Artificial Intelligence thread

[9dashline]

Chinese LLM models appear to be persistently trailing American/Western ones in English favored leader boards like LMSYS, Open LLM, Open Compass, Berkeley Function Call, Can AI Code, etc. Actually, I'm not sure if Open Compass is English favored, but in any case, there's very few Chinese LLM models at the very top of these boards.

By contrast, on Chinese language favored leader boards, Chinese LLMs tend to be more dominant. Which is great and expected.

Wonder if this is a consequence of the sanctions limiting China's access to comparable CPU hardware or if it's a consequence of it just being harder for Chinese researchers to work in English.

Please, Log in or Register to view URLs content!

Consenus is Kling is better than Sora, or at least the version openai demoed a while back

I hope Kling releases the open network weights and forces openai's hands

 

[Overbom]

Here is more details on qwen 2 posted on github
Of course, it’s quite possible that alibaba tweaked its models to do better on the various evaluation platforms. Since it is now made Available on different platforms, we will see if it performs as well as llama

Please, Log in or Register to view URLs content!

Scores lower than Llama 3 on the lmsys (Arena leaderboard). That's for overall, coding, Hard Prompts (overall), Hard Prompts (English) categories

 

[tokenanalyst]

I hope Kling releases the open network weights and forces openai's hands

That ain't happening. What you will likely is see the release of some OS video model by some university and then the slow crawl to parity with commercial models.

 

[tokenanalyst]

Scores lower than Llama 3 on the lmsys (Arena leaderboard). That's for overall, coding, Hard Prompts (overall), Hard Prompts (English) categories

I find the model exceptionally good.
llama-3 has 125K votes vs 12K votes for QWEN2, no enough people testing it, there also some bias because the model compete with llama-3.

 

[9dashline]

Artificial General Intelligence (AGI) is the hypothetical ability of a machine to perform any intellectual task that a human can do. Large language models, such as OpenAI's GPT-3, have demonstrated impressive capabilities in natural language understanding and generation, leading some to speculate about their potential for achieving AGI. However, despite their notable achievements, large language models have inherent limitations that make them unlikely candidates for true AGI. This essay will discuss these limitations, including their lack of reasoning abilities, reliance on massive amounts of data, absence of common sense understanding, and ethical concerns.

A key aspect of AGI is the ability to engage in complex reasoning and problem-solving. While large language models can generate coherent text and answer questions based on the patterns they have learned from their training data, they lack the ability to engage in deductive or inductive reasoning that is essential for AGI. This is because these models primarily rely on pattern matching and statistical associations rather than understanding the underlying logic or principles behind the information they process. Consequently, they are prone to making errors when faced with novel situations or questions that require logical reasoning.

Large language models depend on vast amounts of data for their training, which presents several challenges for achieving AGI. First, the need for extensive data limits the applicability of these models in domains where data is scarce or expensive to acquire. Second, the sheer scale of computational resources required for training large models raises questions about their efficiency and ecological impact. In contrast, humans can learn and generalize from a relatively small number of examples, highlighting the difference between the learning mechanisms of large language models and true AGI.

Common sense understanding is a fundamental aspect of human intelligence, allowing us to make inferences and predictions about the world based on our background knowledge. However, large language models often lack this basic understanding. Despite being trained on vast amounts of text, these models still make mistakes that a human with common sense would not. This is partly because large language models learn from text data alone, which may not fully capture the richness of human experience and understanding. True AGI would require the integration of various types of knowledge, including visual, auditory, and tactile, as well as an understanding of the underlying structure of the world.

While large language models have undoubtedly advanced the field of AI and demonstrated impressive natural language capabilities, they fall short of achieving true AGI due to their lack of reasoning abilities, reliance on massive amounts of data, absence of common sense understanding, and ethical concerns. To reach AGI, researchers must explore alternative approaches that move beyond the limitations of current large language models, incorporating reasoning, efficient learning mechanisms, and a more comprehensive understanding of the world. Addressing the ethical challenges associated with AI development is also crucial to ensure that AGI benefits all of humanity and avoids causing harm.

Please, Log in or Register to view URLs content!

Implications of this are huge, imho this is US plan to extent the dollar hegemony forever...

