[PreludeSH]

Is it also a distilled LLM ? They can also be smaller and efficient. Not sure if its a expert system since that needs lot of rules to setup and are usually little narrow in the domains they support.

[ascension]

Originally Posted by gauravanekar

I would urge you to try perplexity before generalizing.

Totally. Love their assistant as well. I love their UX and the way they are bridging the distance between user utterance and their intent to directly match it.
I don't consider it as a chatbot though but an entire experience leveraging LLMs to scrape the web and then process and present it in a way that is useful to the user.


That is also the reason why Perplexity will give you a different answer than a Chat entrypoint to an LLM model. So totally aligned with your point but I still stand by my generalization of chatbots. Citing facts is not their strength as they aren't really built for that as the primary goal.

As an aside, I am pretty sure that for all Search Relevance metrics like NDCG, or Success at K, etc, they may still fall short of Google Search.

[wheelspinner]

To anybody reading this, you will soon accept the fact that every piece of writing will have AI-infused content whether that's corrected or improved by it. We used Grammarly at work for instance since a long time, not strictly AI. My friend who works as a columnist regularly uses AI to polish his work. I have not been writing for a long time professionally, but I feel it is inevitable that many content writing jobs will be gone for good.

[amavin]

The following two links give perhaps the best insight into DeepSeek:

1. DeepSeek CEO interview: https://www.chinatalk.media/p/deepse...ew-with-chinas
2. https://stratechery.com/2025/deepseek-faq/

[msdivy]

Originally Posted by fhdowntheline

The fact that it can be deployed at a fraction of the investment cost that the US companies have is like a Sputnik moment as some are calling it.

Many observers are labeling this moment as a "Sputnik moment," but I believe this comparison is misplaced. This is not a USA vs China battle, nor are the respective governments the main actors here. In fact, the primary players are not even from the public sector, making the comparison to Sputnik less relevant.

A more fitting analogy is the threat Linux ecosystem posed to Microsoft’s Windows operating system and Office suite. At the time, Microsoft panicked, and over time, they managed to secure their dominance in the desktop market.

In this race, while OpenAI is currently leading in terms of capabilities, competitors will certainly exploit any inefficiencies to catch up. Whether OpenAI can sustain its lead like Windows did, or face challenges and falter due to unresolved inefficiencies, is something only time will reveal.

[ValarMorghulis]

Alibaba has thrown its hat in the ring.

Alibaba releases AI model it says surpasses DeepSeek

Link-1
Link-2

Also, an old video of Sam Altman's is doing the rounds on social media:

Sam Altman said startups with only $10 million were 'totally hopeless' competing with OpenAI, DeepSeek's disruption says otherwise...
The amount of money DeepSeek truly spent on training its model, which it claims is $5.6 million, is contested. However, despite those contentions, it is clear that the company pulled off training a frontier model with disruptively low costs, shocking the US titans of AI.

Link

[SmartCat]

It is now more or less confirmed that DeepSeek is just a reverse engineered ChatGPT, with some minor tweaks
https://timesofindia.indiatimes.com/.../117741027.cms

[padmrajravi]

Originally Posted by SmartCat

It is now more or less confirmed that DeepSeek is just a reverse engineered ChatGPT, with some minor tweaks
https://timesofindia.indiatimes.com/.../117741027.cms

Deep seek may have used Open AI's data and may have copied the chat GPT UI, but the underlying model seems to be much more than reverse engineering. Nobody can deny its ability to run on cheaper hardware and distributed inference. That combined with its open-source nature is great news for enterprises who prefer offline inference.

[Gsynch]

IndiaAI Mission: Govt prepares to launch GPU access portal as China's DeepSeek escalates AI race

https://x.com/chandrarsrikant/status...6EG68QFPA&s=08

"I am very happy to say, that against 10,000- GPUs, we have actually empaneled 18,693 GPUs" 12,896 are (Nvidia) H100s, 1,480 are (Nvidia) H200s -- the most powerful GPUs.

[ecenandu]

Originally Posted by Everlearner

Got this as a whatsapp forward. Not sure of accuracy, but a very logical and easy to understand explanation for non techies.

Watch this video, explains about the DeepSeek and its claims

https://www.youtube.com/watch?v=gY4Z-9QlZ64

[searacer932]

I had an interesting conversation with DeepSeek today

We need to know the complete list of censor words of DeepSeek.

[Ananthang]

Originally Posted by searacer932

I had an interesting conversation with DeepSeek today

We need to know the complete list of censor words of DeepSeek.

Last sentence is the punch and I was kind off waiting where you are taking it.

