Has DeepSeek burst the US tech bubble?

[WatChe]

As a developer with a long experience, I will say this - for me such systems are useful in such directions:
1. generation of primitive but very voluminous code or similar information
2. Searching for specific errors in large projects (he is a bit more attentive than me )
3. quick answer to a question that I can search for myself, but... I save time. But then I make sure I understand the topic thoroughly - because in the end I should have knowledge.

To get knowledge it's advisable to cross check the answers provided by ChatGPT or other AI Chat Bots. My personal experience with these AI chat bots is that at times the answers given by them needed fine tuning. There are online resources which are considered reliable (not every website has authentic information), it's better to check them also rather then solely relying on AI Chat Bots answers.

With code its different because you can easily figure correctness of code by checking whether your code is giving desired output or not.

[STT]

From what Ive read its coding itself which will be changed most of all as AI is most applicable to learning to code anything you tell it which does make for hard competition to people who must sleep and make a profit following instruction for any program request or design.

Some of the biggest changes will be non public facing alterations to streamline production of modern programs.  It could be a good thing for all of us as programs have become sometimes years long in the making and then hard to alter, update, fix and revise once done.

   If we end up with a process that hinders business it has to change, the number one rule of business I try to remember is efficiency will always win not anyone trying to cling onto their particular profitable perspective in the business.   The paradigm of a bubble is it must be burst and that is something good, that would fit if deepseek was just a first volley of competitive efficiency showing up & then we should all cheer progress.

[Royal Cap]

Actually deepseek is a chatbot that can think like a human.  This in itself begs the question of how to resolve this issue.  We can also see what he thinks.  These features are available in the premium version of chatgpt (as far as I know).  Now let's talk about the fall of nvidia.  Actually running this Ai model requires a lot of GPU.  Most of which are trained with nvidia's GPU.  But since the arrival of deepseek, these AI models are being run on their own local machines.  As a result, these models are being run using a very small amount of CPU.  That's why Nvidia's stock has fallen so much.

[DrBeer]

As a developer with a long experience, I will say this - for me such systems are useful in such directions:
1. generation of primitive but very voluminous code or similar information
2. Searching for specific errors in large projects (he is a bit more attentive than me )
3. quick answer to a question that I can search for myself, but... I save time. But then I make sure I understand the topic thoroughly - because in the end I should have knowledge.

To get knowledge it's advisable to cross check the answers provided by ChatGPT or other AI Chat Bots. My personal experience with these AI chat bots is that at times the answers given by them needed fine tuning. There are online resources which are considered reliable (not every website has authentic information), it's better to check them also rather then solely relying on AI Chat Bots answers.

With code its different because you can easily figure correctness of code by checking whether your code is giving desired output or not.

This is the most subtle part of how AI works. As we understand - there is no AI, i.e. intelligence, it's just a language model. For example, with ChatGPT , it was not uncommon to have to fracture and elaborate on a question in order to get an acceptable answer. With claude, the situation is much better. In my opinion he does not suffer from “sclerosis” and “phantom fantasies”

[DanWalker]

tried deepseek, not impressed mainly because it's hard to set up and although their AI chat is free, it has a T&C that you need to agree on where they'd collect your data for training purpose and probably for other things.
but since grok comes out with its reasoning model as well as deep search I didn't really use deepseek anymore, also claude just come up with reasoning model in version 3.7
pretty much deepseek already outdated after just few months, the AI scene is crazy.

in my opinion, such overblown news that deepseek burst the US tech bubbles, meanwhile reasoning models require a lot more compute than ever before and that means US company like NVIDIA just getting higher potential to produce more powerful GPU to appease the AI market.

Any app will collect user data for certain purposes, not just deepseek or apps from China, and depending on the app, there will be different levels of data collection. Most of them will be clearly stated in the terms and conditions, and what's more interesting about deepseek is that it's open source. That means if you suspect an app is collecting more of your data than they say, you can verify it yourself. Unlike closed source apps, we don't know what data they are collecting from us.

I also tried Grok which is built into X and I like it better than any other AI.

Instead of thinking it was created to burst the US tech bubble, why don't we look at the positive side? It brings innovation in cost savings in the field of AI development, creates a competitive environment and gives users more choices.

[o48o]

Actually deepseek is a chatbot that can think like a human. 

No it isn't. It's a language model that doesn't think in any way. It just predicts text and often gets things wrong.

Also alternatives to deepseek and deepseek itself might hit a wall soon if they can't use copyrighted material to train them. Which is actually good in terms of protecting copyrighted material. Deepseek chinese servers will try to keep using it, but they will hit by sanctions and that might turn it unprofitable for them as well.

