Shulou(Shulou.com)11/24 Report--

CTOnews.com, January 9, 2020-after training about 200 billion words and burning tens of millions of dollars, the most powerful AI model in history, "generated training Transformation Model 3" (GPT-3), became an instant hit.

The language AI model created by OpenAI is like an omnipotent player, with a painting style you can't imagine, and no copywriting that it can't output. It can not only create literature, act as a translator, but also write its own computer code. Any layman can use this model and provide an example in a few minutes to get the desired text output.

Researchers at the University of California, Los Angeles found that GPT-3, an autoregressive language model, outperformed ordinary college students in a series of reasoning tests that measure intelligence, according to Xinhua.

The program uses deep learning to generate text similar to human language. GPT-3 has many uses, including language translation and text generation for chatbots. It has 175 billion parameters and is one of the largest and most powerful artificial intelligence models for language processing.

It suddenly occurred to CTOnews.com that OpenAI's ChatGPT seems to have achieved a similar effect. Although it is still based on GGPT-3, this model is known in the industry as "GPT-4". This is also the fourth-generation language model launched by Silicon Valley research laboratories, which poses a threat to search engines, writers, coders, professors and Nickelback around the world.

Of course, according to most experts, compared with the upcoming major version of GPT-4, the current version of ChatGPT can only be said to be an appetizer.

Researchers at the University of California believe that such large-scale language models have reignited the debate about whether humans are better cognitive while providing enough training data. What is particularly interesting is that these models can reason new problems locally without any direct training.

The researchers pointed out that in human cognition, this ability is closely related to analogical reasoning ability, and they made a direct comparison of GPT-3 on a series of analogical tasks, including a new text-based matrix reasoning task closely related to Raven's progressive matrix, and finally found that GPT-3 showed amazing abstract pattern induction ability, matching or even surpassing human ability in most cases.

The end result shows that large language models such as GPT-3 have acquired an "emerging capability" to find zero-sample solutions to a wide range of analogical problems.

Reference: https://doi.org/10.48550/arXiv.2212.09196

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.