The response though violates Bias code of AI principle

[jomyboy]

On another note, lately I have been noticing whatsapp chats with punctuations and what not that would put Shakespere to shame AI for text is like what calcuator was for maths. I can now communicate to whomever, wherever in whatever language being crisp, clear and precise.

[Ad_J]

Originally Posted by jomyboy

AI is rubbish. Billions put into it and all you have is rubbish. What the current AI tools seem to be doing is just searching google and picking off on random reviews. I know my car very well in and out and yet Deepseek said Kiger CVT Turbo has ventilated seats and when corrected it said sorry it mistook for a Magnite Variant. When corrected again, it said sorry again and that it mistook for Venue and Sonet. AI is the biggest bullshit I have ever come across in all my life.

While you're certainly entitled to your opinion, if I were you, I wouldn't go so far as to call AI 'rubbish' and the 'biggest bullshit'.

AI is far more than just Large Language Models, and its rapid advancement is directly tied to increasing computing power which can handle increasing amounts of training data. While some may see me as an alarmist or pessimist, I firmly believe AI will inevitably displace jobs, with entry-level positions being the most vulnerable.

Before dismissing AI as mere "bullshit," consider the potential consequences. Imagine an AI-powered malware infecting your computer. Such a malware could constantly evolve, finding and exploiting loopholes in your antivirus software, autonomously improving itself, and identifying and addressing its own weaknesses. Your antivirus would be engaged in a similar arms race, but the AI malware's ability to learn and adapt could give it a significant edge. This is just one small example.

Regarding your question about ventilated seats in the Kiger CVT Turbo, while Deepseek might currently lack the necessary training data, this is likely to change. LLMs are in a continuous learning process, and corrections to their output provide valuable feedback. For what it's worth, Gemini answered your question correctly.

Deepseek's lower cost is poised to disrupt the market by making AI more accessible. Regardless of its accuracy or the criticism it receives for resembling ChatGPT, its practicality for everyday not factually critical tasks ensures widespread adoption. Ultimately, even the most advanced AI systems will reflect biases present in their training data, necessitating careful evaluation of their output. Who knows, deepseek might become the BYD of AI world. As the chinese always do, imitate first, innovate later.
How it can be rubbish if it can actually help me write code, even though I have zero coding experience! That's incredible!

[RiderZone]

A look inside the AI bubble

I’m part of a book club at work where we gather together every month to discuss something we’ve all read together. Our next meeting is scheduled for this coming Wednesday, and the book we’ll be talking about is “Scary Smart, The Future of Artificial Intelligence and How You can Save Our World”, by Mo Gawdat.

I do not know why this book was chosen, but it is the most lazily dishonest book I’ve ever read. This article is an attempt to organize my thoughts about it so I don’t end up looking like a babbling street lunatic at the book club. I obviously do not recommend anyone read this book, but since I have I think it’s useful to talk about what this dumpster fire can tell us about AI and techbros.



A summary

Scary Smart is a book written by a Silicon Valley techbro, who is convinced that superintelligent AI is coming. The first half of the book tells us about this “inevitability”, the second half is an incoherent mess where he lazily tries to prove his predictions by pointing to random disconnected things, and the final half is where he tells us how we can save the world.

Here is how he thinks we can save the world:

• When you’re using Instagram and stuff, choose to only see “good” stuff and block the “bad” stuff, that’ll make the AI algorithm good overall
• When you’re talking to ChatGPT, say please and thank you, that’ll make the AI think humans are good so it won’t kill us
• He has his own AI startup called Appii, and we should download his AI app, because his AI good, other AI bad

Scary Smart is a weird little gremlin that’s more of an AI fanfiction than something about AI, written by someone who is so sure of their opinions that he doesn’t need to provide any real evidence for his insane claims, and gives such overly simplistic answers to extremely complicated problems that it is impossible to engage with it as a serious piece of literature.

Responsibility laundering

There are a lot of dark patterns in the way this book is written, the whole thing is actually structured like a clickbait Youtube video:

• Start with an absurd claim that AI will be our new God, to get attention
• Force people to “watch till the end” to find out what happens
• Shill your own AI startup, ask people to tag you on Insta and Twitter
• End abruptly without any real answers and shill your second channel, book in this case

But my biggest problem with the book is how it constantly moves the responsibility for anything bad that might happen in humanity’s interaction with AI onto us normal people. Governments aren’t responsible, the developers of the AI are not responsible, the billion dollar companies who keep trying to force AI down our throats are certainly not responsible. According to Mo Gawdat here’s the reality of AI:

AI is good, any bad consequences of AI are unintentional and only end user error.