But who knows what happens eventually when we have this much new rogue players on economics.

[iBaba]

To get knowledge it's advisable to cross check the answers provided by ChatGPT or other AI Chat Bots. My personal experience with these AI chat bots is that at times the answers given by them needed fine tuning. There are online resources which are considered reliable (not every website has authentic information), it's better to check them also rather then solely relying on AI Chat Bots answers.

With code its different because you can easily figure correctness of code by checking whether your code is giving desired output or not.

It's very necessary to say that your ability to explore the best part of these chat bots is how best you can arrange your questions and prompts, if you know how to generate good prompts, you will get good responses. I have tried DeepSeek too like I tried ChatGPT, and ChatGPT still offers a better interface design and smooth response. DeepSeek may probably be more smarter in giving response at the barest minimal of question twisting and intelligence, but it is still important for DeepSeek to work on their interface to enable easygoing activities particularly in terms of formatting of texts and responses.

[WatChe]

It's very necessary to say that your ability to explore the best part of these chat bots is how best you can arrange your questions and prompts, if you know how to generate good prompts, you will get good responses. I have tried DeepSeek too like I tried ChatGPT, and ChatGPT still offers a better interface design and smooth response. DeepSeek may probably be more smarter in giving response at the barest minimal of question twisting and intelligence, but it is still important for DeepSeek to work on their interface to enable easygoing activities particularly in terms of formatting of texts and responses.

You have to spend time in arranging questions and prompt in way that AI chat bot can get idea about what exactly you want to know. It's like asking something from a wise person, the exactness of answer depends on how well you placed your question. 
As per my personal experience with DeepSeek since start of February, it has gone up in performance since it's initial launch in Jan 2025. Initially I have seen it crashing on some queries and this issue is very much settled. But for now, ChatGPT is the king in AI chat bot domain because it has an algorithm that's matured over a period of time.

[DrBeer]

To get knowledge it's advisable to cross check the answers provided by ChatGPT or other AI Chat Bots. My personal experience with these AI chat bots is that at times the answers given by them needed fine tuning. There are online resources which are considered reliable (not every website has authentic information), it's better to check them also rather then solely relying on AI Chat Bots answers.

With code its different because you can easily figure correctness of code by checking whether your code is giving desired output or not.

It's very necessary to say that your ability to explore the best part of these chat bots is how best you can arrange your questions and prompts, if you know how to generate good prompts, you will get good responses. I have tried DeepSeek too like I tried ChatGPT, and ChatGPT still offers a better interface design and smooth response. DeepSeek may probably be more smarter in giving response at the barest minimal of question twisting and intelligence, but it is still important for DeepSeek to work on their interface to enable easygoing activities particularly in terms of formatting of texts and responses.

One of the finer points of working with “AI” (in quotes for a reason) is that you need to be VERY precise and very qualitative in formulating the question or task. Unfortunately, we must realize that “AI” does not have the most important thing - intelligence. That's why unless we formulate a sentence so that it could be correctly processed by a large language model, we won't get an adequate answer or a normal code

[Reatim]

One of the finer points of working with “AI” (in quotes for a reason) is that you need to be VERY precise and very qualitative in formulating the question or task. Unfortunately, we must realize that “AI” does not have the most important thing - intelligence. That's why unless we formulate a sentence so that it could be correctly processed by a large language model, we won't get an adequate answer or a normal code

it is quite literally called artificial intelligence

all its “intelligence” is supposed to come from you because you are the basis of ai some people feed ai prompts that wouldn’t get them the precise answer they want while others can easily garner information with only just submitting one prompt compared to others who have to submit multiple that is simply tedious

[X-ray]

One of the finer points of working with “AI” (in quotes for a reason) is that you need to be VERY precise and very qualitative in formulating the question or task. Unfortunately, we must realize that “AI” does not have the most important thing - intelligence. That's why unless we formulate a sentence so that it could be correctly processed by a large language model, we won't get an adequate answer or a normal code

it is quite literally called artificial intelligence

all its “intelligence” is supposed to come from you because you are the basis of ai some people feed ai prompts that wouldn’t get them the precise answer they want while others can easily garner information with only just submitting one prompt compared to others who have to submit multiple that is simply tedious

honestly current AI just works better in context because of its nature that it works like statistics to predict the next word based on provided context.
something like deepseek can have reasoning model but at the end of the day it's just text prediction (CMIIW) that's why people are waiting for AGI which might present something new.
as of now event current LLM is already good enough.

