DeepSeek’s AI Breakthrough: The Democratisation of Artificial Intelligence

Author(s): Kapardhi kannekanti

Originally published on Towards AI.

A Chinese Company called Deepseek recently dropped a bombshell in the AI world with the new model called R1. Tech Experts are calling it AI’s “Sputnik moment”, a game-changer that could shake up how we build and use AI. Here’s the full story in plain English.

The Secret Sauce: Smart Memory Usage

Think of AI like a brain that needs memory to think. Most AI companies use a system that takes up lots of memory (32-bit), like writing everything in a really detailed way. DeepSeek did something clever — they used a lighter system (8-bit or FP8) that’s like writing in shorthand. Sure, you lose some details, but it works almost just as well. What’s the big deal? Well, DeepSeek trained their AI for just $5.6 million in three months. Compare that to U.S. companies spending billions — we’re talking about paying for a bicycle instead of a fleet of Ferraris.

Dirt-Cheap AI Access

Here’s where it gets interesting. DeepSeek is offering their API for peanuts — about 3% of what OpenAI charges for ChatGPT. Even when they raise prices, it’ll still be 75% cheaper than the competition. One developer ran 200,000 tests for just 50 cents — that’s cheaper than a cup of coffee.

Power Without the Price Tag

The cool part? This cheaper AI isn’t worse — it’s keeping up with the big boys like OpenAI and Google’s latest models. Even better, you don’t need a supercomputer to run it. While other AIs need massive data centers, DeepSeek’s model can run on a decent gaming PC, and smaller versions work on phones.

Big Tech’s Reaction: Mixed Feelings

The tech giants are scrambling to figure out what this means:

• Microsoft’s cloud business could make bank offering cheaper AI
• Amazon Web Services (AWS) could host these efficient models
• Apple’s M-series chips are perfect for running this kind of AI
• Google might be sweating since their massive computing advantage matters less now

Open Source vs. Proprietary

Meta’s AI chief dropped some truth: this isn’t just about China vs. USA — it’s about open source (shared code) vs. closed source (secret code). DeepSeek built on public research and shared their work back. It’s like they used everyone’s recipes to make a better dish, then shared their improved recipe with the world.

What This Means for the Future

The cost of high-level AI has crashed — we’re talking a 1000x price drop in 18 months. That’s like a $30,000 car suddenly costing $30. When tools get this cheap, crazy things happen.

Remember when cloud storage was expensive and we counted every megabyte? Now we upload countless photos without thinking. The same thing’s happening with AI — as it gets cheaper, we’ll use it everywhere.

More Efficiency = More Usage

There’s a funny effect in tech called “Jevons Paradox” — when something gets more efficient and cheaper, people don’t use less of it — they use way more. It’s like when Netflix made watching movies super easy — people didn’t watch less, they binged more than ever.

The Bottom Line

DeepSeek just kicked down the door to the AI club. Before, only tech giants could play with the best AI. Now? Any developer with a decent computer can join in. It’s like moving from expensive mainframe computers to personal PCs — suddenly everyone can innovate.

The real winner here isn’t any single company — it’s everyone who wants to use AI. The technology is becoming a utility, like electricity or internet access. We’re not just talking about cheaper versions of what we already have — we’re talking about entirely new possibilities.

As costs drop and access grows, the next big AI breakthrough might come from a college student’s laptop rather than a tech giant’s data center. The AI race isn’t about who has the biggest computer anymore — it’s about who can use these tools in the smartest way.

Find the repository here!

Passionate about AI, ML, and tech? Let’s connect and collaborate!
Twitter. LinkedIn. GitHub. [email protected].

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI