Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

The Turing Award winner LeCun:ChatGPT has great limitations, and the life span of the autoregressive model is no more than 5 years.

2025-04-05 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >

Share

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

Turing Award winner Yann LeCun talks about AI: the future is open source.

The first half of this year is the most magnificent half of the AI session.

Behind the rapid development of all kinds of GPT and even the embryonic form of AGI, there are two camps of people with different views.

One school believes that the generative AI led by ChatGPT is very powerful and can lead to a big wave of revolutionary waves, so there is no problem to continue to push forward.

The other side thinks that we are moving a little too fast. If you don't say no, you have to stop for a while. And morally, there are too few matching constraints.

As a master of AI, Yann LeCun has a different view on this.

Completely forbidden? Infeasible LeCun said he was not surprised by ChatGPT's performance and did not approve of suspending artificial intelligence research.

"this may produce the opposite result as expected. "

He said that artificial intelligence, as an amplifier of human intelligence, may be the origin of the new Renaissance.

The big language model of ChatGPT is autoregressive. AI is trained to extract words from a corpus of up to 1.4 trillion words to predict the last word in a given sentence sequence, the next word that must appear.

The relevant research carried out by Claude Shannon in the 1950s is based on this principle.

The principle remains the same, but it becomes the size of the corpus and the computing power of the model itself.

LeCun said, "at present, we can not rely on such models to grow and coherent text, these systems are not controllable." For example, we can't directly ask ChatGPT to generate a text that targets a 13-year-old child.

Second, the text generated by ChatGPT is not 100% reliable as a source of information. The function of GPT is more like an auxiliary tool. Just like the existing driving assistance system, you have to keep the steering wheel on autopilot.

Moreover, the life of the autoregressive language models we know today is very short, five years is a cycle, and five years later, no one will use the model of the past.

The focus of our research should be on finding a way that these models can be controlled. In other words, the AI we are going to study is an AI that can reason and plan according to a given goal, and the standards that can ensure its security and reliability are consistent. This kind of AI can feel emotion. "

You know, a large part of human emotion has to do with the achievement of goals, that is, with some form of expectation.

With such a controllable model, we can generate long and coherent text.

LeCun's idea is to design enhanced models in the future that can mix data from different tools, such as calculators or search engines.

Models like ChatGPT are only text-trained, so ChatGPT's understanding of the real world is incomplete. If we want to develop further on this basis, we need to learn something related to the sensory perception and structure of the whole world.

These more complex content can not be achieved simply by reading the text, which is one of the biggest challenges in the next few years.

Open source is the end of the desire for power, which is unique to human beings. And AI doesn't have this desire just because it gets stronger and stronger.

Only the human species knows how to make laws to ensure that individual actions do not harm the common interest too much.

OpenAI, which started as an open research project, has now been closed. OpenAI said nothing about his work, and reversing the situation is not common in the field of research.

The problem is that training a language model is expensive and costs tens of millions of euros, so startups can't afford it.

This is also the main reason for the merger of Microsoft and OpenAI, which requires the group's common computing power to improve its future model. That's why DeepMind and Google brain finally merged.

LeCun said that eventually, in terms of the market, developers will move towards the common ecology of an open platform. It would be bad if only a small number of companies control this kind of technology.

Historically, both Facebook and the renamed Meta have been actively promoting open basic research, such as LlaMa, an open source project.

In the early 1990s, Sun Microsystems and Microsoft fought for the right to operate the server. Keep in mind that all Internet technologies that can always gain a foothold are open source.

LeCun concluded by saying that at present, the key to preventing such open and open source AI platforms is a legal issue. Such a similar open source platform is essential if the European Union wants to promote the structure of the AI industry in the future.

Reference:

Https://twitter.com/USBEKetRICA/status/1648597311843450881

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

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.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

IT Information

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report