[DrBeer]

One of the finer points of working with “AI” (in quotes for a reason) is that you need to be VERY precise and very qualitative in formulating the question or task. Unfortunately, we must realize that “AI” does not have the most important thing - intelligence. That's why unless we formulate a sentence so that it could be correctly processed by a large language model, we won't get an adequate answer or a normal code

it is quite literally called artificial intelligence

all its “intelligence” is supposed to come from you because you are the basis of ai some people feed ai prompts that wouldn’t get them the precise answer they want while others can easily garner information with only just submitting one prompt compared to others who have to submit multiple that is simply tedious

I am well aware that the name “AI” is a marketing name for a solution that can be called as : A big language model for human interaction, with a knowledge base pre-trained with neural networks, on a huge amount of open data I.e. AI today is essentially a model for comfortable interaction in a native language with a huge knowledge base

[Free Market Capitalist]

This is the most subtle part of how AI works. As we understand - there is no AI, i.e. intelligence, it's just a language model. For example, with ChatGPT , it was not uncommon to have to fracture and elaborate on a question in order to get an acceptable answer. With claude, the situation is much better. In my opinion he does not suffer from “sclerosis” and “phantom fantasies”

Well, not yet in the human sense but I'll summarize what Sam Altman told Joe Rogan in an interview about what he thought about AI 10 years ago. He categorized what he believed they could do into:

1) Repetitive tasks, such as putting an AI on a robot and having it go about tightening screws or carrying things along a defined path.
2) Blue collar tasks, he believed they could be helpful but wouldn't amount to much.
3) Creativity. About this I believed that it would be impossible for these models to have any skill.

Well, it seems that time is going to prove him wrong even on point 3.

I just think AI nowadays is like the internet on 1995. We still have a lot to see.

[DaNNy001]

As a developer with a long experience, I will say this - for me such systems are useful in such directions:
1. generation of primitive but very voluminous code or similar information
2. Searching for specific errors in large projects (he is a bit more attentive than me )
3. quick answer to a question that I can search for myself, but... I save time. But then I make sure I understand the topic thoroughly - because in the end I should have knowledge.

To get knowledge it's advisable to cross check the answers provided by ChatGPT or other AI Chat Bots. My personal experience with these AI chat bots is that at times the answers given by them needed fine tuning. There are online resources which are considered reliable (not every website has authentic information), it's better to check them also rather then solely relying on AI Chat Bots answers.

With code its different because you can easily figure correctness of code by checking whether your code is giving desired output or not.

Deep seek isn't really bad but I had a slight problem with it, after asking a question on something philosophical it started giving me but numerical results at first before give me the answers that I needed to know.... I kept wondering why that happened, I think it was a technical glitch or something but i don't think I've had such issues with chatgtp. I have been using ai bots for a while now and I think it's best to cross check them with Google or Wikipedia because somethings that they provide might not be accurate...but still, deepseek came out of the blue and plans on taking chatgtp out of business but I wonder why I'm not hearing much about it These days.

[l99l]

To get knowledge it's advisable to cross check the answers provided by ChatGPT or other AI Chat Bots. My personal experience with these AI chat bots is that at times the answers given by them needed fine tuning. There are online resources which are considered reliable (not every website has authentic information), it's better to check them also rather then solely relying on AI Chat Bots answers.

With code its different because you can easily figure correctness of code by checking whether your code is giving desired output or not.

It's very necessary to say that your ability to explore the best part of these chat bots is how best you can arrange your questions and prompts, if you know how to generate good prompts, you will get good responses. I have tried DeepSeek too like I tried ChatGPT, and ChatGPT still offers a better interface design and smooth response. DeepSeek may probably be more smarter in giving response at the barest minimal of question twisting and intelligence, but it is still important for DeepSeek to work on their interface to enable easygoing activities particularly in terms of formatting of texts and responses.

ChatGPT platform is generally stable and its interface is simple and stylish, in my opinion it is a few steps ahead of Deepseek. In terms of Privacy and Data Usage, I can put both of them in the same bag, there is no difference in this regard.
People have formed certain habits in Chatgpt usage and I think that the majority of users are satisfied with this situation with the data they have collected so far.
I do not think that users who are satisfied with ChatGPT will prefer DeepSeek.

[WatChe]

Deep seek isn't really bad but I had a slight problem with it, after asking a question on something philosophical it started giving me but numerical results at first before give me the answers that I needed to know.... I kept wondering why that happened, I think it was a technical glitch or something but i don't think I've had such issues with chatgtp. I have been using ai bots for a while now and I think it's best to cross check them with Google or Wikipedia because somethings that they provide might not be accurate...but still, deepseek came out of the blue and plans on taking chatgtp out of business but I wonder why I'm not hearing much about it These days.

