SPEECH: The future is the large-scale model system centered on the conversational language computing large-scale model!

On June 26, 2023, Yu Kai, the co-founder and chief scientist of Aspire, delivered a keynote speech "Conversational General Artificial Intelligence and Specialized Language Models" at the Fifth Global Intelligence Expo. He said that the large model is the foundation of artificial intelligence. In the new era, large language models and conversational language models are the core of further breakthroughs in the entire artificial intelligence large model. The specialized large language model is the core of the real general artificial intelligence development. After the meeting, Yu Kai, the co-founder of Aspire, accepted an exclusive interview with CSDN and shared topics related to the current big model.

 

CSDN: Because you mentioned in your speech that the DFM-2 large model that will be released next month, can you introduce it to us?

Yu Kai: In 2021, our Jiaotong University laboratory released the first generation of DFM. The English translation is the dialogue base model. In fact, we also use a general generative architecture. Now, we are the second generation of DFM. Let’s talk about the name first. DF stands for Dongfeng, and it is also related to my country’s "Dongfeng No. 1". It is also larger in scale than the first generation. We also hope that he can become an artificial The core strength in the field of intelligence, the specific details will be discussed in detail at our press conference next month, and please look forward to our second generation of DFM.

CSDN: In terms of future planning and strategy, under the trend of artificial intelligence, does Aspire have any other plans?

Yu Kai: In the whole wave of artificial intelligence, general-purpose capabilities have already come up. The most important strategic view of Aspire is to use general-purpose capabilities to conduct dedicated dialogues. The first block, we want to focus on it with its traction. To shape the artificial intelligence industry, we emphasize conversational artificial intelligence. The second most important thing is that these things can fall into our daily life. It is a major trend to upgrade connected smart hardware to language-interactive information hardware. The third thing is the digital assistant in the vertical field. Aspire will also use the assistant to reshape this part of scientific research.

CSDN: In your speech, you also mentioned that both artificial intelligence and large models are facing some challenges. In the face of this ever-changing market demand and technology trends, how does Aspire respond?

Yu Kai: One of our very important capabilities is called model adaptation. The general model is a must. We first need to have the best general model technology in the world, but the general model cannot solve all problems. It can solve many To C problems. , because Aspire is mainly To B, highly customized is still difficult to solve, so we Aspire to do model migration, rapid migration, tool migration, software and hardware integration to do migration, to deal with various personalized problems , we need the ability to personalize the model, which is technology, algorithm, engineering technology platform, and tool architecture. We have developed a series of related technologies for this.

CSDN: At present, many manufacturers are making large-scale models. What are the advantages of our large-scale models?

Yu Kai: The first one is more distinctive. The tens of billions of scale models that Aspire is now making, use it first and talk later. It is our characteristic to start with the end in mind. The second feature is that Spies has always believed that the two concepts of conversational artificial intelligence and large models are indispensable. The new opportunities brought about by large models must be in the field of conversational language computing. These things and The combination of large models back and forth is a great feature of Spies, because language is the center of all human scheduling. It is our special idea to emphasize conversational language computing, so I don’t think large models are the future. I think the large-scale model system with the core of the conversational language computing large-scale model is the future.

CSDN: For the large model released this time, is there any part that can be disclosed to us in the industry application?

Yu Kai: Aspire is working on two major issues. The first is that smart hardware should be upgraded to information hardware and interactive hardware, such as smart cockpits for cars. The second is also like The eight characters of Bichi, communicate everything and take care of everything, how can we empower 20 billion intelligent hardware with large models, countless professional knowledge problems that need to be dealt with, how to complete them, how to do professional vertical assistant.

CSDN: Finally, under the trend of large-scale model fire, what do you think of the risks of artificial intelligence from your perspective?

Yu Kai: Artificial intelligence is very similar to the invention of the steam engine. There were rickshaws in the past, but now they are invisible, so some industries will definitely be replaced by the emergence of new tools. The most dangerous is without a doubt the white-collar workers who do basic repetitive brain power, so some of his functions will be replaced. But at the same time, all the things that are replaced will create new industries. What is replaced is a class of skills, which will require humans to master a new class of skills, for example, who can use large models well, so I think While eliminating some old skills, it must be to master new skills. For human beings, it will push people to move forward. Therefore, the personal learning ability will always need to move forward.

Guess you like

Origin blog.csdn.net/FL63Zv9Zou86950w/article/details/131450627