(Musk is CIA from the getgo he is just putting on a fake show, so read beyond the headlines)

Overnight half the world will be using US based openai AI.... every apple device and iphone will have the latest chatgpt (as API) baked into the iOS....

Each and every one of these devices will be spying for Uncle Sam, and worst yet, this one move will ensure as cloud intelligence gets baked into every layer of every product and every service, that the world will be beholden to US AI services and this AI ecosystem will be gate kept on condition of using US dollar....

First gold, then oil, now AI is the final frontier of USD extension.....

With Apple and Nvidia as hardware, Google Play and Andriod, and Azure and OpenAI as the cloud and intelligence layer, this is the US plan to turn the tide around and make good on its PNAC plan that it enacted at the turn the century circa 2001

The anti Huawei campaign, the TSMC thing and EUV ban, its all come down to this moment. I feel like US is going all in and even jumping the shark on AI because its America's last real hope and best objective bet at beating China in the great power competition

Just like with Nvidia GPU once gamers jumped to DLSS and game devs baked that into their games, there was no going back and AMD cards never held a chance, the DLSS API held everyone captive... likewise US is attempting in one fell swoop to do the same but with openai and azure as the AI and cloud layer, getting the rest of the world onto its platform so it can start collecting usury tax yet once again, all spearheaded by the move of Apple baking in openai to its iOS....

Perhaps China's best bet is to rug pull while it still can and do AR sooner rather than later, maybe before end of 2024

 

[gabriel.shenton]

Please, Log in or Register to view URLs content!

Implications of this are huge, imho this is US plan to extent the dollar hegemony forever...

(Musk is CIA from the getgo he is just putting on a fake show, so read beyond the headlines)

Overnight half the world will be using US based openai AI.... every apple device and iphone will have the latest chatgpt (as API) baked into the iOS....

Each and every one of these devices will be spying for Uncle Sam, and worst yet, this one move will ensure as cloud intelligence gets baked into every layer of every product and every service, that the world will be beholden to US AI services and this AI ecosystem will be gate kept on condition of using US dollar....

First gold, then oil, now AI is the final frontier of USD extension.....

With Apple and Nvidia as hardware, Google Play and Andriod, and Azure and OpenAI as the cloud and intelligence layer, this is the US plan to turn the tide around and make good on its PNAC plan that it enacted at the turn the century circa 2001

The anti Huawei campaign, the TSMC thing and EUV ban, its all come down to this moment. I feel like US is going all in and even jumping the shark on AI because its America's last real hope and best objective bet at beating China in the great power competition

Just like with Nvidia GPU once gamers jumped to DLSS and game devs baked that into their games, there was no going back and AMD cards never held a chance, the DLSS API held everyone captive... likewise US is attempting in one fell swoop to do the same but with openai and azure as the AI and cloud layer, getting the rest of the world onto its platform so it can start collecting usury tax yet once again, all spearheaded by the move of Apple baking in openai to its iOS....

Perhaps China's best bet is to rug pull while it still can and do AR sooner rather than later, maybe before end of 2024

truly amazing explanation. The current "AI" gold rush looks really strange, language models do not have intelligence, the speed and strength of the AI push is so strong, one might suspect deep state involvement.

now everything makes sense, "AI" ecosystem will be enhancing USD for the next few decades, the people who came up with the reserve currency system was the true genius, the best business model in the world, collecting taxes constantly day in and day out.

Without EUV, it's hard for China to catch up as time pass by.

 

[9dashline]

truly amazing explanation. The current "AI" gold rush looks really strange, language models do not have intelligence, the speed and strength of the AI push is so strong, one might suspect deep state involvement.

now everything makes sense, "AI" ecosystem will be enhancing USD for the next few decades, the people who came up with the reserve currency system was the true genius, the best business model in the world, collecting taxes constantly day in and day out.

Without EUV, it's hard for China to catch up as time pass by.

AI is rapidly becoming embedded in every layer of technology, a trend driven by U.S. efforts to secure global dominance. Microsoft's latest updates to Windows 11 and Azure, featuring the Copilot option—a generative AI based on large-language models—illustrate this shift. Not only are everyday applications like Notepad integrated with AI, but this trend extends to the Internet of Things, with AI-enabled devices ranging from refrigerators to smartwatches making API calls to services like OpenAI, leveraging Azure Cloud and NVIDIA GPUs.

