Sora cannot replace humans! Amazon engineer asserts: Real work conflicts cannot be resolved by AI

This week, when the OpenAI video AI tool Sora appeared, it was a big hit.

The fear of "not being able to keep their jobs" has truly hit many people.

Picture

However, Cameron Gould, an engineer at Amazon, believes that there is no need to be so afraid of AI, and it will not cause We lose our jobs.

Why? Here’s why.

The contradiction of Prompt

First of all, when generative AI is applied to creative content, it usually generates images, videos and general Nonfiction content.

Providing a generative AI model prompt is like dealing with a naughty child whose unique skill is created by "interpreting your words in a literal-oriented way" confusion.

Using accurate words is crucial because even tiny details can have a significant impact on the outcome, leading in a completely different direction.

But at the same time, you need to provide as much detailed information as possible so that the model does not misunderstand you.

This kind of contradictory operation is simply terrible!

Having a one-way conversation with a computer, trying to understand why it failed to produce the results you want, can give people the urge to hit the wall for three hours straight...

Picture

Gould said that in his career as a software engineer, he has actually become accustomed to this kind of one-way dialogue with computers. But the experience that generative AI gave him was definitely worse.

It is not a predictable programming language with idempotent functions.

Picture

It is completely a black box and can produce completely different results almost every time with the same input .

Even if you ask the AI ​​to maintain consistency in the prompt, it may not listen to you.

Companies worth billions have been troubled by this reality.

Recently, a Canadian court ruled that Air Canada must pay compensation to one of its passengers after a chatbot provided misleading advice that caused the passenger to pay nearly a million dollars for his ticket. times the cost.

Picture

The speed is really fast, but...

Of course, although There are flaws in consistency, but generative AI can help us produce things quickly.

Picture

For example, the picture above was created using generative AI and it only took a few seconds to make. Out.

So, it is possible that AI can allow us to produce pictures faster than artists.

And Sora also allows the general public to generate more complex media materials many times faster than before.

So, since one person can now do the work of several people, do we really not need so many artists, software engineers, and copywriters?

No, the details of it are wrong

The answer is not that simple.

When using generative AI to create art, you may often think: "No, this is not what I want."

This This is also a view that often appears on forums.

Some people may argue that this is because your prompt lacks details and you need to give a more specific prompt.

This point of view does make sense.

Pictures

What is the difference between our own pictures and professional films that have grossed over $100 million at the box office?

What is the difference between our own personal blog and the blog of a professional full-time writer?

What is the difference between the game video title we made ourselves and the 30-hour AAA video game title?

The answer is in the details.

AI生图

It’s the details that separate amateurs from professionals.

If we want to use generative AI to create work that is as impressive as professional content, we will need a prompt with a lot of detail.

Also, it takes a lot of trial and error to get the model to produce what we want.

Generative AI is just a tool, similar to autocomplete, compiler, spell checker or any other auxiliary tool.

It won't do any work for you except its own drudgery.

Let’s go back to Sora’s example.

It can only generate silent videos. There's no audio, no dialogue, no objects speaking or making sounds.

##Picture

But if we want to make a movie with it, we need to use it——

Generate script?

Read the script?

Generate a video to accompany a script?

What are the aspects of modern movies that make them interesting?

Suddenly, we need to integrate a bunch of different generative AI tools to get the job done.

Of course, OpenAI has a music generation tool called Jukebox, which can add music to videos, but it cannot add dialogue.

Picture

How to add dialogue to a video? The task sounds too daunting.

We not only need to generate a video of a specific scene, but also need to make the mouth movements and lines of the characters in the scene consistent.

Additionally, the video generator needs to know how the words are spoken. Not just the tone, but the rhythm.

Imagine how much manpower it would take to solve all these details?

The details become quite difficult.

At every level of every step, there are numerous details that impact material consequences.

Picture

Gould said that he recently read a blog post that he liked very much, "Reality has an amazing amount of detail."

picture

The details of reality are infinite, but the tokens of AI models are limited.

It can only take you so far, and after that, we need to rely on other people to get the job done.

By far, humans still do the heaviest work.

Pictures

Some important details about colonizing the universe

Less details, more Change

Focusing on details sounds really tiring.

What would happen if you kept it simple? What if you only need to provide a short description that only takes a few minutes to write?

Eventually, you will get a variety of results, most of which are completely inconsistent with your ideas.

You need to find the one you want among hundreds or even thousands of results, because prompt is too general.

You may never find a working result.

If the prompt is too simple, the model’s wild imagination will provide unlimited arrangements for you to choose from.

For example, enter this prompt - "a man", and you will get all kinds of men.

Picture

But actually, what I want is a guy with blue hair, brown eyes, a mustache, and A man in a black jacket and hood.

If you only enter "a man" as a prompt, how long will it take to find the exact output?

There is a balance between depth and breadth where you can minimize your effort.

This requires a lot of critical thinking about the details and a lot of searching.

AI will not replace us

So, why do we say AI will not replace us?

Because companies do not exist to solve small problems such as generating images and video clips.

Enterprises solve larger and more complex problems for their stakeholders.

The most difficult part of solving these problems is the logistics.

Solving large, complex problems requires the time and energy of a large group of problem solvers, all of whom need to be organized by other problem solvers to keep things moving. .

If you grab two people at random and ask them to self-manage and organize their work, it is likely that their standards and schedules will be inconsistent.

If you want them to work together, you have to set standards so that they work at the same time and produce the same quality of results.

Cohesion is a strength, and you need to do some work to make individual units consistent.

A team running alone can complete the task smoothly; but when we introduce multiple teams into a single environment, a lot of conflicts will occur.

Picture

When multiple interdependent teams work towards a goal, a certain amount of cohesion is required to achieve meaningful completion thing.

This "cohesion" work is actually very complex and often requires the resolution of interpersonal conflicts, which are often illogical and emotional.

Obviously, AI cannot resolve this type of conflict.

To sum it up is——

• Businesses make money by solving big problems for many people/businesses
• Big problems are difficult to solve because many teams are involved
• We need to carefully plan how to coordinate these teams to resolve interpersonal conflicts and maintain cohesion
• People are the most reliable resource when it comes to resolving interpersonal conflicts

Obviously, robots simply cannot do these jobs, and their performance will be terrible.

AI won’t take your job

When we take a closer look at the current state of generative AI, this fact is obvious——

Currently there is no way to automate the work.

Tip Interfaces and models lack autonomy and critical thinking, resulting in key limitations.

Yes, generative AI can indeed allow us to improve work efficiency, but that’s all.

Picture

This sentence has been agreed by more and more people: AI will not take away your job.

The real threat to you is someone who is more proficient in using AI tools than you are.

Generative AI is a great resource to help you speed up, but it won’t completely automate your work.

All you have to do is add them to your toolbox and master them skillfully.

Reference:

https://www.php.cn/link/52156f7194073f084237a48a1c2b7ea4

The above is the detailed content of Sora cannot replace humans! Amazon engineer asserts: Real work conflicts cannot be resolved by AI. For more information, please follow other related articles on the PHP Chinese website!