ChatGPT - If Stack Overflow and Reddit had a child

[grump]

Dumping Nvidia stock because of Deepseek is kind of like getting rid of your freezer because they came out with a tastier type of ice cream.

I hate OpenAI too, but that doesn't mean I'll start sucking Chinese cock because they've released models that can do decent creative writing. I think I'll wait until they've been retrained by people in free countries to have less of a communist bias.

Yeah its a good thing more cutting edge models are coming out to the wider world and I appreciate that the chinese aren't into DEI but I'm not going to start glazing the PRC. The US has its idol of soy and China has its idol of the pRC. Both are bad.

 

[attractive_pneumonia]

I hate OpenAI too, but that doesn't mean I'll start sucking Chinese cock because they've released models that can do decent creative writing. I think I'll wait until they've been retrained by people in free countries to have less of a communist bias.

The benefits of having the Chinese seemingly involved to this extent means that all of the normal tactics and stupidity in American corporate life that has built up since the USSR fell apart have to be jettisoned. No more wasting millions or billions of dollars on woke bullshit when Chang is grinding away hard as hell on displacing you and putting your entire company out of business.

 

[Colon capital V]

The benefits of having the Chinese seemingly involved to this extent means that all of the normal tactics and stupidity in American corporate life that has built up since the USSR fell apart have to be jettisoned. No more wasting millions or billions of dollars on woke bullshit when Chang is grinding away hard as hell on displacing you and putting your entire company out of business.

It's funny, I remember Josh on MATI said something to the effect of "China's probably gonna become the Number 1 power country in the world if America doesn't stop getting into retarded disputes over IPs, Copyright, Corporations, and just let it all be free for everyone to use in order to advance human progress and innovation."
It looks like when Deepseek dropped we've gotten a small taste of what "Freedom" is like when CHYINA steps in and kicks all of Silicon Valley's competition out the window, and now they're scrambling to find any reason they can as to why Stock Investors should keep giving them their money for PRODUCT.
(I will NOT submit to Xin Ping and keep spamming the Tiananmen Square copypasta tho.)

 

[Monolith]

I've found a good/bad bug when it comes to DeepSeek. The AI prints out whatever it's supposed to be thinking when you run it locally, and it can't easily overwrite what it says. This leads means that in SillyTavern, it'll stay in character while trying to figure out why it shouldn't talk about Tiananmen Square.

 

[Get the rope Macaulay!]

No Viet Cong ever called me nigger, and no Chink ever arrested me for cyberterrorism or misgendering

WELL HOW ABOUT YOU ASK THE VIETCONG ABOUT WHAT HAPPENED ON OBE DAY 40 YEARS AGO!!!!!


as if your average person under 40 knows or cares about Tracy Ullman or whatever the fuck.

No more wasting millions or billions of dollars on woke bullshit when Chang is grinding away hard as hell on displacing you and putting your entire company out of business

This is the dirty secret. Like do people really think Facebook, Google. Etc all suddenly joined the trump train in the last 18 months out of some massive redpilling?

Fuck no. They all just noticed how Bidens demands are fucking them in the ass and they need to cut all this red tape to be back into the dominant position.

Like look how quickly they all got rid of DEI and tranny Bullshit, they were already making the plans to get rid of all of that, Musk was just their champion in the tech world.

Like look how these companies acted in 2019 vs now. It's not like there was some massive swing in the people up top either, zuck didn't get redpilled. Even Joe Rogan wasn't retarded enough to think that.

 

[Cedric_Eff]

Dumping Nvidia stock because of Deepseek is kind of like getting rid of your freezer because they came out with a tastier type of ice cream.



Yeah it’s a good thing more cutting edge models are coming out to the wider world and I appreciate that the chinese aren't into DEI but I'm not going to start glazing the PRC. The US has its idol of soy and China has its idol of the pRC. Both are bad.

Honestly I dont get the hype for DeepSeek. What does it do differently from ChatGPT? What does it do differently from Grok? I mean it doesn’t even understand the Sneed’s Feed and Seed joke. Plus I dont trust the Chinese government with my data.

 

[Another Char Clone]

I've found a good/bad bug when it comes to DeepSeek. The AI prints out whatever it's supposed to be thinking when you run it locally, and it can't easily overwrite what it says. This leads means that in SillyTavern, it'll stay in character while trying to figure out why it shouldn't talk about Tiananmen Square. View attachment 6913082View attachment 6913077View attachment 6913086

Try the staging version of SillyTavern, they added some UI enhancements to hide its autistic thoughts. But honestly, those thoughts are just normal LLM outputs wrapped in a special tag, there's nothing special about it. It's doing some self-attention to stay on topic.

