Michael UngarCEO of the Industry 4.0 Club

The Limitations of AI: Why Human Touch Matters

18 Nov 2025 · 1 min

Watch on YouTube

0:56

In this monologue, the speaker highlights three core limitations of current AI systems. First, the data used to train AI models is largely unfiltered, leading to persistent biases that, despite mitigation efforts, remain present and require conscious awareness. Second, AI should be viewed as an augmentative tool rather than a human replacement; the speaker likens AI to CAD software, which accelerates work but does not guarantee quality design. Finally, the speaker argues that the final creative spark—the "last 5%" that distinguishes mediocre from excellent music, art, or food—remains uniquely human, emphasizing that true excellence still depends on human judgment and creativity.

Chapters

So the kinds of content that are put in there, you know, haven't been filtered. And a lot of the work that went into making sure that it was usable was to try to take out some of these biases, but it's not quite there yet. So that's something to be conscious of. And it's not a replacement for humans. And I think, you know, there's a— and this is something that everybody's concerned with. And I started my career working as a mechanical designer, and I use CAD software, which in some ways— enabled me to be way faster than I was without using CAD software, but that still doesn't make it, you know, like I had to make it a good design or bad design. So in some ways, in its current state, it's a tool, but to actually do something good, and if you think about, you know, the difference between crappy music and good music, you know, crappy art and good art, crappy food and good food, like that magic that makes something excellent, that last little 5% is something that, you know, still humans are the only ones that can do.