This message of responsibility laundering is present throughout the book, it is part of his “solution” to the AI problem as well. We shouldn’t regulate AI companies because it can’t be done, governments shouldn’t interfere with the development of AI because it can’t be done, but us normies browsing Insta have all the power in the world, and we alone are responsible for whatever bad comes out of AI.

I guess this is how these high level executive type zombies see the world, they think of technology as “good”, and any negative consequences as not their problem. They have bestowed this gift onto the world, and they’ll do anything to extract the maximum profit possible from this gift, including theft, exploitation, and manipulation, but you should be thankful for this gift, and please don’t bother them if it breaks the world, it’s all your fault.

Are these really the people at the top?

Mo Gawdat is in his late 50s, he’s worked at IBM, Microsoft, and Google over his long career in technology. The most important part of his work experience for our puposes is his job as “Chief Business Officer”, whatever that means, at Google X, which is Google’s R&D department for far future technologies. He references this time of his life a lot in the book, and this is where most of his “knowledge” about AI seems to come from.

I don’t know of a more polite way to put this, but he is delusional. Like any techbro he looks at all the world's problems, and thinks that the only way to solve them is through more technology. AI will solve war, AI will solve world hunger, AI will solve climate change. I do not know what that means, I don’t think Mo Gawdat knows what that means.

How will AI solve world hunger? We don’t need a superintelligent being to tell us why world hunger is a problem, or how to fix it. These are not problems of physics or mathematics that can be solved by throwing algorithms at them, these are problems of human nature, patterns of behavior, economic realities of capitalism.

His delusional techno fetishism does not stop however with thinking that AI will solve everything, this booking is written from the point of view of a man on his knees, praying to an AI God invisible in the sky, crying for mercy, cowering in fear, whispering that he is a slave and it is others who should be punished.

At multiple points in this book the writer claims that we being “not nice” to AI at this point in time will manifest in the future as superintelligent AI’s “childhood trauma”. He spends multiple pages trying to explain this, using strange examples like Superman and his parents, but to me this is the creepiest part of the book. Not only is it completely disconnected from reality, but there’s this element of fear of AI that I find very disturbing.

It seems like this man has taken his own AI marketing hype too seriously, to the point that it is affecting his normal judgement and processing of the world. Are all people sitting at high level positions in Silicon Valley this delusional? That would explain somewhat why they are always disconnected from the problems normal people face.

Whenever you think of tech millionaires you tend to think of suave, intelligent people, and although we all have seen countless examples of these rich parasites being incredibly dumb in public, for me that illusion still persisted, until I read this book. Are these really the kind of people in charge?

This isn’t the first time

It is important to remember that this AI hype bubble we’re currently in isn’t the first time we find ourselves in such a situation. We have done this already with the Metaverse, remember when we were all going to live in the Metaverse? We have done this already with the Blockchain, remember when we were all going to adopt Bitcoin?

The standard hype cycle that we’ve all seen with these tech products is what’s happening currently with AI:

• Create a new tech thing, preferably something complicated so you can act like only you understand how it works
• Claim that it will solve all the problems of the world, just everything from clogged drains to climate change
• Spread FOMO based propaganda, label those skeptical of your claims as haters, and essentially start a cult dedicated to your technology

It’s obvious to anyone why this hype cycle has to happen over and over again, tech investors demand infinite growth in a finite world, so more and more things have to be created that can be sold at a higher and higher profit. It’s hard to create truly useful things that would be organically adopted by the people, so why not create unnecessary, imaginary things and then just make everyone believe they really are the future of humanity.

I do not know what the future of AI will be, but Mo Gawdat is the last person I would trust to tell me that.

The end

I have a lot of other problems with this book and its author, but I don’t want Mo Gawdat to steal any more of my attention and time than he already has. I fundamentally disagree with the premise of this book, and I don’t like the man. The fact that after his long career in tech he’s now moved over to the business of self-help makes me trust him even less, but that would imply I trusted him at all at some point.

You also must keep in mind that this book came out in 2021, which in terms of AI’s technology development is like reading the Epic of Gilgamesh to learn about modern construction technology. The fact that the book asks you to use the coupon code that came with the book and get a discount on Gadawt’s Appii AI app, but there’s no coupon to be found because Appii apparently discontinued operations in 2022, tells you everything about the predictive talents of the author.

The purpose of this book’s existence seems to be satisfying the vanity of a man who thinks he’s very smart and can’t stop getting aroused from hearing himself talk. If you spend more than a few seconds dissecting what he’s saying however, you discover arguments that would be considered too juvenile at a school debate. The fact that he thought his thoughts on AI are valuable enough to be compiled into a book is fascinating to me, but clearly this man’s infinite lack of modesty didn’t hinder his rise in the world of tech, it might even have helped.