AI can make movies. What happened to the big model in the past 14 months？

"There are two things that the more I think about, the more I feel awe. They are the starry sky overhead and the moral law in the heart. "

Wen/Ba Jiuling

February 16, 2024

The boundary between virtual and reality

Began to blur completely

To paraphrase "Three-body"

"Reality, no longer exists"

Subverting all this

It was released by OpenAI yesterday.

The first Wensheng video model "Sora"

That night.

Movies and gamers "stayed up all night"

Sora is very good

It can generate a 60-second video.

There are also "single video multi-angle lens" and "world model"

Very, very exaggerated functions.

this means

AI has understood the concepts of light and shadow, gravity and collision.

Can understand and simulate the physical laws of the world.

Even a few descriptions can be reproduced.

Scenes in the classic game My World

In other words,

AI can finally make a movie.

More exaggerated.

Hollywood may be "one pot at a time"

Following the occupation of the graphic field by AI

The advent of Sora

Announce "video content"

This last and strongest fortress of mankind has been breached.

In terms of content generation,

Has gradually entered the era ruled by AI.

And the starting point of all this

It happened in November 2022.

* This article combines "Foreseeing the 2024 Wu Xiaobo Factory Year-end Show"

The content of the halftime speech "12 months of the big model"

After reviewing ChatGPT’s official launch.

"Evolution" of Global Megamodel and Major Public Opinion Events

The pictures in this paper are all generated by Midjourney.

Back to the beginning

November 30, 2022

A company called OpenAI.

Silently launched an experimental product-

ChatGPT3.5

Under the "super power" of the founder Altman,

Just two months later,

ChatGPT global monthly users soared to 100 million.

Break the fastest record ever set by TikTok.

In the same month, OpenAI also launched the Plus version.

Started commercial operation.

How awesome is ChatGPT?

In the qualification examination for practicing doctors and lawyers in the United States

ChatGPT got qualified grades.

The two most difficult exams for human beings were easily overcome.

So in the rapidly growing January,

89% of American college students use it to do their homework.

No wonder Bill Gates said excitedly

The significance of the birth of ChatGPT

No less than the birth of PC or Internet.

And Huang Renxun compared ChatGPT to

"iPhone moment" of artificial intelligence

February 2023

OpenAI introduces API programming interface of GPT.

Just connect the API

You can do your own AI application.

You can even build a big model.

After the programming interface is opened.

"Shell AI" has mushroomed.

China’s "big model war" is poised to break out.

Fudan University released the first ChatGPT-like model in China.

Baidu, Alibaba Cloud, Netease and other enterprises.

Announced research and development of ChatGPT products one after another

……

Outbreak and crisis

The 4th month of ChatGPT’s launch.

Baidu released it on March 16th.

ERNIE Bot, China’s first generative language model.

Throughout 2023

ERNIE Bot’s call volume

It is the sum of all large model calls in China.

In the field of fast iterative technology

Leave early.

Is to take the lead.

Just the day before the release of ERNIE Bot.

OpenAI has released a multimodal version of hundreds of billions of parameters.

GPT4.0

The parameters and performance have been improved at epic level.

Then fully implanted into the Office family bucket.

What is even more shocking is that

Midjourney, the AI drawing tool, has been updated to the fifth edition.

It generated a photo of a couple in the background of the 1990s.

Completely detonate the Chinese Internet.

We finally saw a new technology.

Challenge or exceed human ability.

At the same time, it caused panic among scientists and business circles.

OpenAI’s angel investor Musk

More than 1,100 experts jointly issued an open letter.

Require all AI laboratories to suspend immediately for 6 months.

Italian government orders blockade of ChatGPT

Samsung electronics employees were arrested for ChatGPT

Causing the disclosure of company secrets

350 global artificial intelligence experts jointly warned

"AI may bring extinction risks to human beings"

ChatGPT is a tool.

Deconstructionist philosopher stigler.

Put forward "technical ethics"

He said that the process of human development

Is a process of being iterated by tools.

Revolutionary tools not only triggered the big explosion of productivity

New culture and new ideology will also be born.

ChatGPT today is like

Discussion on "Alienation" of Labor to People during the Industrial Revolution

And the fear and controversy after the invention of the atomic bomb

Accelerate in the dispute

Everything revolutionary

Always speed up in disputes.

So that we can’t match it.

May 2023

Huawei Cloud Pangu Meteorological Model

The track of Typhoon Mawa was predicted 10 days in advance.

Alibaba Cloud released the big language model "Tongyi Qianwen"

Google Releases Big Language Model PaLM2

……

June 2023

ChatGPT’s application is online

Microsoft’s Bing and Apple iPad.

Japanese local government takes the lead in trying ChatGPT office.

Chinese Academy of Sciences released Zidong Taichu 2.0.

Then came July.

META released 70 billion parameters.

Open source large model base Llama2

Android version of ChatGPT

Launched in the United States, India, Brazil and other countries.

