DeepSeek v Silicon Valley: A David v Goliath story?

DeepSeek: The New Star in the AI Galaxy

Imagine, if you will, a quiet ripple in the vast ocean of artificial intelligence, slowly building momentum until it becomes a wave that has the tech world sitting up and taking notice.

This is the story of DeepSeek, a Chinese AI startup that has made quite the splash since its inception in 2023. Founded by Liang Wenfeng — an individual who deftly juggles the realms of hedge funds and AI — DeepSeek is not your run-of-the-mill tech endeavour.

Funded solely by High Flyer, a hedge fund also led by Wenfeng, DeepSeek operates with an enviable degree of independence. This unique setup allows the company to focus on long-term research without the constant hum of investor demands in the background.

DeepSeek first caught the public's eye with the release of DeepSeek Coder in November 2023, an open-source model tailored for coding tasks. But it was the arrival of DeepSeek LLM, a 67-billion-parameter behemoth, that really set the cat among the pigeons in the world of large language models.

However, the real game-changer was DeepSeek-V2, launched in May 2024. With its strong performance and budget-friendly pricing, this model sparked a price war among Chinese tech giants like ByteDance, Tencent, Baidu, and Alibaba, forcing them to rethink their pricing strategies. This was no gentle ripple; this was a tidal wave of change.

The Great Tech Stock Tumble

Now, let's take a moment to travel across the ocean to the bustling world of US tech stocks. On January 27, 2025, a day that will likely be remembered by many a stock analyst, US markets took quite the tumble.

More than $1 trillion in market cap vanished into thin air, with the S&P 500 dipping nearly 1.5%, and the tech-heavy Nasdaq Composite shedding over 3%. The hardest hit were the chip firms, with Nvidia alone suffering a 17% plunge—an eye-watering $589 billion loss in market cap, marking the worst single-day loss in history.

What caused this financial tempest, you ask? It was none other than DeepSeek's latest AI model outperforming OpenAI's ChatGPT in several tests, all while being developed at a fraction of the cost. This model, running on modest hardware, is 20 to 40 times cheaper than its counterparts from OpenAI.

The sheer cost-effectiveness of DeepSeek’s model threw the necessity of the current pace of capital expenditure and technology upgrades by US hyperscalers into question.

The Secret Sauce: How DeepSeek Did It

So, how did DeepSeek achieve what seemed impossible? The answer lies in a clever restructuring of AI model foundations, focusing on software-driven resource optimization rather than relying solely on hardware.

Despite facing US export controls that limited access to the latest chips, DeepSeek employed a series of ingenious engineering tweaks. These included custom communication schemes between chips to improve data transfer, memory-saving techniques, and reinforcement learning methods to reduce computational power needs.

The result?

A significant reduction in costs compared to traditional large language models. For example, training one of DeepSeek's latest models cost a mere $5.6 million—a stark contrast to the $100 million to $1 billion usually estimated for such feats.

This cost efficiency is mirrored in their API pricing for DeepSeek-R1, which is a fraction of OpenAI's rates. With input tokens priced at $0.55 per million and output tokens at $2.19 per million, DeepSeek is undercutting OpenAI’s API rates of $15 and $60, respectively.

What This Means for Background Checks

Now, let's bring this back to a topic near and dear to us at Verify360: the world of background checks. With more advancements and other AI models replicating the process to drive the costs down, the implications AI's advancements are as profound as they are exciting.

1. Lightning-Fast Screenings:

Traditionally, background checks could be an arduous affair, often taking days or even weeks. But AI-driven platforms promise to shrink this timeline dramatically, conducting checks in mere hours by using AI to comb through public records and criminal databases with the precision of a seasoned detective.

2. Sharper Fraud Detection:

We've all seen those resumes that read more like works of fiction. With AI, these fabrications can be sniffed out with remarkable accuracy. Machine learning and blockchain technology work in tandem to verify data in real-time, unearthing discrepancies before they become costly mistakes.

3. Ongoing Vigilance:

By 2025, AI will enable continuous background checks, keeping employers informed of any significant changes in an employee's background post-hiring. This means staying ahead of potential legal issues or changes in licensure, ensuring that the workforce remains compliant and trustworthy.

4. Bias-Free Evaluations:

Bias has long plagued traditional background screening methods, but AI offers a path to fairness. By evaluating candidates based on objective criteria, AI helps to level the playing field, providing an equal chance for all.

Conclusion: The Road Ahead

As AI continues to shape the future of hiring and background checks, both employers and job seekers must be ready to embrace these changes. Companies that invest in AI-powered tools will find themselves with a competitive edge, attracting top talent with newfound efficiency.

By 2030, AI technology will not only streamline processes but also enhance accuracy and fairness in hiring.

So, as we stand on the brink of this new era, let us be prepared to ride the wave of innovation that AI promise. The future is bright, and the possibilities are endless.