With 38.8% of tweets providing positive feedback, other users wonder if the AI’s content filtering is more restrictive than its competitors, as it sidestepped certain topics.

But is it truly on track to challenge ChatGPT, or is it simply another flash in the pan?

Between January 20th and January 29th, the AI model, developed by Chinese startup DeepSeek, received 47.3% neutral tweets and 13.9% dissatisfied users, but the strong preference for DeepSeek stood out. 

A noteworthy 86.8% of users who explicitly compared different language models showed a preference for DeepSeek, noting its affordability and performance. 

ChatGPT, by comparison, earned the preference of just 12.1%, with even fewer users favouring other models like Claude and Gemini.

Here’s how they stacked up across key aspects:

• Performance – Some users found DeepSeek faster and better suited for complex reasoning, while others still preferred ChatGPT’s more nuanced responses.
• Cost – DeepSeek’s affordability was a major advantage, especially among users who found ChatGPT’s pricing restrictive.
• Accuracy – While many found DeepSeek impressive, some noted gaps in precision when dealing with complex or ambiguous queries.
• Hallucinations – Like many AI models, DeepSeek occasionally generated misleading or incorrect information, leading some users to question its reliability.

“DeepSeek’s rapid adoption highlights a strong demand for cost-effective AI models. Many users highlighted price as a key factor when comparing LLMs. However, concerns about accuracy persist. If DeepSeek can balance affordability with reliability, it has the potential to reshape the AI market,” said an AI research analyst at TopView.

Even with its growing popularity, some users have spoken about DeepSeek’s content moderation policies. A particularly striking observation involved the model’s tendency to censor sensitive topics, prompting several users to wonder whether the moderation came from external regulations. 

One user pointed out, “Fun demos of DeepSeek’s new r1 shutting itself down when asked about topics the CCP doesn’t like. But the censorship is obviously being done by a layer on top, not the model itself. Anyone have the open-source version running and able to test whether/how much it also censors?”

This type of censorship has led to fears that DeepSeek could be subject to political influences, particularly the Chinese government’s regulations. A user remarked, “DeepSeek’s newest AI model is impressive—until it starts acting like the CCP’s PR officer. Watch as it censors itself on any mention of sensitive topics.”

Performance comparisons between DeepSeek and ChatGPT have also been a main part of the discussions. Many users found DeepSeek faster and more effective at handling complex reasoning, while others still preferred ChatGPT for its nuanced responses. 

However, the lack of precision when DeepSeek encountered ambiguous queries has been noted by some users as a drawback. Nonetheless, the affordability factor has kept the tool in users’ good graces, especially when compared to ChatGPT’s pricing.

In terms of user preferences, DeepSeek was a top choice among the 372 users who specified their preferred large language model (LLM). With 86.8% preferring DeepSeek, its key selling points were cost and speed, although criticisms on accuracy and censorship continue to simmer.

With DeepSeek attracting high attention, many wonder if it will be able to overcome the concerns  around censorship and maintain its affordability, or will competitors like ChatGPT continue to dominate the field? Time will tell if DeepSeek will become a resilient competitor in the AI space.

This analysis was made possible by TopView, which provides in-depth insights using AI-powered tools to explore user sentiments and preferences.

DEPLOY was built in collaboration between the Australian National University (ANU) and the National Cancer Institute (NCI) in the US.

Currently, diagnosing brain tumours often involves a lengthy wait for DNA methylation profiling, a complex and resource-intensive process that can take weeks. 

This delay can be agonizing for patients and hinder the timely initiation of treatment, especially for aggressive tumours.

“The current gold standard for identifying different kinds of brain tumours is DNA methylation-based profiling,” explains Dr Danh-Tai Hoang, a researcher at ANU. “But the time it takes to do this kind of testing can be a major drawback, often requiring several weeks or more when patients might be relying on quick decisions on therapies. There’s also a lack of availability of these tests in nearly all hospitals worldwide.”

DEPLOY tackles this challenge by leveraging deep learning, a form of artificial intelligence. It analyzes microscopic images of patient tissue samples (histopathology images) to predict the tumour subtype and infer the DNA methylation state. 

This essentially reveals which genes are active within the tumour, important information for doctors to determine the most effective treatment course.

The development of DEPLOY involved a three-pronged approach. The first model classifies brain tumour subtypes directly from the images. The second one focuses on predicting DNA methylation-based on those same images. 

Finally, the third model considers patient demographics like age and sex to further refine the tumour classification. In combining these models, DEPLOY achieves almost perfect accuracy.

Researchers precisely tested DEPLOY using data from over 4,000 patients across the US and Europe. The results were outstanding, with DEPLOY consistently achieving a 95% success rate in identifying tumour subtypes. However, the actual potential of DEPLOY goes beyond accuracy.

“Furthermore, when given a subset of 309, particularly difficult-to-classify samples, DEPLOY was able to provide a diagnosis that was more clinically relevant than what was initially provided by pathologists,” highlights Dr Hoang. 

This means that DEPLOY can act as a valuable second opinion for pathologists, potentially leading to more precise diagnoses and improved patient outcomes.

Countries with limited access to advanced diagnostic tools, like Nigeria, face a huge burden of brain tumours. The high cost of DNA methylation profiling and the lack of infrastructure often create a diagnostic bottleneck. Here, DEPLOY brings hope.

“So, there’s an urgent need to do something in the meantime, to give the oncologist some initial recommendations on what is the tumour subtype, which in turn may inform on how to tailor the treatment,” explains Dr Kenneth Aldape, Chief of the Laboratory of Pathology at NCI. 

Through the elimination of the need for expensive and time-consuming DNA methylation profiling, DEPLOY could expedite diagnoses and ensure that patients receive timely treatment.

DEPLOY’s underlying technology can be adapted to diagnose other cancers, ensuring a future where AI helps in early and accurate cancer detection. This, in turn, can lead to better treatment outcomes and a brighter future for countless patients around the world.