This acceleration in AI deployment reflects a broader strategy. The U.S. government appears to be pushing AI as a "winner-takes-all" race, aiming to be the first to capture the flag and secure a long-term competitive advantage. By embedding AI deeply into every product and service, the U.S. positions itself at the top of a global technological pyramid. This push ensures that other nations, driven by market forces, must adopt these technologies or risk falling behind, thereby increasing their dependence on U.S. innovation.

AI is becoming the great equalizer and intellectual multiplier, automating tasks that once required human labor. This transformation allows countries to cut costs and streamline operations, yet it also consolidates power within governments, making populations more reliant on state-controlled technologies. In essence, AI pits the workforce of each country against an insurmountable competition, reinforcing governmental control.

As AI becomes ubiquitous, the financial benefits flow back to the U.S. Nations worldwide will pay for AI services in dollars, further entrenching U.S. economic hegemony. This strategy ensures that while other nations adopt and integrate AI, they do so under the umbrella of American technological supremacy, maintaining the U.S. at the pinnacle of global power.

In summary, the global integration of AI is not just a technological shift but a strategic maneuver by the U.S. to cement its dominance. As AI continues to evolve, its pervasive influence will reshape international relations and economic dependencies, reinforcing America's position at the top.

 

[9dashline]

While the U.S. appears poised to dominate the global AI landscape, relying heavily on the latest advancements from companies like NVIDIA, there are significant geopolitical vulnerabilities that could disrupt this trajectory. One of the critical points of concern is the dependency on Taiwan Semiconductor Manufacturing Company (TSMC) for NVIDIA's GPUs, which are essential for AI technologies.

TSMC is based in Taiwan, just 90 miles off the coast of China. Given the long-standing tensions and China's ongoing claims over Taiwan, any escalation could have severe consequences. If China were to force reunification with Taiwan, either through economic pressure or military force, TSMC's operations would be jeopardized. This risk is particularly concerning as the Arizona fabs, intended to diversify and secure chip manufacturing for the U.S., have yet to come online.

China has been vocal about its intentions towards Taiwan for over 70 years, and recent provocations in the Philippines add to the regional instability. Should China decide to act, it would be a strategic move to disrupt the U.S. AI strategy. This so-called "rug pull" would not only impact NVIDIA but also other tech giants like Apple, which rely on TSMC for their chips.

A conflict that takes TSMC out of commission, whether through direct control by mainland China or through damage during conflict, would create a massive supply chain shock. The disruption would extend beyond GPUs to include smartphones, batteries, and other tech products. Rare earth elements, crucial for many of these technologies, are also predominantly sourced from China, further exacerbating the potential impact.

In a scenario where the U.S. and China engage in a conventional war, the ramifications would be global. A scorched-earth approach could cripple both economies, but the immediate effect would be a significant delay in the U.S. AI strategy. The U.S. relies on a continuous supply of advanced chips to maintain its technological edge and economic hegemony. Without TSMC's capacity, the pace of AI development would slow dramatically, undermining the U.S.'s ability to leverage AI for sustained dominance.

In conclusion, while the U.S. aims to lead the AI revolution, this strategy is precariously balanced on the stability of TSMC and the geopolitical landscape of the Taiwan Strait. Any conflict involving Taiwan could derail the U.S. plans, highlighting the critical need for diversifying supply chains and securing domestic chip manufacturing capabilities to mitigate such risks.

 

[bebops]

Isn't TSMC and others are rushing to build new plants all over the world? In a few years, an attack on Taiwan will not disrupt the global supply flow of chips.

 

[9dashline]

Isn't TSMC and others are rushing to build new plants all over the world? In a few years, an attack on Taiwan will not disrupt the global supply flow of chips.

OpenAI plans to release gpt5 in november, which they claim is AGI level.

Apple just signed a deal with them to bake openai into every Apple device at os layer

There may not even be a few years left for China to act.... its clear US at the very highest level is pushing this like there is no tomorrow... a sort of Manhattan and Apollo project combined...

Currently America is still deindustrialized, once AI takes over, it will reindustrialize rapidly... this is beyond just chips and EUV but broadly all supply chains

China needs to rugpull on USD, and supply chains in general and tsmc in specific etc etc before that point of inflection can be allowed to happen