DeepSeek: All you need to know about the AI gamechanger

The internet has gone crazy over a new chatbot — DeepSeek.

If you have glanced at tech headlines and social media lately, you might have noticed this new name buzzing through the chatter.

This week, Chinese AI chatbot DeepSeek did something out of thin air: It slipped past ChatGPT, TikTok, and Instagram to claim the top spot on Apple's App Store. It is the brainchild of a Beijing-based lab you have probably never heard of, until now.

Seemingly overnight, its sudden fame has tech circles and Wall Street asking two questions: How did this happen? And could it shake America's dominance on the AI race?

What is DeepSeek, anyway?

Well, you already know ChatGPT and what it does? DeepSeek can do most of them, and better in most cases.

Built on what its creators call a "reasoning-focused" AI model, it is designed to handle complex queries with startling speed. Think quick summaries of dense reports, real-time data analysis, or even troubleshooting tech glitches.

DeepSeek had to achieve all these edges over its competitors with some low-tier GPUs as graphics card makers are restricted from exporting higher-tier products to China.

Unlike older chatbots that often sound like they are reading from a textbook, DeepSeek focuses on truly understanding the conversation. Early users praise its ability to remember things, like your preference for bullet points over paragraphs, and adjust its tone depending on whether it is chatting with a client or a friend.

It is also multilingual, supporting over a dozen languages, which is what makes it talk of the — globe! Not to mention, DeepSeek is open-source and you can run it on your own computer — in case you are geeky enough.

DeepSeek vs. ChatGPT: The chatbot showdown

Let us address the elephant in the room: is not this just another ChatGPT clone? Not really. While both tools fall under the "AI chatbot" umbrella, they are built for different crowds.

ChatGPT, developed by OpenAI, is the jack-of-all-trades — great for brainstorming ideas, drafting essays, or even generating poetry. But ask it to analyse live stock market trends or debug a chunk of code, and it might stumble.

DeepSeek, meanwhile, seems to be gunning for the "specialist" label. According to the founders, its training data includes vast technical manuals, financial reports, and scientific journals, making it a hit with professionals in fields like engineering, finance and healthcare.

In my test, I gave a calculus problem from a book to both chatbots. Both provided similar answers, but DeepSeek's explanation was more detailed and it took less time than ChatGPT's response.

In my second attempt, I gave both bots a document to derive insights from it, again, both of them provided similar answers — with DeepSeek being a bit faster and ChatGPT — more conversational.

According to Geekforgeek's comparison, DeepSeek surpasses ChatGPT in coding related tasks as well.

DeepSeek's biggest advantage is its completely free access with no query limits. In contrast, while ChatGPT has a free version, advanced features require payment. DeepSeek's APIs are also much more cost-effective, making them an appealing choice for developers and businesses. 

For example, DeepSeek charges just $0.55 per million input tokens and $2.19 per million output tokens — significantly lower than OpenAI's rates of $15 and $60 for the same services.

DeepSeek had to achieve all these edges over its competitors with some low-tier GPUs — as GPU makers are restricted from exporting higher-tier GPUs to China. Given the efficiency DeepSeek has achieved, this has raised the eyebrows of the Silicon Valley bosses, some claiming that the company acquired high-end GPUs illegally, and a thorough investigation is on its way.

However, DeepSeek claims it legally acquired Nvidia's H800 chips in 2023 and also possesses the H20s, which are still allowed in China. This essentially means they had to achieve an unforeseen level of power efficiency.

Wall Street stunned

Here is where things get juicy. DeepSeek's rise has not just captivated tech enthusiasts, the Chinese AI firm has sent shockwaves through the entire tech industry.

This caused a massive selloff that wiped $600 billion from Nvidia's market value on a single day, and in total the number is nearly a trillion, so far. The Nasdaq index, heavily weighted with tech stocks, dropped by 3%, and some of the world's richest men saw their fortunes shrink overnight.

Jensen Huang, Nvidia's co-founder and CEO, was among the hardest hit. His personal wealth plunged by $20.7 billion, bringing his net worth down to $103.7 billion. The loss pushed him from 10th to 17th place in global billionaire rankings.

Oracle chair Larry Ellison lost even more — $27.6 billion — as Oracle's stock tumbled by 14%, knocking him from third to fifth place among the world's richest. In total, tech billionaires saw almost a hundred billion disappear in a single day.

Some experts say the reaction is overblown. They believe AI development still depends on the best hardware, and DeepSeek's rise will not change that. But with chip export controls in place, China's AI ambitions may still face hurdles ahead.

One thing is clear at this point: the AI race just got a new frontrunner. Whether it will stay ahead or face challenges is yet to be seen. As for the rest of us, maybe it is time to ask DeepSeek for some stock market tips.