Technology

The ideologues of Silicon Valley are in model collapse.

To train an AI model, you need to give it a ton of data, and the quality of output from the model depends upon whether that data is any good. A risk AI models face, especially as AI-generated output makes up a larger share of what’s published online, is “model collapse”: the rapid degradation that results from AI models being trained on the output of AI models. Essentially, the AI is primarily talking to, and learning from, itself, and this creates a self-reinforcing cascade of bad thinking.

We’ve been watching something similar happen, in real time, with the Elon Musks, Marc Andreessens, Peter Thiels, and other chronically online Silicon Valley representatives of far-right ideology. It’s not just that they have bad values that are leading to bad politics. They also seem to be talking themselves into believing nonsense at an increasing rate. The world they seem to believe exists, and which they’re reacting and warning against, bears less and less resemblance to the actual world, and instead represents an imagined lore they’ve gotten themselves lost in.

Let’s continue once again our magical trip into GUI Wonderland, and leave timesharing behind with the extremely capable Three Rivers / ICL PERQ and its stunning graphical capabilities! If you want to come with us, feel free to check this article out!

Disponibile anche in 🇮🇹

cross-posted from: https://lemmygrad.ml/post/5443422

Thoughts?

The Chinese expert community is focused on competition between China and the USA for AI domination. China's local AI companies in contrast to country's authorities are skeptical towards achieving global AI leadership by 2030 amid the American sanctions, lack of state support, variety of technologies and lack of staff. Many of the Western countries acknowledged China's lagging from the developed world.

The recent article of China's minister of science and technology Yin Hejun rates highly the AI achievements of China. It mentioned the volume of local AI market in 2023 was 500 billion yuan, the number of AI companies is above 4400, among these companies 108 with capitalization over 1 billion US dollars. About 500 institutes of high education introduced programs to study AI.

The article highlights the 2017 plan to develop the next generation AI to make China the global AI leader by 2030.

But according to the unanimous opinion of local experts, China is still behind the US. Americans are almost monopolized the global AI market by bringing Open AI's advanced ChatGpt to the masses. Fast development of this tool became possible due to the feedback from users, including China where this service is officially banned, and this made the gap between America and China even bigger.

China's local companies responded with creation of series of competitive LLM products that are the basis for generative instruments which are used at least for now only on China's local market. At the moment the technological giants (Baidu, Alibaba, Iflytek) achieved only the level of ChatGpt 3.5 and plan within the year to surpass the actual ChatGpt 4.

However, the Americans also keep pushing and taking leadership on the market of generative AI. OpenAI's text to video Sora was presented in February and has no Chinese analogs and repeats the ChatGpt history.

According to the representatives of IT companies, the pace of AI development in China is on hold by a number of factors. In March the head of Iflytek Liu Qingfeng pointed that technological gap between China and America is not decreasing because of the US sanctions that limited access to the global LLM achievements. He believes the solution is promoting the AI by relying on local projects. As example he mentioned the LLM model of his company Spark that got a push to development after the Western sanctions and by making a contract with another sanctioned Chinese company Huawei.

The head of China Electronic Corporation Zeng Yi also said that the lack of state support of local developers, start-ups and IT companies in the AI race makes the gap between China and the West even bigger. The businessman asks the developers to cooperate with the local producers since the Western technologies may at any moment lose functionality and support.

The representatives of the local IT business believe the uncontrolled development of national hardware base requires too much resources and hinders the pace of AI development. The chairman of JD.com's Technology Committee Peng Cao notes that scattered technical chip solutions make it hard to develop software. This way local companies should consolidate their efforts to create AI platforms and lower their amounts.

At the same time, China's developers are still highly dependent on Western software solutions. This March the analysts of Beijing Academy of Artificial Intelligence marked in their report to China's State Council Li Qiang the problem of high popularity of Western tools with free access. According to their data the biggest part of local AI developers uses 2023 LLAMA model of American company META, which owns Facebook and Instagram and is among global AI leaders, for their own LLMs.

According to Xiaomi founder Lei Jun the key flaws of Chinese AI ecosystems are lack and turnover of staff. Only 40% represent the qualified personnel while the graduates prefer West to build careers. For instance, in Sora two of thirteen employees are from China. Xiaomi believes the only way to achieve the ambitious aim of overtaking the US and becoming the global AI leader is tight cooperation of state and business to bring back the best and young professionals to work on the breakthrough projects.