A.I. News Chinese AI Assistant Shocks Big Tech with Low-Cost, Efficacy

[Gandalf_The_Grey]

Content source

https://www.thurrott.com/a-i/316361/chinese-ai-assistant-shocks-big-tech-with-low-cost-efficacy

A Chinese AI assistant called DeepSeek has roared to the top of the iPhone App Store charts, sending Big Tech stocks into a tailspin and raising new questions about the investments these firms have made. How is it possible that this thing is so good?

...

What’s most troubling about DeepSeek is that it’s able to rival the best AI models provided by Big Tech and its AI partners, but at a fraction of the cost. The Wall Street Journal reports that DeepSeek’s owners spent just $4 million training its latest models, compared to the hundreds of millions or billions of dollars that companies like OpenAI and Anthropic spend on similar work. And Microsoft, notably, is spending $80 billion in its current fiscal year just to build out the infrastructure required to run these models.

DeepSeek isn’t without problems–it won’t answer sensitive questions about China, for starters–but its instant success triggered a Big Tech/AI stock sell-off on Friday. As I write this, Nvidia’s stock price is down 11.4 percent, Microsoft is down 17 percent, and Google is off 6.3 percent. I’m curious if this will have a long-term impact on these companies, which have thus far dominated this new AI era.

Chinese AI Assistant Shocks Big Tech with Low-Cost, Efficacy

A Chinese AI assistant called DeepSeek has roared to the top of the iPhone App Store charts, sending Big Tech stocks into a tailspin.

www.thurrott.com

 

[Marko :)]

It's Chinese. No thanks!

 

[Victor M]

It's Chinese. No thanks!

But it is open source, which means you can download, inspect it, learn from it or modify it, and then call it your own. ( not too sure about the MIT License particularly). Being owned by a communist coutnry doesnt matter in this case. It is not a closed source Tik Tok. I for one would never use a Chinese AV, but this thing is open source !

 

[cartaphilus]

What people don't realize is not that China does it, it's how they do it. They copy and steal the Intellectual Property and code. I don't know if they did that here or not since I don't know this from Adam.

Of course they can provide the "same" looking code since they never developed it. Time and $$$ goes towards R&D and learning via trials and failure how to perform the task. Sure I can grab the recipe and make it identical or nearly identical but what I am missing are the notes, the lessons learned, what it took to make it. Those notes and lessons learned aid in further development of the item. That is what Chinese are missing. That is why they can copy but they can't innovate. Since the innovate potion was never taken since it can't be, that has to be taught, shown and worked through.

That is also why if someone were to send today's cell phone technology to let say 1980, not much would change. Sure, the "current" tech would most likely advance by 15% to 20% but it wouldn't be revolutionary. You can't just make the item, you have to learn how to make the item. You have to create the substrate that the item is made from. You can't just suddenly reproduce a 4 nm process during an epoch where the smallest item is in micro scale, you just don't have the know how.

 

[Marko :)]

But it is open source, which means you can download, inspect it, learn from it or modify it, and then call it your own. ( not too sure about the MIT License particularly). Being owned by a communist coutnry doesnt matter in this case. It is not a closed source Tik Tok.

What open source essentially means is you can see the code and build the software yourself to run it locally. You still don't know what (re)sources China used to teach this AI. In the article, you can clearly see it's heavily censored. This is exactly the reason why no one should use it.

 

[SeriousHoax]

I have been using it for more than a month almost daily and it's very good. It's almost similar to ChatGPT. Some answers are almost identical in answer quality and pattern that they use. So, they definitely copied the essence of ChatGPT. Google Gemini/Claude/Llama on the other hand are a bit different in style though Google Gemini for 9/10 things I search for gives me the worst answer. Talking about DeepSeek like any other AIs they often make things up which is their biggest issue. But it's faster at answering and much faster at generating codes than the free version of ChatGPT.
Apparently DeepSeek's censoring happens on the server side which is understandable (as even the USA made ChatGPT and Gemini censor many government related things). If you run DeepSeek locally then it can answer those China related topics but probably not all of them obviously.

 

[Victor M]

it's heavily censored. This is exactly the reason why no one should use it.

Nobody is saying we should 'trust' the code as is. We have our own programmers who can examine the code, extract useful portions from it. I dont think this is a case of introducing poisonous code. The censoring may also be done with limiting the training data: no data on china; so no answers are available.

If you must take a cynical angle to it, I would say that their open sourcing the code is an act of destablization.They would be able to foresee that the market would go crazy. It is the kind of thing that a hacker would do, publish the exfiltrated data, publish the credit card numbers, make the company owe up to being hacked. The benefit to them is political instead of monetary.

 

[Marko :)]

The censoring may also be done with limiting the training data: no data on china; so no answers are available.

