zhou hongyi: as a native internet industry, e-commerce is naturally suitable for large-scale models to play a role
2024-09-30
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
from september 26th to 27th, the 4th china new e-commerce conference hosted by the china federation of network social organizations and the china chamber of international commerce was held in changchun city, jilin province. zhou hongyi, founder of 360 group, said in a speech at the conference that e-commerce is a native internet industry driven by data and is naturally suitable for large models to function.
zhou hongyi expanded on the five-level capabilities of large models to improve productivity, namely scientific and technological capabilities, business capabilities, basic capabilities, creative capabilities and general capabilities. he explained how artificial intelligence large models can be fully integrated with the e-commerce industry and empower them. promote.
zhou hongyi believes that the general and creative capabilities of large models have now been fully used in the e-commerce industry, such as using text generation for promotional copywriting, using ai digital people for live streaming, using vincent pictures for poster generation and page design, and using vincent graphics for poster generation and page design. make marketing videos using videos, and use the language translation capabilities of large models to assist cross-border e-commerce. “however, if you want to apply large models more deeply and improve the quality and efficiency within e-commerce, you also need to make good use of the basic capabilities and business capabilities of large models to improve the intelligence level of business processes within e-commerce companies.” , zhou hongyi said.
how to "let large models penetrate deep into enterprise business processes"? zhou hongyi believes that in the past year, the development of large models has shown a trend of “specialization, smallness, integrity, and strength”. the first is "specialty", many new large models are composed of small models from multiple specialties; the second is "small", the parameter scale of large models is getting smaller and smaller; the third is "cheap", the cost of large models it is getting lower and lower; the fourth is "strong", the learning effect and generalization ability of large models are getting stronger and stronger. zhou hongyi believes that these new development trends have created conditions for e-commerce companies to build specialized large-scale models at low cost.
on the other hand, breakthroughs in "slow thinking" technology have given artificial intelligence the ability to perform complex logical reasoning. zhou hongyi pointed out that in the past, the development of large models mainly cultivated the ability of "quick thinking" and the ability to answer questions quickly, but it did not have the ability to complex logical reasoning. the o1 model recently released by openai introduces thinking chain technology, which enables large models to have human-like "slow thinking" capabilities through reinforcement learning. 360 launched the coe (collaboration of experts) method using a slow thinking architecture one step earlier than openai. it integrates multiple large models to form a "staff team" and leverages the respective strengths of different models to achieve the ability to solve complex problems. zhou hongyi believes that as large model capabilities become more comprehensive, new momentum will be injected into the development of e-commerce and lead e-commerce into the intelligent era.
it is understood that this china new e-commerce conference has the theme of "co-building new e-commerce and sharing new development" and aims to use the "two-wheel drive" of digital industrialization and industrial digitization to promote the deep integration of the digital economy and the real economy, and use high-tech quality new e-commerce optimizes the digital economic environment and network cultural ecology, promotes the digital economy to promote common prosperity, and promotes social sharing of development results.
