China just wrecked all of American AI. Silicon Valley is in shambles.

Apollo Creed

Look at your face
Supporter
Joined
Feb 20, 2014
Messages
56,721
Reputation
13,583
Daps
213,095
Reppin
Handsome Boyz Ent
I don't understand any of this but I do know that the media loves sensationalizing shyt. It may very well turn out that this DeepSeek shyt is a temu chat-gpt and everything goes back to normal.

For my line of work I'd be interested in a free image generator that rivals MidJourney, or a free video generator (probably a fantasy though)

China has wild work culture that people may be ignoring when it comes to them working to building something and cheaper from the human labor stand point. That said if USA got 500 Billy to toss around then we should have something insanely better like yesterday
 

bnew

Veteran
Joined
Nov 1, 2015
Messages
60,074
Reputation
8,966
Daps
165,993




AI research team claims to reproduce DeepSeek core technologies for $30 — relatively small R1-Zero model has remarkable problem-solving abilities​


News

By Jowi Morales

published 11 hours ago

It's cheap and powerful.


The DeepSeek logo against a hexagonal textured background


(Image credit: DeepSeek)

An AI research team from the University of California, Berkeley, led by Ph.D. candidate Jiayi Pan, claims to have reproduced DeepSeek R1-Zero’s core technologies for just $30, showing how advanced models could be implemented affordably. According to Jiayi Pan on Nitter, their team reproduced DeepSeek R1-Zero in the Countdown game, and the small language model, with its 3 billion parameters, developed self-verification and search abilities through reinforcement learning.

Pan says they started with a base language model, prompt, and a ground-truth reward. From there, the team ran reinforcement learning based on the Countdown game. This game is based on a British game show of the same name, where, in one segment, players are tasked to find a random target number from a group of other numbers assigned to them using basic arithmetic.

The team said their model started with dummy outputs but eventually developed tactics like revision and search to find the correct answer. One example showed the model proposing an answer, verifying whether it was right, and revising it through several iterations until it found the correct solution.

Aside from Countdown, Pan also tried multiplication on their model, and it used a different technique to solve the equation. It broke down the problem using the distributive property of multiplication (much in the same way as some of us would do when multiplying large numbers mentally) and then solved it step-by-step.

Image 1 of 2

BYzFEMp.jpeg


(Image credit: Jiayi Pan / nitter)

Xw10vD6.jpeg


(Image credit: Jiayi Pan / nitter)

The Berkeley team experimented with different bases with their model based on the DeepSeek R1-Zero—they started with one that only had 500 million parameters, where the model would only guess a possible solution and then stop, no matter if it found the correct answer or not. However, they started getting results where the models learned different techniques to achieve higher scores when they used a base with 1.5 billion parameters. Higher parameters (3 to 7 billion) led to the model finding the correct answer in fewer steps.

But what’s more impressive is that the Berkeley team claims it only cost around $30 to accomplish this. Currently, OpenAI’s o1 APIs cost $15 per million input tokens—more than 27 times pricier than DeepSeek-R1’s $0.55 per million input tokens. Pan says this project aims to make emerging reinforcement learning scaling research more accessible, especially with its low costs.

However, machine learning expert Nathan Lambert is disputing DeepSeek’s actual cost, saying that its reported $5 million cost for training its 671 billion LLM does not show the full picture. Other costs like research personnel, infrastructure, and electricity aren’t seemingly included in the computation, with Lambert estimating DeepSeek AI’s annual operating costs to be between $500 million and more than $1 billion. Nevertheless, this is still an achievement, especially as competing American AI models are spending $10 billion annually on their AI efforts.
 

Scustin Bieburr

Baby baybee baybee UUUGH
Joined
May 3, 2012
Messages
21,698
Reputation
11,227
Daps
125,220
Facts, China isn’t perfect but Communists already aren’t that fond of Capitalist oligarchies. Worse when it’s an oligarchy that’s actively attempting to sabotage you in order to maintain subservience and record profits for their shareholders.

If we don’t stop this shyt from within we’re going to be viewed as an existential threat to the rest of the planet. We already are an existential threat to every living creature and natural mechanisms that have been developed over millions of years. The rest of the world just had to catch up and it’s beginning to happen.

If the world collectively slid on us it would be the least shocking thing ever.

Throughout history the country who spends the most on their military uses it to wage war on others and attempt to control them. Eventually they run into a force too powerful for even their precious military to overcome and they're destroyed.

It happened to Rome. It happened to France. It happened to Germany. It happened to Mongolia. It happened to Japan. It happened to England.

America is next. WW3 will be started by the USA and the whole world will have to set aside their differences and put our leaders down for good. Canada, the EU, China, they're all talking more about investing in their military and starting to earmark funds. America is not a peace loving country and eventually trump or one of his fans with power will leave nato then stupidly make a move on a nato country like Canada or will back up Russia as it attacks a NATO country and then it's on.

Now before anyone talks about nukes, let us keep in mind that nukes are controlled by computer systems. If you shut down the system, the missile never leaves the silo. If you shut down the electronics in a nuclear sub, the warhead never leaves the watercraft. If you can scramble the systems of a bomber with nuclear ordinance, it falls into the ocean before it can get to your countries. If you develop ai that can reroute missiles, your weapon will literally backfire and destroy the launch site.

If anyone doesn't believe that countries are working on countermeasures right now and the focus on AI and computing is a key part of that strategy, I'd like to live in your world because that shyt seems blissful.
 

Micky Mikey

Veteran
Supporter
Joined
Sep 27, 2013
Messages
16,332
Reputation
3,119
Daps
92,443
How long do we have to wait to get a relatively inexpensive and open source model that'll be on par with o3?
 

morris

Superstar
Joined
Oct 8, 2014
Messages
16,760
Reputation
5,036
Daps
36,860
You do realize the term “public investment” refers to projects intended for civilian use right?

Lol

I guess Iraq and Afghanistan were public investment projects?
you make a great point but it looks like public funding started it via ARPA (I'm no history buff and had to look it up)

Yes, the internet was essentially invented through public funding, specifically from the U.S. Department of Defense's Advanced Research Projects Agency (ARPA), which funded the development of the ARPANET, the precursor to the modern internet, in the late 1960s.
Key points about the public funding of the internet:

ARPANET:
The initial network that became the foundation of the internet was called ARPANET, which was funded by ARPA.
Military origins:
The initial motivation for developing the internet was to create a robust communication network for military use in case of a nuclear attack.
Research and development:
The government funding allowed researchers to develop the core protocols and technologies that underpin the internet, like TCP/IP.
 

IIVI

Superstar
Joined
Mar 11, 2022
Messages
12,467
Reputation
2,998
Daps
43,586
Reppin
Los Angeles
Ok, after a lot of experimentation and using it to do some tasks.

It’s 100% legit.

Like other have mentioned, the fact you can use it all day while the PAID ChatGPT o1 version caps out per day makes DeepSeek annihilate it. I even straight up chose it over using ChatGPT because I like how it shares its reasoning. I’ll be cancelling my ChatGPT until they make a better version.
 
Last edited:

Michael's Black Son

Blanket Jackson
Supporter
Joined
Sep 30, 2013
Messages
53,135
Reputation
15,575
Daps
233,241
Reppin
New York City & Neverland Ranch
Oh now it's gonna be a race to the bottom. Them get rich off of AI schemes got exposed :russ:

These random AI companies are the 2025 shīt coins at this point.

Next thing you know someone will say they can beat Deepseek with a Speak and Spell
 
Last edited:
Top