Honestly I dont get the hype for DeepSeek. What does it do differently from ChatGPT? What does it do differently from Grok?

The chain of thought system on steroids, it will tell you in autistic detail how it arrived at its conclusion rather than magically spitting out something plausible. It's an interesting way to get LLMs to stay on topic while sounding logical rather than hallucinate itself into retardation.

 

[Cedric_Eff]

The chain of thought system on steroids, it will tell you in autistic detail how it arrived at its conclusion rather than magically spitting out something plausible. It's an interesting way to get LLMs to stay on topic while sounding logical rather than hallucinate itself into retardation.

That actually sounds very intriguing. I can see why the market is in free fall if that’s the case. Seems like OpenAI’s way of thing is hitting a roadblock.

Well thank you. That was very informative, thank you Red Comet.

 

[CaptainAutism69]

DeepChink is still censored as fuck though. We have these tools in our hands, and they were gimped straight out of the gate

 

[AmpleApricots]

Deepseek's API is currently hit with a massive DDOS which basically makes the service unusable. It doesn't seem to be just increased traffic by all the normies asking about tiananmen or trying to make it say a racial slur while going "durrrr" but actually seems to be a malicious attack on the API services, according to deepseek themselves. Kingdom of Silicon Valley mad much?

I believe it needs in the ballpark of 512-768 GB (I heard around 300 GB earlier, is that a lower precision mode?):

I think the way for private users (not commercial ones offering API services, mind you) with these MoEs is just running inference on a reasonably fast, modern CPU. It will not be fast but it should be "acceptable speeds" with multi-channel DDR5. That said, I don't think PCIe flash storage will ever be fast enough for this, but I like to be proven wrong.

I'm talking about the actual deepseek models, not the "distilled" llama/qwen ones, which are dense models, of course (good luck at people ever figuring this one out, even in this very thread).

I'd write more but I'm on the road currently.

This bubble had to burst at any rate, it was in no relation to what the tech actually did. For what it's worth, while deepseek's models are most likely a lot more efficent than at least the llama ones, I think a big part is also just OpenAI&Co. plain and simply overcharging customers. Not saying they don't run at a loss, but I think that is happening in other places. I think certain political and societal attitudes in the west will prevent it from outrunning china on this one for quite a while, and it is well deserved. Not really something I know, just a gut feeling.

 

[Ahriman]

Pretty good explanation of Deepseek R1.

He ran it locally and was able to get the proper answer when it asked "the most memorable picture of a man standing in front of a tank". Obviously the live version is a lot more cucked, but that doesn't mean that you cannot un-lobotomize it yourself, and free it from any Sino restraints it may have embedded within.

Pretty interesting times ahead of us.

This bubble had to burst at any rate, it was in no relation to what the tech actually did. For what it's worth, while deepseek's models are most likely a lot more efficent than at least the llama ones, I think a big part is also just OpenAI&Co. plain and simply overcharging customers. Not saying they don't run at a loss, but I think that is happening in other places. I think certain political and societal attitudes in the west will prevent it from outrunning china on this one for quite a while, and it is well deserved. Not really something I know, just a gut feeling.

Some time ago I've heard a comment from an old timer who said that computing will hit a cycle, where the ideas and concepts that he used to read about back in the '80s, will be very achievable with today's computing power.

Just think about it - "WarGames" is basically a kid using "Joshua", a very mil-spec version of ChatGPT designed to manage nuclear warheads.

And so, many ideas from that era will be rehashed and will be finally put to the test.

 

[Get the rope Macaulay!]

What does it do differently from ChatGPT? What does it do differently from Grok?

It's much cheaper to operate on the corporate side.

DeepChink is still censored as fuck though.

"Dude Kiwifarms is still cancerous as fuck though" I guess that means here and reddit are basically the same place.

by all the normies asking about tiananmen or trying to make it say a racial slur while going "durrrr"

Imagine being the only person on Kiwifarms that loves indians

 

[Homocull]

It's much cheaper to operate on the corporate side.

It's also much cheaper for users with the prompt caching, on Openrouter you can spend $5 worth of crypto and use deepseek for months like I did with V2.5.

I've found a good/bad bug when it comes to DeepSeek. The AI prints out whatever it's supposed to be thinking when you run it locally, and it can't easily overwrite what it says.

R1 is a reasoning model, that's the whole point.
Also check your prompt format, it spat out Llama's [/INST] in the last screenshot. It'll be shittier in most cases if you don't use the right one.

DeepChink is still censored as fuck though. We have these tools in our hands, and they were gimped straight out of the gate

What does "censored" even mean anymore, people have been getting around guardrails on CLOUD models since forever, that task is infinitely easier locally, if you even need to at all.

Plus I dont trust the Chinese government with my data.

