGPT, or embrace open source?
- Closed models are polished and fast to adopt, but you’re locked into their ecosystem. One API change can break your software overnight.
- Open source models give you control, transparency, and data sovereignty. That’s why banks, law firms, and anyone handling sensitive information increasingly favor them.
The Road Ahead: Exciting but Messy
So, where does this leave us? According to Nedjalko, we’re still far from “true” artificial general intelligence. Today’s models are powerful, but ultimately they’re pattern machines with short attention spans.
That said, the pace of progress is dizzying. New frameworks, libraries, and models appear every few weeks – many of them open source. For developers and businesses, this means endless opportunities to experiment, innovate, and build responsibly.
The real trick? Don’t chase the flashiest, biggest models. Pick the right tool for your problem and always pay attention to where your data goes.
Final Takeaway
AI isn’t magic, but it’s no longer science fiction either. It’s a fast-moving field full of promise and…well, pitfalls. Whether you’re building a chatbot, automating document review, or experimenting with video generation, the takeaway is simple:
- Stay curious.
- Experiment often.
- And don’t believe every AI myth you hear.
Because ultimately, the future of AI won’t be defined by bigger models or cooler demos, but by how thoughtfully we put these tools to work.
Want to go deeper into AI myths and realities?
Check out our Tech Savvy Talks podcast episode “AI Myths and Legends”, where Nedjalko Milenkov joins Simeon Petkov to bust misconceptions – from “you always need massive data” to “AI is just a black box.”
It’s a practical, no-nonsense look at what AI can (and can’t) do, with lessons from systems that have been built, broken, and improved.
