The Deepseek Wave

The Deepseek Wave: Who’s under risk actually? In this post, we are gonna look at the first TSUNAMI of an Innovation that happened in the AI field.

Disclaimer: This is not an endorsement of Deepseek family of LLMs but rather an information of an important innovation in the AI field.

In this site, we often talk about LLMs like “The Amazing ChatGPT”, Phi family of LLMs, LLAMA. I am an ardent user of ChatGPT. I used to be a premium user. But, as I got another competitor LLM (though I hate it) through our company, I stopped using the ChatGPT premium subscription .

But, ChatGPT is the best LLM available right now. It is very much like ourselves but only as our own personal assistant, who can do most of our average knowledge tasks, available 24x7, and understand us the best with all our conversational history as another source of data, ChatGPT is unstoppable. Could you imagine yourself as your assistant? How easy life would be if our assistant is us? An intelligent-being putting our values first and prioritises us. We are not there quite yet, we will be there soon.

Now, suddenly, a wave happened, as big as a TSUNAMI, in the name of DEEPSEEK. It crashed around 830 billion dollars from the American Stock Exchange. Deepseek family of LLMs which is a Chinese product has shaken the AI field and rightly so. AI Industry is the current hot field. It’s where the action is happening and will happen for a long time. But, just one problem. LLM Training took 5 Billions to make something as amazing as ChatGPT o3. Deepseek changed it by getting trained in 5 Millions (actually 6, but let’s keep 5 for simplicity), that’s around 1/1000 of 5 Billions. That’s a big innovation. Deepseek has shown that even above average LLMs don’t have to be too costly. Then, imagine the below average ones, or, the low efficient open ones (cough-cough-LLAMA), who would use them.

Now, let’s look at the ones that have lost their place to DEEPSEEK. It’s not OpenAI. They are for profit and most importantly, they’re closed source. They had become for-profit and as true capitalists they will find exploitative ways to stay profit. The current real losers, are the open weight ones like LLAMA. They championed being open source, while just being open weight. They were the flag bearers for open LLMs, AI teams from Meta came to India for canvassing too. Open weight ones cannot afford to be very costly. But LLAMA was. So, DEEPSEEK has dethroned LLAMA right now and taken its rightful top place in the open weight LLMs space.

So, now all the players are trying to bring the training cost down. That’s good for the average user of LLM. But, the demand for hardware for training will be there. They will just add more features on the existing LLMs and the make the training cost worth it.

The current wave is to make an ethical and moral AI. And India, the birthplace of morals and ethics, with her rich literature on the subject, can actually surf the current wave.

Could India finally become a Global Spiritual Leader?

What do you think? Let me know your thoughts in the comments below!