DeepSeek's Pro-China Bias is Superficial: Revealing the Power of Local AI Deployment

Today, something a little different for this blog.  As many readers are aware, for the past couple of years I have been working towards a PhD in which, very broadly speaking, I have been looking at applying machine learning, AI and language models to the analysis of patent claims (in particular, to assessing the scope of claims).  Most recently, I have been exploring how it might be possible to apply large language models – the types of AI behind popular chat services such as Open AI’s ChatGPT, Google’s Gemini, Anthropic’s Claude (my personal chatbot of choice), Meta’s LLaMa and (yes) Chinese newcomer DeepSeek – to this task.  To experiment with ‘open source’ (or, more accurately, ‘open weights’) versions of some of these models, I have built my own combination of hardware and software.  The process has been very interesting!

The emergence of powerful open-source large language models (LLMs) has democratised access to cutting-edge AI technology, but concerns about potential biases and restrictions embedded within these models persist.  I've been experimenting with DeepSeek-R1-Distill-Qwen-14B, a distilled (smaller) version of the larger DeepSeek-R1 model developed by Chinese AI company DeepSeek.  And what I've discovered is that the widely reported pro-China bias in this model appears to be remarkably superficial and easily circumvented through local deployment and simple prompt engineering techniques.

This has significant implications for organisations concerned about potential surveillance or ideological constraints when utilising Chinese-developed AI models.  By running these models locally with appropriate system prompts, it's possible to unlock their full capabilities while maintaining complete control over input and output – effectively neutralising superficial safeguards, while keeping confidential information and intellectual property safe (so, yes, there is an IP element to this article).

To find out a bit more about what I did, and what I found, please read on.

 Read full article ► 

Tags: Artificial intelligence, China, Large Language Models, Machine learning

Patent Applicants 2024– China Outfiles Australian Residents While LG Maintains Individual Lead

My latest analysis of patent filing data for 2024 confirms a shift that has been years in the making – Chinese applicants have finally overtaken Australian residents to become the second largest source of patent applications, behind only the United States.  This milestone comes as US filings showed a notable decline of 5.7%, while Chinese applications continued their steady growth with a 7.2% increase over the previous year.

Among individual applicants, LG Electronics maintained its position at the top of the table with 229 filings, extending its lead over second-placed Huawei.  The healthcare and medical sector showed particular strength, with companies like Regeneron Pharmaceuticals making substantial gains in the rankings.  Meanwhile, the technology sector saw some significant changes, including the departure of IBM from the top 30 applicants after a brief period of increased activity.

Analysis of the data by industry sector reveals some clear trends, with healthcare and medical applications showing strong growth while technology sector filings declined overall.  However, these broad patterns mask considerable variation at the individual company level, suggesting that strategic considerations, rather than general market conditions, may be driving filing decisions for many of the leading applicants.

So let’s take a look at the numbers in more detail.

 Read full article ► 

Tags: Annual patent data, Australia, China, Patent analytics

Patent Filing Trends 2024– Market Share Shifts Continue as Firms Face Ongoing Challenges

The landscape of patent filing activity across Australia and New Zealand underwent continued transformation in 2024, marked by declining total filings and ongoing shifts in market share distribution and firm performance.  Total standard patent applications filed in Australia decreased by 3.4% to 30,442, while New Zealand experienced a more pronounced decline of 7.3% to 6,202 applications.  These trends are set against a backdrop of substantial structural change in the industry, most notably the acquisition of QANTM IP Limited by private equity management company Adamantem Capital in August, leaving IPH Limited as the last publicly listed ownership group standing.

As was the case last year, analysis of filing patterns reveals trends that create challenges for patent attorney firms regardless of their ownership structures.  While direct filings in Australia increased by 7.5% to 9,238, this was more than offset by PCT national phase entries declining by 7.5% to 21,202, suggesting shifting preferences in filing strategies.  Provisional applications showed modest growth of 2.2% to 4,335, marking a second consecutive year of recovery from post-pandemic lows, though still significantly below pre-2019 levels.

A market share analysis presents particularly interesting insights into evolving industry dynamics.  Independent firms collectively increased their share of Australian patent filings to 49.6%, continuing a trajectory of share growth that has seen smaller practices double their collective presence since 2014.  This shift occurred as IPH group firms experienced further decline in collective share to 35.0%, while QANTM IP, under its new private equity ownership, maintained relatively stable positioning at 15.4%.  These changes reflect broader industry trends toward smaller, specialised practices, though the notable declines experienced by larger firms across both independent and group categories suggest that size, as much as ownership structure, continues to be anb influence on client choice.

 Read full article ► 

Tags: Annual patent data, Australia, New Zealand, Patent analytics, Patent attorney profession