Skip to main content
  1. Home
  2. Computing
  3. Emerging Tech
  4. News

Digital Trends may earn a commission when you buy through links on our site. Why trust us?

DeepSeek readies the next AI disruption with self-improving models

Add as a preferred source on Google
DeepSeek AI chatbot running on an iPhone.
Nadeem Sarwar / Digital Trends

Barely a few months ago, Wall Street’s big bet on generative AI had a moment of reckoning when DeepSeek arrived on the scene. Despite its heavily censored nature, the open source DeepSeek proved that a frontier reasoning AI model doesn’t necessarily require billions of dollars and can be pulled off on modest resources.

It quickly found commercial adoption by giants such as Huawei, Oppo, and Vivo, while the likes of Microsoft, Alibaba, and Tencent quickly gave it a spot on their platforms. Now, the buzzy Chinese company’s next target is self-improving AI models that use a looping judge-reward approach to improve themselves.

Recommended Videos

In a pre-print paper (via Bloomberg), researchers at DeepSeek and China’s Tsinghua University describe a new approach that could make AI models more intelligent and efficient in a self-improving fashion. The underlying tech is called self-principled critique tuning (SPCT), and the approach is technically known as generative reward modeling (GRM). 

In the simplest of terms, it is somewhat like creating a feedback loop in real-time. An AI model is fundamentally improved by scaling up the model’s size during training. That takes a lot of human work and computing resources. DeepSeek is proposing a system where the underlying “judge” comes with its own set of critiques and principles for an AI model as it prepares an answer to user queries. 

This set of critiques and principles is then compared against the static rules set at the heart of an AI model and the desired outcome. If there is a high degree of match, a reward signal is generated, which effectively guides the AI to perform even better in the next cycle. 

The experts behind the paper are referring to the next generation of self-improving AI models as DeepSeek-GRM. Benchmarks listed in the paper suggest that these models perform better than Google’s Gemini, Meta’s Llama, and OpenAI’s GPT-4o models. DeepSeek says these next-gen AI models will be released via the open-source channel. 

Self-improving AI?

The topic of AI that can improve itself has drawn some ambitious and controversial remarks. Former Google CEO, Eric Schmidt, argued that we might need a kill switch for such systems. “When the system can self-improve, we need to seriously think about unplugging it,” Schmidt was quoted as saying by Fortune.

The concept of a recursively self-improving AI is not exactly a novel concept. The idea of an ultra-intelligent machine, which is subsequently capable of making even better machines, actually traces all the way back to mathematician I.J. Good back in 1965. In 2007, AI expert Eliezer Yudkowsky hypothesized about Seed AI, an AI “designed for self-understanding, self-modification, and recursive self-improvement.”

In 2024, Japan’s Sakana AI detailed the concept of an “AI Scientist” about a system capable of passing the whole pipeline of a research paper from beginning to end. In a research paper published in March this year, Meta’s experts revealed self-rewarding language models where the AI itself acts as a judge to provide rewards during training.

Microsoft CEO Satya Nadella says AI development is being optimized by OpenAI’s o1 model and has entered a recursive phase: “we are using AI to build AI tools to build better AI” pic.twitter.com/IHuFIpQl2C

— Tsarathustra (@tsarnick) October 21, 2024

Meta’s internal tests on its Llama 2 AI model using the novel self-rewarding technique saw it outperform rivals such as Anthropic’s Claude 2, Google’s Gemini Pro, and OpenAI’s GPT-4 models. Amazon-backed Anthropic detailed what they called reward-tampering, an unexpected process “where a model directly modifies its own reward mechanism.”

Google is not too far behind on the idea. In a study published in the Nature journal earlier this month, experts at Google DeepMind showcased an AI algorithm called Dreamer that can self-improve, using the Minecraft game as an exercise example. 

Experts at IBM are working on their own approach called deductive closure training, where an AI model uses its own responses and evaluates them against the training data to improve itself. The whole premise, however, isn’t all sunshine and rainbows.

Research suggests that when AI models try to train themselves on self-generated synthetic data, it leads to defects colloquially known as “model collapse.” It would be interesting to see just how DeepSeek executes the idea, and whether it can do it in a more frugal fashion than its rivals from the West. 

Nadeem Sarwar
Nadeem is the Managing Editor at Digital Trends.
After iPad Pro and MacBook Pro, the iMac could be the next in line for an OLED screen upgrade
iMac with M4

The iPhone got an OLED panel in 2017, while the iPad Pro followed in 2024. Even the MacBook Pro is expected to follow later this year or early next year. But what about the iMac?

According to TrendForce, the iMac could get an OLED upgrade. There's no timeline yet, but the direction is clear. Apple wants to replace its current display technologies with OLED, raising the bar for color quality for both regular users and professionals.

Read more
This $1,299 gaming PC wants to be a Steam Machine without waiting for Valve
Valve’s Steam Machine dream is already real in MetaPC's new prebuilt
MetaPC's Steamroller is a new Steam Machine rival

Valve’s Steam Machine may be the face of SteamOS, but the platform isn't exclusive to it. A big announcement after Steam Machine's unveiling was that SteamOS would be arriving on systems outside of the new hybrid console. Now, MetaPCs is one of the first to take advantage of this by opening the preorders for the Steamroller, a new prebuilt gaming desktop that ships with SteamOS installed by default.

Though Steamroller is not trying to be a tiny console-like cube. It is a normal desktop PC with standard parts and a real upgrade path. The system costs $1,299 and is listed with a preorder date of July 3, 2026.

Read more
This cheap Steam Machine clone sounds too good to be true because it probably is
A Chinese Steam Machine clone claims impossible hardware at an impossible price
A Chinese rip-off of the Steam Machine

Valve’s new Steam Machine has already caused plenty of sticker shock. So it's no surprise that a flood of cheaper alternatives is hitting the online market. Valve is currently charging over $1,000 for its tiny-living-room SteamOS PC, and of course, people are trying to offer the same feel for less money,

One listing from China is a great example, but it looks a little too suspicious. According to VideoCardz, a Steam Machine-style mini PC listing shared on Reddit claims to offer a compact SteamOS system with a 2TB SSD, AMD Ryzen 5 5500 processor, Radeon RX 6750 GRE 10GB graphics, 16GB of DDR5 memory, and a price of 4,680 RMB, or roughly $688. This sounds incredible... if it were true.

Read more