Tech conferences today feel like Groundhog Day. Every keynote starts with dramatic music, a slide deck with a buzzword explosion (“AI-powered deep tech innovation”), and a founder claiming their Deep tech startups will ‘revolutionize the world.’ But scratch beneath the surface, and you’ll find the revolution is, more often than not, a fancy wrapper around someone else’s hard work—specifically, generative AI frameworks like ChatGPT, Claude, or Gemini.
What Exactly Is “Deep Tech”?
Once upon a time, deep tech meant breakthroughs in quantum computing, nanotechnology, or biotech. It was a realm for visionaries who spent years tinkering in labs, pursuing bleeding-edge science. Now? Deep tech has apparently been redefined to mean “customizing OpenAI’s API and slapping a shiny new interface on it.”
Here’s the drill:
- Pick a Framework: ChatGPT, Claude, Gemini, or one of the open-source models.
- Rebrand It: Call it a “personalized conversational agent for X industry.” Bonus points if you describe it as “disruptive.”
- Profit: Or at least try to, while burning VC funds faster than a server running an unoptimized neural net.
Generative AI: The Not-So-Secret Sauce.
To be fair, frameworks like ChatGPT and Gemini are marvels of engineering, representing years of research and billions of dollars in investment. But the irony is that while these tools embody deep tech, many startups using them… don’t.
Copy-Paste Innovation: The tech stack of many “deep tech” startups begins and ends with integrating a pre-trained AI model into a user-facing app.
Applications Masquerading as Inventions: Whether it’s a “healthcare AI assistant” or a “writing enhancement tool,” most are just thinly veiled applications of generative AI. Calling it deep tech is like calling a WordPress blog a “bespoke publishing platform.”
But Wait, There’s More… Sort Of.
The problem isn’t the use of generative AI frameworks—it’s the pretence. Startups could simply say, “We’re leveraging the power of generative AI to create value in X.” Instead, they claim to have cracked the AI code themselves, implying they’ve built something revolutionary from scratch.
Take a closer look, and you might find:
No Proprietary Models: Few startups are developing their own large language models (LLMs). Training these models requires massive datasets, computational power, and research teams—resources that are beyond most startups.
Minimal Differentiation: The product offerings often overlap to the point where you can swap one for another without noticing.
VCs, Are You Paying Attention?
Investors are complicit in this ecosystem of “deep tricks.” Many are dazzled by AI pitches without digging deeper. As long as the deck mentions GPT-4 and throws in a chart showing exponential growth (with no context), it’s money time. This creates a cycle where startups focus more on marketing buzzwords than building genuinely innovative technology. “AI-powered” is the new “cloud-enabled” of the 2010s—a term that’s broad enough to mean everything and nothing.
A (Gentle) Wake-Up Call for Startups.
If you’re going to call yourself a deep tech startup, here are some suggestions: Do Something Core: Develop a unique algorithm, improve model efficiency, or create your own training dataset. Drop the Facade: If you’re using OpenAI’s API, own it. Innovation can still happen in the application layer, but pretending you’ve reinvented AI is just embarrassing. Focus on Real Problems: AI doesn’t need another app to write emails for people. Solve challenges in healthcare, energy, or education.
Conclusion.
Generative AI has democratized access to powerful technology, and that’s a good thing. But deep tech startups need to embrace a little humility. Building a user-friendly wrapper around ChatGPT isn’t revolutionary; it’s product development. As for the term “deep tech,” perhaps it’s time to reclaim its meaning. Until then, let’s call today’s “deep tech” startups what they really are: applications of deep tech, not the creators of it. And that’s fine—but let’s not confuse the two.
“Deep trick or deep tech? You decide.“