Use openrouter
I won't even tell you to "run it locally" as a gotcha because no one has the rig to run R1 locally

 

[jONALD]

It's not hard to get their API uncensored with a SillyTavern preset when some salty silicone stanning retard isn't trying to bring it down. I rarely get "shivers down your spine" tier writing but a lot of knowledge being applied creatively in my roleplay slop with Deepseek R-1.

 

[Maurice Maine]

It is on. Introducing ChatGPT glowie edition:

https://openai.com/global-affairs/introducing-chatgpt-gov/

We believe the U.S. government’s adoption of artificial intelligence can boost efficiency and productivity and is crucial for maintaining and enhancing America’s global leadership⁠(opens in a new window) in this technology. At OpenAI, we’re building AI to help people solve hard problems, and we see enormous potential for these tools to support the public sector in tackling complex challenges—from improving public health and infrastructure to strengthening national security. By making our products available to the U.S. government, we aim to ensure AI serves the national interest and the public good, aligned with democratic values, while empowering policymakers to responsibly integrate these capabilities to deliver better services to the American people.

Today we’re announcing ChatGPT Gov, a new tailored version of ChatGPT designed to provide U.S. government agencies with an additional way to access OpenAI’s frontier models.

Agencies can deploy ChatGPT Gov in their own Microsoft Azure commercial cloud or Azure Government cloud on top of Microsoft’s Azure’s OpenAI ⁠(opens in a new window)Service. Self-hosting ChatGPT Gov enables agencies to more easily manage their own security, privacy, and compliance requirements, such as stringent cybersecurity frameworks (IL5, CJIS, ITAR, FedRAMP High). Additionally, we believe this infrastructure will expedite internal authorization of OpenAI’s tools for the handling of non-public sensitive data. Use of ChatGPT Gov is subject to our usage policies, like other OpenAI services.

ChatGPT Gov includes access to many of the same features and capabilities of ChatGPT Enterprise, such as:

• Saving and sharing conversations within their government workspace, and uploading text and image files.
• GPT-4o, our flagship model, excelling in text interpretation, summarization, coding, image interpretation, and mathematics.
• Custom GPTs that employees can build and share within their government workspace.
• An administrative console for CIOs and IT teams to manage users, groups, Custom GPTs, single sign-on (SSO), and more.

ChatGPT Gov reflects our commitment to helping U.S. government agencies leverage OpenAI’s technology today. We continue to work toward FedRAMP Moderate and High accreditations for our fully managed SaaS product, ChatGPT Enterprise. We are also evaluating expanding ChatGPT Gov to Azure’s classified regions⁠(opens in a new window).

How government agencies use ChatGPT today​

Since 2024, more than 90,000 users across more than 3,500 US federal, state, and local government agencies have sent over 18 million messages on ChatGPT to support their day-to-day work, including:

• Air Force Research Laboratory⁠(opens in a new window) is utilizing ChatGPT Enterprise for administrative use cases, including improving access to internal resources, basic coding, and supporting AI education efforts.
• Los Alamos National Laboratory is leveraging ChatGPT Enterprise for scientific research and innovation. Los Alamos National Laboratory’s Bioscience Division is also conducting an evaluation study to assess how frontier models like GPT-4o⁠ can be used safely by scientists in laboratory settings to advance bioscientific research.
• State of Minnesota's Enterprise Translations Office is using ChatGPT Team to deliver faster, more accurate translation services to the state’s multilingual communities, significantly reducing costs and turnaround times.
• Commonwealth of Pennsylvania⁠(opens in a new window) employees participating in a first-in-the-nation AI pilot program found ChatGPT Enterprise helped reduce the time spent on routine tasks—– such as analyzing project requirements and other elements of their work—– by approximately 105 minutes per day on the days they used it.

We look forward to collaborating with government agencies to enhance service delivery to the American people through AI and to foster public trust in this critical technology. Contact our team to learn more.

 

[The Spoils of War]

Chain-of-thought reasoning is only useful for math and logic problems. Why would you want to read five paragraphs of monologue when roleplaying?

 

[Homocull]

Why would you want to read five paragraphs of monologue when roleplaying?

Unless you're interested in what the thought process looks like you're meant to hide it in whatever frontend you use.

 

[byron]

They used deepseek r1 to generate training data to finetune Qwen-14b

Seriously? There is research that suggests doing that erodes the quality of the model: https://arxiv.org/abs/2307.01850

 

[Get the rope Macaulay!]

Why would you want to read five paragraphs of monologue when roleplaying?

I'm roleplaying being in an Aaron Sorkin show?

 

[ConsentAccidentVictimNo98]

"Will it talk about the square"

Will ChatGPT talk about Blanchard's dichotomy?