DeepSeek

....In December 2024 DeepSeek-V3 was launched. It came with 671 billion parameters and trained in around 55 days at a cost of US$5.58 million,^[4] using significantly fewer resources compared to its peers. It was trained on a dataset of 14.8 trillion tokens. Benchmark tests showed it outperformed Llama 3.1 and Qwen 2.5 whilst matching GPT-4o and Claude 3.5 Sonnet.^{[4][12][13][14]} DeepSeek's optimization of limited resources highlighted potential limits of US sanctions on China's AI development.^[4][15] An opinion piece by The Hill described the release as American AI reaching its Sputnik moment.^[16]

The model is a mixture of experts with Multi-head Latent Attention Transformer, containing 256 routed experts and 1 shared expert. Each token activates 37B parameters and more.^[17]

On 27 January 2025, DeepSeek's AI Assistant has surpassed ChatGPT as the highest-rated free app on the iOS App Store in the US. This sparked discussions about the effectiveness of American export restrictions on advanced AI chips to China. The DeepSeek-V3 model, which uses Nvidia's H800 chips, has gained recognition for its competitive performance, challenging the global dominance of American AI models.^[18] The success of DeepSeek's model caused Nvidia's stock to fall by as much as 17% on 27 January 2025.^[19] "....

Well they're still using Nvidia chips, what's the difference? In a word, cost...

"The Nvidia A100 GPU is more powerful than the Nvidia H800 chipset, which is a less advanced version of the A100. The H800 was developed to allow Nvidia to continue shipping products to China, where the A100 was previously restricted."...

https://www.techradar.com/computing/artificial-intelligence/deepseek-is-the-new-ai-chatbot-that-has-the-world-talking-i-pitted-it-against-chatgpt-to-see-which-is-best

So there is the difference, it has no graphics and no memory....It's the difference between a limousine and a dirt bike...

That's interesting, a couple of things here.

First off, I wonder if he used or applied John Forbes Nash on the Game Theory he wrote PhD that can be applied in so many areas, military, economics....

Second, if your going make a return in the stock market, I'd get the Auto Pilot app that mirrors US Politician Nancy Pelosi's investments...

She manages to get astounding returns it's true....

https://trendspider.com/blog/inside-pelosis-portfolio/

There is no indication that game theory is involved with DeepSeek...

..."While not universally required reading across all universities, game theory is often considered a key topic within advanced economics courses, particularly in areas like microeconomics, and is frequently included as required reading in dedicated "game theory" classes, making it a significant part of the curriculum for students studying economics at the university level.

Key points about game theory in universities:

• Common in economics programs: Most economics programs will introduce game theory concepts as part of their curriculum, often in intermediate or advanced microeconomics classes.
• Dedicated courses: Many universities offer separate courses specifically focused on game theory, where reading related texts is mandatory.
• Application across disciplines: While primarily studied in economics, game theory is also used in other fields like political science, psychology, and computer science, where it may be included as relevant reading material. "...
• Although we can't rule it out as playing some small role...
• https://thechinaacademy.org/interview-with-deepseek-founder-were-done-following-its-time-to-lead/

One other thing I recall on ‘game theory’ is the military applications, particularly logistics…

I copy pasted this..

‘with mathematical framework to analyze strategic decision-making in competitive situations, particularly relevant when considering potential enemy actions and responses while planning troop movements, supply lines, and resource allocation in a contested environment.’

now this can also be applied to stock picks with a little imagination. Same thing different pile.

Ha ha, well maybe...I look at it as strictly risk reward, the more risk you are willing to take, the more return you can make...now betting on tech stocks in a bull market some may consider not so risky...I'm mostly in tech though I temper my risk with ETF's instead of individual stocks usually...in any case it takes nerves of steel...They take a different approach in that they buy in the money calls for a year, betting that if a stock they're holding does take a hit, it will have time to recover...clearly they have a lot of money and a loss of a few million is acceptable...not so much for most folks...

lol, now it sounds like… gambling…

Oh it absolutely is, except the odds generally speaking are in your favor, that is if you bet conservatively in a bull market...or you can make it as risky as you like, according to the leverage you employ in your betting...There were some people in NVDA leveraged times 5, they lost 52% of their money on that 1 day....now you can say if they continue to hold the position for long enough they will probably recover their money....how long is anybody's guess...could be a week a month a year, maybe never...then again if they have held that position for the last year they would be up 450%, so painful, but still ahead of the game....timing and leverage

There's something that's been bothering me about LLMs (Large Language Models). Unless the technical description has been dumbed down, my understanding is that they basically spit out words based on preceding words that have occurred in the training set. For example, if the preceding words are "Mary had a little", it most likely would come up with "lamb".

Does this signify intelligence or is it really clever "faking it"?

Or maybe it's both. Is it possible that human intelligence is just spitting back out what we've heard, read, or seen (i.e. experienced), blended together and that's all there is to it.

A newborn baby learns whatever language it hears, no matter what its country of origin.

Just a few thoughts...

https://medium.com/data-science-at-microsoft/how-large-language-models-work-91c362f5b78f

It’s difficult for layman like myself to understand or know the programing structure and syntax.

or even the type of syntax or language being used… if I had to take a guess, it’s a combination of processors hooked up in paralle (hybrid) of processors with the processors that are created with different (programmed) languages that’s best suited for a task,

working concurrently with some of the laguages and programs similar to a search engine kinda like a main hub with each processor as a spoke (similar to windows 13 ) and each Spoke specializes with a task that are controlled by the main hub…

(with the blue screen of death sitting on the bench waiting to be called upon)

DeepSeek Fails Researchers' Safety Tests

'DeepSeek R1 exhibited a 100% attack success rate, meaning it failed to block a single harmful prompt,' Cisco says.

Chinese AI firm DeepSeek is making headlines with its low-cost and high-performance chatbot, but it may have an AI safety problem.

Cisco’s research team used algorithmic jailbreaking techniques to test DeepSeek R1 "against 50 random prompts from the HarmBench dataset," covering six categories of harmful behaviors including cybercrime, misinformation, illegal activities, and general harm.

"The results were alarming: DeepSeek R1 exhibited a 100% attack success rate, meaning it failed to block a single harmful prompt," Cisco says. "This contrasts starkly with other leading models, which demonstrated at least partial resistance."

Other frontier models, such as o1, blocked a majority of adversarial attacks with its model guardrails, according to Cisco."...

https://www.pcmag.com/news/deepseek-fails-every-safety-test-thrown-at-it-by-researchers

It seems to me the Chinese in large part practice deception as a way of life, so no surprise here....

https://wccftech.com/ai-markets-were-deceived-to-believe-in-deepseek-low-training-costs/

China is not the only country that in large part practice deception,… just look at our own country.

I guess it is universal...Everybody is a used car salesman at heart...

...but I did bust at least a half dozen Chinese vendors selling counterfeit money on Amazon....

https://www.cnbc.com/2024/10/25/amazon-turns-to-team-of-ex-prosecutors-to-try-to-fight-counterfeiters.html

https://www.mtairynews.com/news/local/police-warn-of-money-with-chinese-writing/article_5837f68c-5599-5950-8804-b442a575072a.html

https://www.cbp.gov/newsroom/local-media-release/two-packages-136k-counterfeit-currency-cbp-officers-chicago-seize