If that was the case, AI would immediately respond how it has no data. But as seen from this example, it starts to write and then it suddenly stops and basically says "we can't talk about that". It has data, but China censored it about certain topics.

 

[Victor M]

I could do that with a few lines of code inserted into the answer generation part. If topic=china then halt.
Anyways, lets see what happens.

 

[badboy]

I read a funny news from Russian users who tried this AI. He refused to admit his mistake when he answered incorrectly, moreover he started to justify the Chinese Communist Party when it came to the events in Tiananmen Square.

So, no thanks. I personally don't need that kind of AI. It's like asking AI Kandinsky from Russia about the war in Ukraine.

 

[Behold Eck]

I could do that with a few lines of code inserted into the answer generation part. If topic=china then halt.
Anyways, lets see what happens.

Time will tell and lets hope it`s something good but I wouldn`t use because of concerns already mentioned.

Regards Eck

 

[TairikuOkami]

So while the rest of the world is like, lets be careful about AI's capabilities, China unleashes it's full potential tainted with it's coding, just to see, what happens. What would go wrong.

 

[cartaphilus]

So while the rest of the world is like, lets be careful about AI's capabilities, China unleashes it's full potential tainted with it's coding, just to see, what happens. What would go wrong.

View attachment 287318

Considering that their top of the line submarine sunk in port because they decided to dive without closing all the hatches....I am not worried. The sub was still tied to the pier.

 

[Gandalf_The_Grey]

DeepSeek exposes database with over 1 million chat records

DeepSeek, the Chinese AI startup known for its DeepSeek-R1 LLM model, has publicly exposed two databases containing sensitive user and operational information.

The unsecured ClickHouse instances reportedly held over a million log entries containing user chat history in plaintext form, API keys, backend details, and operational metadata.

Wiz Research discovered this exposure during a security assessment of DeepSeek's external infrastructure.

The security firm found two publicly accessible database instances at oauth2callback.deepseek.com:9000 and dev.deepseek.com:9000 that allowed arbitrary SQL queries via a web interface without requiring authentication.

Apart from all the concerns that arise from DeepSeek being a China-based technology company, meaning it has to comply with aggressive data access requests from the country's government, the company does not appear to have established a solid security stance, placing sensitive data at risk.

The exposure of user prompts is a privacy breach that should be very concerning for organizations using the AI model for sensitive business operations.

Additionally, the exposure of backend details and API keys could give attackers a way into DeepSeek's internal networks, privilege escalation, and potentially larger-scale breaches.

Earlier this week, the Chinese platform was targeted by persistent cyberattacks, which it appeared unable to thwart, forcing it to suspend new user registrations for nearly 24 hours.

DeepSeek exposes database with over 1 million chat records

DeepSeek, the Chinese AI startup known for its DeepSeek-R1 LLM model, has publicly exposed two databases containing sensitive user and operational information.

www.bleepingcomputer.com

 

[cartaphilus]

DeepSeek exposes database with over 1 million chat records

DeepSeek exposes database with over 1 million chat records

DeepSeek, the Chinese AI startup known for its DeepSeek-R1 LLM model, has publicly exposed two databases containing sensitive user and operational information.

www.bleepingcomputer.com

The only difference between then and now is that now we know that they know. If you can't keep your henhouse closed the wolf will eventually get in.

This reminds me the DSLReports / BroadbandReports (which is dead now) 2011 database hack . Where hackers stole username and password data all of it was unsalted plain text. yay

 

[Brownie2019]

BEIJING, Jan 27 (Reuters) - Chinese startup DeepSeek's launch of its latest AI models, which it says are on a par or better than industry-leading models in the United States at a fraction of the cost, is threatening to upset the technology world order.
The company has attracted attention in global AI circles after writing in a paper last month that the training of DeepSeek-V3 required less than $6 million worth of computing power from Nvidia H800 chips

https://www.reuters.com/technology/artificial-intelligence/what-is-deepseek-why-is-it-disrupting-ai-sector-2025-01-27/

 

[nickstar1]

Well if the United States actually cared about helping it's own people rather then putting on political clown shows maybe we would have caught up to technology. But instead, we all want to attack each other.

 

[Aleeyen]

Like any other technology this so called AI technology will also become cheaper and cheaper.

 

[Zero Knowledge]

I'll leave the anti Chinese comments alone and focus on the wider implications.

But competition is good for the economy, good for society and good for people, the problem arises when one or two companies become monopolies thus controlling the market and setting noncompetitive prices which tends to happen in modern economies.

In other words cartels form and control the market to the detriment of society. Quality goes down and prices go up due to players manipulating the market which they have control off

ChatGPT (And other LLM's based in USA/EU) were a monopoly and a new model from China is good the world, more competition pushes innovation which in the end we benefit from.

 

[TairikuOkami]