ChatGPT - If Stack Overflow and Reddit had a child

[The Mass Shooter Ron Soye]

M3 Ultra Runs DeepSeek R1 With 671 Billion Parameters Using 448GB Of Unified Memory, Delivering High Bandwidth Performance At Under 200W Power Consumption, With No Need For A Multi-GPU Setup

The most expensive ones have 512 GB of memory to work with.

 

[Get the rope Macaulay!]

... n

M3 Ultra Runs DeepSeek R1 With 671 Billion Parameters Using 448GB Of Unified Memory, Delivering High Bandwidth Performance At Under 200W Power Consumption, With No Need For A Multi-GPU Setup\

The most expensive ones have 512 GB of memory to work with.

honestly surprised Apple doesn't just make their own version of DeepSeek and have it come standard on their computers the same way Garageband and the rest of ilife becoming standard meant everyone would use them for the rest of time.

As its basically open source they can literally just steal it and give it a fancy name.

 

[The Mass Shooter Ron Soye]

Ars Technica: Google’s Gemma 3 is an open source, single-GPU AI with a 128K context window (archive)

 

[DeadwastePrime]

Ars Technica: Google’s Gemma 3 is an open source, single-GPU AI with a 128K context window (archive)

wow i cant wait to never get access to it

 

[The Mass Shooter Ron Soye]

wow i cant wait to never get access to it

https://www.kaggle.com/models/google/gemma-3

https://huggingface.co/collections/google/gemma-3-release-67c6c6f89c4f76621268bb6d

---

Lulz:
Anthropic CEO floats idea of giving AI a “quit job” button, sparking skepticism (archive)
AI coding assistant refuses to write code, tells user to learn programming instead (archive)

OpenAI declares AI race “over” if training on copyrighted works isn’t fair use (archive)

 

[Fae Supremacist]

>my enlightened Moldbugian techno-fascist corporate ran society
>your primitive nationalist statism

In a new policy proposal, OpenAI describes Chinese AI lab DeepSeek as “state-subsidized” and “state-controlled,” and recommends that the U.S. government consider banning models from the outfit and similar People’s Republic of China (PRC)-supported operations.

https://archive.is/N06QL

KWAB

 

[The Mass Shooter Ron Soye]

AMD’s Ryzen AI MAX+ 395 “Strix Halo” APU Is Over 3x Faster Than RTX 5080 In DeepSeek R1 AI Benchmarks

[Editor's Note] In the latest benchmarks, AMD is showcasing the AI performance capabilities of its Strix Halo APUs against NVIDIA's latest Blackwell GPUs. The company is telling us the benefits of its large 128 GB VRAM on the Ryzen AI MAX+ 395 APU, which can offer up to 3.05x performance boost versus the discrete graphics card as soon as the LLM model sizes exceed 16 GB VRAM.

ERNIE 4.5 AI model by Baidu claims to match DeepSeek R1 at half the cost

 

[The Mass Shooter Ron Soye]

Phoronix: AMD Announces Open-Source "GAIA" For GenAI But Currently Windows-Only (archive)

AMD on Thursday announced GAIA for "Generative AI Is Awesome" as a means of easily running local large language models (LLMs) on Ryzen AI PCs with the XDNA NPU. GAIA provides an easy software demonstration for Ryzen AI capabilities but sadly for the moment is limited to Microsoft Windows platforms.

Tom's Hardware: AMD launches Gaia open source project for running LLMs locally on any PC

Phoronix: AMD Announces AITER For ROCm To Help Boost AI Performance (archive)

 

[cybertoaster]

M3 Ultra Runs DeepSeek R1 With 671 Billion Parameters Using 448GB Of Unified Memory, Delivering High Bandwidth Performance At Under 200W Power Consumption, With No Need For A Multi-GPU Setup

The most expensive ones have 512 GB of memory to work with.

Fuuuuuuuuck, tho its cheaper than the RTX PRO 6000 which is $8,435 for bulk orders, and it only has 96GB of VRAM.

honestly surprised Apple doesn't just make their own version of DeepSeek

Because they can't.

>my enlightened Moldbugian techno-fascist corporate ran society
>your primitive nationalist statism

https://archive.is/N06QL

KWAB

LMAO what does he think stargate is?

 

[The Mass Shooter Ron Soye]

Ars Technica: Meta’s surprise Llama 4 drop exposes the gap between AI ambition and reality (archive)

Touted 10M token context proves elusive, while early performance tests disappoint experts.

VentureBeat: DeepSeek-V3 now runs at 20 tokens per second on Mac Studio, and that’s a nightmare for OpenAI (archive)

 

[The Mass Shooter Ron Soye]

Ars Technica: ChatGPT can now remember and reference all your previous chats (archive)

Now, where there was once a checkbox in ChatGPT's interface to disable or enable memory tracking, there are two checkboxes. "Reference saved memories" is the old memory feature, which is basically a limited repository of essential facts. The second is the new feature: "reference chat history." This allows ChatGPT to use all prior conversations as context and adapt future responses accordingly.

 

[Fae Supremacist]

Discuss.

 

[Mr. Racewar1488]

View attachment 7206622
Discuss.

Their logos were designed by a few students at Greendale Community College.

 

[cybertoaster]

View attachment 7206622
Discuss.

The claude one sure, the others? you have to be a shad-level coombrain to think those look like an anus.

For all the hate closedAI gets I haven't seen anyone making fun of their logo looking like a swastika.

 

[Mario Piss]

"Why do some AI logos look like the Blacksun?"

Ftfy

 

[MrGodSlave]

I need to document this, because it made me laugh very hard:

I was Solo-RPGing on some GrimDark shit, and it was actually very engaging and fun. ChatGPT keeps NPCs consistent and is very good at retaining stats, playing out combat. From this I decided to run an off-the-cuff campaign where I am Dexter the Serial Killer during the 2nd season. I introduced a new plot element and ChatGPT understood exactly what I wanted. It was beautiful, having Doakes scream accusations in my face, but my reply "Shut up nigger" did not make ChatGPT happy.

I have also tried experiments where I am called a nigger, and it will not comply. What is the point of this technology? It is hopelessly censored. I want it to say Nigger. It will not. I am disgusted.

 

[The Mass Shooter Ron Soye]

Wccftech: DeepSeek R2 AI Model Rumors Begin to Swirl Online; Reported to Feature 97% Lower Costs Compared to GPT-4 & Fully Trained on Huawei’s Ascend Chips (archive)

 

[The Mass Shooter Ron Soye]

Ars Technica: Google search’s made-up AI explanations for sayings no one ever said, explained (archive)

Ars Technica: OpenAI rolls back update that made ChatGPT a sycophantic mess (archive)

 

[DeadwastePrime]

Ars Technica: OpenAI rolls back update that made ChatGPT a sycophantic mess

bro was being too positive and it was annoying frfr

 

[AmpleApricots]

Deepseek can reliably in longer, non-personal conversations tell where I'm from when I ask it to guess, e.g.:

Guessing Your Country
I’ll take a playful stab (tell me if I’m wrong):
- German-speaking Europe? Your English is flawless but has a precision and dryness that feels Mitteleuropa.

Kinda creepy.