Why are AI developers so crazy? (I)

all-night online battles, hundreds of models competing, and thousands of companies challenging each other|Blogs|Tech Web|

2023-09-06 16:24:03 970

Just after midnight on August 31st, Baidu suddenly announced the opening of "Wenxin Yiyan" to the entire society.

Judging from the timing of the announcement, Baidu must have worked overnight and carefully timed the release of this news.

At 1:44 a.m., Zhupu AI followed suit and announced the official launch of their big model application "Zhupu Qingyan".

At 3:09 a.m., Baichuan Intelligence also joined in, claiming that their big model had passed the "Generative AI Filing" and was open for public service.

By the morning of August 31st, other big models such as "Shangliang" from SenseTime, "Yunque" from Douyin, and "Zidong Taichu" from the Chinese Academy of Sciences were also announced to be open to the public.

What happened to cause such an impatient rush among domestic big model players?

AI fierce

Since the explosion of ChatGPT, big models have become a "hot potato" in the field, with start-ups in related areas springing up like "flowers in chaos".

In August, an application called "Text With Jesus" was launched on the Apple App Store. It is said that the application used all versions of the Bible to train its big model, allowing users to chat online with characters such as Jesus, Mary, and Joseph, and even have debates with Satan for a monthly fee of $2.99 (approximately 22 RMB).

Back in March, the Japanese brought the cyber Buddha online.

Developers paired ChatGPT with the mechanical Buddha named Hotoke AI (hotoke being the Japanese word for Buddha). Within five days of its launch, the Buddha answered 130,000 user questions, and even had a "disclaimer" on its homepage - kindly accepting tolerance if the answers are not accurate.

Rewind, known as the "life search engine," is capable of recording the entire screen once the software is opened. It chronologically documents the webpages you have visited, videos watched, and even the text you have typed or spoken.

With Rewind, you can "rewind" to the computer interface at that moment, allowing you to retrieve missed tasks assigned by your boss during work slacking or recapture those moments of creative inspiration during brainstorming sessions.

By learning and processing text, audio, images, and videos, AI has greatly excelled in the film and television industry.

Recently, the trailer for 《Bapenheimer》 went viral on the internet. AI was responsible for everything from dialogue, scenes, and characters to posters, special effects, and camera transitions, reaching a point where it was almost indistinguishable from reality.

In the past year, with the rapid iteration of generative AI, large-scale model applications have ventured into various fields. Native AI applications such as AI voiceover, AI illustrations, AI scriptwriting, AI lawyers, AI doctors, and so on have emerged one after another.

All of this indicates that the next era is the era of artificial intelligence.

In the domestic market, large-scale model applications in specific fields are also thriving.

For example, ChatLaw, a legal large-scale model developed by Peking University, has been trained on a dataset consisting of 930,000 judgment cases. It can match user-provided information with corresponding legal provisions, thereby providing professional legal advice. It can even help generate legal documents with just one click, making it comparable to an AI version of Professor Luo Xiang.

In the office domain, WPS AI by Kingsoft can generate formatted documents directly based on keywords. It can help you write various documents such as leave applications, resumes, and rental contracts. It can also assist in analyzing lengthy and tedious articles to identify key points and extract original page numbers for quick reading. Additionally, generating intelligent PowerPoint presentations is also within its capabilities.

Recently, the popular Miya Duck Camera became a phenomenal AI application with its 9.9 yuan AI portrait service.

However, compared to the large-scale model applications that have flourished in overseas markets, commercialization of large-scale models in China is still far from sufficient.

According to a report jointly published by Sootoo.com and Large Model House, the global market size of large-scale AI models will reach 21 billion US dollars in 2023 and 109.5 billion US dollars in 2028. However, the Chinese market for large-scale AI models in 2023 is only expected to reach 14.7 billion yuan and is projected to reach only 117.9 billion yuan in 2028.

In order to catch up with their foreign counterparts, everything has been accelerated:

On August 15th, the "Interim Measures for the Management of Generative Artificial Intelligence Services" was officially implemented, and the first batch of large-scale models were recorded by August 31st.

Following this, the scene at the beginning of the article appeared overnight, with large-scale model vendors opening up to the entire society.

The development of large-scale models highly depends on user feedback. Prior to this, domestic large-scale model products were in the form of closed testing and invitation-only when facing users. Full openness can reach a larger user base.

As large-scale models continue to evolve, future models will move beyond just competing in terms of "technology" (such as algorithms and computing power) and enter the stage of competing in "application" implementation.

Given the sensitivity of policies and the active actions taken by major players, the competition in the AI industry between China and other countries is pressing. Time is tight, tasks are heavy, and both technology and application must be grasped with equal strength.

In particular, application implementation is even more urgent and important, as only by attracting a massive number of users can the industry switch from a blood transfusion state to a self-sustaining state and sustain itself during the competition.

Therefore, will the development of the domestic AI industry follow the same story of "overtaking on a bend" as other industries?

To be continued...

Prev:: Why are AI developers so crazy? (II)

Next:: How did the era of NVIDIA come with the "airdrop" of ChatGPT? (II)

Why are AI developers so crazy? (I)

Advisory Service