JD.COM launched the "big model of words and rhinoceroses"

The language model is in full swing

AI applications are also closely followed.

Two topical products appeared this month.

One is the "cheating glasses" of Stanford University.

One is the "crew robot" of the Finnish company.

However, during this period,

Two American writers sued OpenAI.

It is the first ChatGPT copyright infringement case in the world.

What is even more worrying is that

The anxiety of AI robbing human rice bowl is also increasing day by day.

In this regard, kevin kelly is writing a new book

The world after 5000 days answers.

Facing AI

Humans need the ability to ask questions and imagination more.

"Hundred Models Battle"

By August,

OpenAI launched ChatGPT enterprise edition.

More and more organizations are involved in great changes.

this moment

More than 80% of Fortune 500 companies

ChatGPT products have been adopted

The China model is also accelerating.

Baidu "ERNIE Bot" opens its services to the public.

Tencent’s "mixed-element big model" entered the application beta.

Zhipu Huazhang launched AI assistant "Zhipu Qingyan"

……

And with September 2023,

Alibaba Cloud’s Great Model of "Universal Meaning and Thousand Questions"

Officially open to the public

There are more than 100 large models in China.

See the big model of collective "volume"

Li Yanhong reminded

China needs AI native applications in the order of 1 million.

But you don’t need 100 big models.

And now the emergence of Sora

A saying turned out to be a prophecy.

By October,

OpenAI has developed to provide

Map reading, surfing the Internet, drawing, data analysis, etc.

one package service

META launches generative AI advertising service

AI copywriting, AI drawing, automatic delivery

Since November 2022,

After Pandora’s box was opened,

A "devil" called generative artificial intelligence

Has been iterating rapidly.

It is important to run early.

But more important is the iteration speed.

And at this point,

There are 238 large models in China.

"the sheller" and "the jogger"

Will face a brutal knockout

Key change

At least for OpenAI.

November is an extraordinary month.

It not only held the first developer’s day

Also launched the GPT store.

And officially announced the custom GPT—GPTs.

Global developers exceeded 2 million.

However, only 11 days later,

Sam Altman was suddenly dismissed by the board of directors.

Back in power after 48 hours.

This dramatic event means

Innovation path under the premise of technical ethics

There are still major discussions and disputes.

It will still torture us

Organization, world outlook and ideology

It was also in November.

The concept of AI assistant has become hot.

Bill Gates, the big money owner of OpenAI

Predict that within five years

Everyone will have their own AI personal assistant.

It will completely change the way of life of mankind.

Of course, OpenAI is not invincible.

December 6th.

Google released Gemini Gemini

In the large-scale multi-task language understanding test

Beyond human experts for the first time

Sam Altman knows very well.

Google is bound to be its biggest rival.

A new plan that has been brewing for a long time.

Be ready to come out soon.

New changes in 2024

Step into 2024

The AI era is accelerating like Moore’s Law.

King fried in January

Still depends on OpenAI.

It’s officially launched in GPT store.

It is "Apple Store in AI field"

At present, the number of GPT in the store has exceeded 3 million.

In the following February,

Sam Altman announces "secret plan"

In pursuit of computing power

He plans to raise 5-7 trillion dollars.

Build a chip factory belonging to OpenAI

This is not only to crack the "stuck neck" of the computing chip.

Backstab severely, NVIDIA of Huang Renxun.

It is also the way to continue to lead the opponent.

However, in these two months,

Competitor Google is not idle either.

The multi-modal large model Gemini Ultra 1.0 was released.

Quickly upgrade to version 1.5 within one month.

And unify all its AI brand names.

Comprehensive benchmarking GPT-4

At present, in terms of text processing length and ability

It has surpassed GPT-4 Turbo.

Processing ability of AI big model

Broaden by "epic"

But just two days later,

With the detonation of Sora, a deep-water bomb,

OpenAI seems to have the last laugh.

tag

The evolution of ChatGPT and other big models

Like ape-man evolved into homo sapiens.

And all this happened only in

In just 14 months,

Masayoshi son, the founder of Softbank, once predicted

Generative artificial intelligence (AIGC)

Is entering a wide range of applications.

General artificial intelligence (AGI)

Will become a reality in 10 years.

10 times smarter than humans

Super artificial intelligence (ASI)

Will be realized within 20 years.

10 thousand times smarter than humans

And the emergence of Sora

In the eyes of some insiders,

The realization of AGI will be shortened from 10 years to 1 year.

While marveling at the exponential evolution of AI

We still have to ask ourselves a question.

Generative or universal artificial intelligence

What does it mean for each of us?

What is the technical ethics behind it?

Is it possible that mankind will be destroyed by this?

Our own children.

As Kant said,

There are two things that I think more about.

The more awe I feel.

They are the starry sky overhead.

And the moral law in my heart

Facing the change of artificial intelligence

We should maintain great awe of technology and starry sky.

At the same time, we must firmly believe in human morality.

Author | Tao Xu | with pictures | Zhao Jie

Editor in charge | He mengfei | Editor | He mengfei