Regarding accuracy of results provided by these AI Chat Bots, there is latest incident where a Norway man has registered a complaint against ChatGPT after he was accused of killing his two sons by ChatGPT. I don't say that every answer provided by these Chat bot is false but we need to be watchful in what we are getting from these AI Chat Bots.

Or may be we stop asking questions related to "hallucinations" - where information is invented and presented like a fact by AI.

https://www.bbc.com/news/articles/c0kgydkr516o

[btc78]

To get knowledge it's advisable to cross check the answers provided by ChatGPT or other AI Chat Bots. My personal experience with these AI chat bots is that at times the answers given by them needed fine tuning. There are online resources which are considered reliable (not every website has authentic information), it's better to check them also rather then solely relying on AI Chat Bots answers.

people who take ai’s words as if it were 100% accurate are most likely to spread misinformation ai are not always factual sometimes i ask something and it gives me an answer that is outdated which makes it inaccurate in today’s world so it is always the best to not rely on ai completely and still make our own research and find articles and papers actually written by humans

[beerlover]

It's very necessary to say that your ability to explore the best part of these chat bots is how best you can arrange your questions and prompts, if you know how to generate good prompts, you will get good responses. I have tried DeepSeek too like I tried ChatGPT, and ChatGPT still offers a better interface design and smooth response. DeepSeek may probably be more smarter in giving response at the barest minimal of question twisting and intelligence, but it is still important for DeepSeek to work on their interface to enable easygoing activities particularly in terms of formatting of texts and responses.

ChatGPT platform is generally stable and its interface is simple and stylish, in my opinion it is a few steps ahead of Deepseek. In terms of Privacy and Data Usage, I can put both of them in the same bag, there is no difference in this regard.
People have formed certain habits in Chatgpt usage and I think that the majority of users are satisfied with this situation with the data they have collected so far.
I do not think that users who are satisfied with ChatGPT will prefer DeepSeek.

It is miles better than deepseek, not just a bit. What people who are not in AI doesn't realize is that deepseek took the opensource part of chatgpt back when it was opensource, had the code, and worked on top of it, so they have managed to cut the costs by a large amount thanks to chatgpt itself.

But what you can see from results, by simply asking the same question to both of them, you will see how it is different. Chatgpt has 4.5 as their latest ow, and comparing to what deepseek has, the difference is huge, and that is not what chatgpt even has as latest within the company, they have 5, just not to public yet. Deepseek can't keep spending this little amount, and ever imagine catching the level of Chatgpt ever, they need to increase their funding.

[DrBeer]

As a developer with a long experience, I will say this - for me such systems are useful in such directions:
1. generation of primitive but very voluminous code or similar information
2. Searching for specific errors in large projects (he is a bit more attentive than me )
3. quick answer to a question that I can search for myself, but... I save time. But then I make sure I understand the topic thoroughly - because in the end I should have knowledge.

To get knowledge it's advisable to cross check the answers provided by ChatGPT or other AI Chat Bots. My personal experience with these AI chat bots is that at times the answers given by them needed fine tuning. There are online resources which are considered reliable (not every website has authentic information), it's better to check them also rather then solely relying on AI Chat Bots answers.

With code its different because you can easily figure correctness of code by checking whether your code is giving desired output or not.

Deep seek isn't really bad but I had a slight problem with it, after asking a question on something philosophical it started giving me but numerical results at first before give me the answers that I needed to know.... I kept wondering why that happened, I think it was a technical glitch or something but i don't think I've had such issues with chatgtp. I have been using ai bots for a while now and I think it's best to cross check them with Google or Wikipedia because somethings that they provide might not be accurate...but still, deepseek came out of the blue and plans on taking chatgtp out of business but I wonder why I'm not hearing much about it These days.

No one is saying Deepseek is a bad solution. I'm just saying that the hype around it is unjustifiably high. And for me, the moment when some facts, not the most pleasant for the system, were revealed, also became negative. If someone uses a system and it satisfies him, then it is a good system for this consumer, it's simple But I didn't find any noticeable advantages in it, plus, some “dark spots” in its history create a negative attitude to the product.

[WatChe]

people who take ai’s words as if it were 100% accurate are most likely to spread misinformation ai are not always factual sometimes i ask something and it gives me an answer that is outdated which makes it inaccurate in today’s world so it is always the best to not rely on ai completely and still make our own research and find articles and papers actually written by humans

Getting answers from AI seems an easy solution compared to searching the same query on google and visiting multiple sites. I have noticed that if you ask an AI Chat Bot same question multiple times then answer is different all the time. It might be because AI Chat Bots are still in there initial phase and with time we will see improvement in there performance. But for now we need to remain careful about the results we are getting from these AI Chat Bots, they must be compared with reliable sources.