"With immediate effect, the management decided to completely stop four types of outsourcing expenses related to creative design, solution writing, copywriting and short-term employees indefinitely."

Some time ago, an internal email from Blue Cursor was widely circulated online, which magnified the "unemployment anxiety" of people under the AI wave.

A report by investment bank Goldman Sachs suggests that 300 million jobs worldwide will be replaced by ChatGPT-generated AI, with lawyers and administration most likely to be laid off.

OpenAI, which created ChatGPT, also recently published an academic paper that systematically argues for the replacement of existing occupations by AI. The paper concludes that 80% of people will be affected, and 19% will be "greatly affected," and OpenAI says bluntly that blue collar jobs are relatively safe and white collar jobs are more at risk, with programming and writing being the most affected.

After interviewing three entrepreneurs who work in design and creative jobs, we found that the argument that AI is "taking jobs away" is perhaps too pessimistic.

But there is no denying that the nature and requirements of many jobs are changing, that various parts of the workflow are being reshaped, and that we are on the verge of an era of working with AI.

Game art: outsourcing basic execution to AI

Peter is a technical art team leader at a game startup.

According to him, the work related to game art is quite subdivided in the whole production line. The scenes and dynamic images of game characters are determined through 7 stages: "requirements - UI and original drawings - models - binding and modeling - animation - simulation - technical art".

Peter's point of view is that AI tools are already almost capable of mapping design, model part of the prop design, but at this stage they can't replace any of them completely.

For example, the model part is divided into character model and scene model. The character model, which needs to conform to the character's personality settings, technically involves the high-precision construction of various details such as hair and skin, which is currently unattainable with AI mapping. The same goes for the macro scene models, such as the global map of Song of Ice and Fire, which AI does not have such a huge imagination for the time being.

But the game props, environmental decoration of these models, more detailed, trivial, not so important in the scene; secondly, the shape and more basic, before Peter will be handed over to outsourcing or interns, now AI will be able to complete very well.

In other words, Peter's creative team's establishment has not been reduced, but AI did replace some of their previous needs for interns and outsourcing teams.

Then there's the mapping job, which involves unfolding a 3D model and mapping it on a 2D plane. Now you just need to give the texture you need in the Prompts, and AI will generate a texture that's pretty much the same, and the designer can modify it based on what AI generates.

For common generic mapping tasks, if the AI is well guided, it can output 70-80 points; then the original artist still needs to refine it to meet the delivery requirements.

In addition to some basic work assistance in the formal design process, AI can also make the preliminary internal communication more efficient. For example, in the concept design stage, AI can generate more image references in the same amount of time than manual drawing.

For small companies, this is a good thing," Peter said, "small companies have one carrot and one hole, and the previous demand is incomplete, with AIGC, the output is greatly improved, no need to outsource, two or three people can meet the concept needs of a project.

And in large companies with more than enough staff, the original artist will be more saturated. There may even be a surplus of personnel, and this time the ability to master AI graphics is very important competitiveness.

Many large factories in the game department have been a large number of use of AI to assist in drawing, and even set up AI art center. The threshold of AI painting is not as low as we think: AI is an advanced tool, and people are still needed to master it. It is the consensus of game art practitioners that "you can't not use AI, you can't rely on AI".

To use Stable Diffusion well, it requires a lot of effort to learn. For example, in the locally deployed Stable Diffusion, it is not enough to use only natural language to make a better picture, you need to use some command line and code to do it.

But Peter also admits that the speed of progress of AI tools gives him a sense of urgency. Previously, it was difficult to handle the details of the hands when drawing a figure with AIGC tools. Now, Stable Diffusion supports fine-tuning a certain part of the figure.

Peter even had an idea: "Can I record all the data about what I've mastered, how I've adjusted and controlled a model, and so on, and then feed it to the AI to learn and generate a digital doppelganger of my own? That way, as long as there are enough graphics cards, my capacity is unlimited."

Sounds, a bit crazy, but who knows if it will come true in the future?

Creative Studio: AI doesn't guarantee delivery yet

Zhai Dong and Young, who both came from graphic design, set up their own creative studios respectively to provide design outsourcing services for large companies.

Because the design process is relatively independent, the output is better measured, and the work is mostly project-based, the "individual studio" model is quite popular in the design industry. It is a more common career choice for designers with outstanding personal ability to set up their own studio after accumulating certain client resources.

Designers, according to different directions can be subdivided into illustration, graphic design, space design.

Zhai Dong's feeling is that although the major design categories will be affected by AIGC, but at present, the biggest impact is still the illustrator, graphic design and spatial design are relatively less affected.

He doesn't plan to hire full-time illustrators in the next few years unless the candidates have more complex abilities - such as planning skills, business thinking, etc. "What illustrators need to present is often not abstract content, but figurative expression, the more like the better, which is what AIGC is good at.

The higher-order needs of graphic design require designers to have user thinking, copywriting thinking, and certain abstraction ability, which are beyond AIGC's reach.

Young's company mainly provides new media graphic design materials for commercial real estate, and he holds the same view: "If a graphic designer only has the technical ability to operate software, lacks aesthetics, and lacks understanding of customer needs, even without AI, it doesn't quite meet our needs.

Young's studio now has five graphic designers, each with the ability to serve clients independently, more like project managers who understand design.

Another important reason why graphic design is less affected by AI is that AI drawing is less variable and manipulative, and elements cannot be split up. "Usually the client's needs will be more customized, for a small element fine-tuning, such as changing the color, adjusting the size, although the operation of the basic, but the controllability of AI is still not as good as manual.

The spatial designer, who ultimately delivers a 3D model, is only able to do a certain viewpoint of a drawing.

But Young's experience is that AIGC can really improve productivity. For example, when you show a client the intention drawing, you can use AI to produce the drawing directly, so you don't have to spend time and effort to model and present for a client who hasn't paid yet. In addition, AI can also help space designers broaden the boundaries of their thinking and provide new inspiration.

Zhai Dong's studio is already using AI tools extensively in the production process of the project. This includes style drawings, intention drawings, process drawings for initial contact with clients, and inspirational references for internal discussions. "Some of them will refine multiple angles or modeling based on the AI drawings.

For studio owners like Zhai Dong and Young, AIGC can help them improve session efficiency, not delivery efficiency.

Old jobs disappear, new opportunities come

When we look at something, we always have two choices: optimistic, and pessimistic. You can choose one or the other, or you can choose both.

Peter, Young and Zhai Dong have all made the "optimistic" choice.

"I don't see any job type being completely replaced by AI, but rather that these tools will make our jobs easier," Peter says. Peter says.

Young added, "Overall, AIGC is good for the creative industry, avoiding a lot of redundancies.

Zhai Dong believes that both this round of AIGC and the previous automated SaaS tools can help the formation of "super individuals" and solve the problem of "dumplings in a teapot": "With AIGC, a planner with no art skills can render a complete set of plans.

In a letter to investors, the CEO of Blue Cursor expressed a similar view: "We will see a group of 'super individuals' stand out. The best people will quickly master new tools, use the wealth of knowledge and problem-solving skills embedded in the models, and be far more productive than others through human-machine collaboration with AI work partners."

Both Zhai Dong and Young's creative studios have set up planning positions. They both said that the planning posts will not be reduced, but will be replaced: "Employees who are not as efficient as AIGC, for example, may have to be replaced to find a planner who is more proficient in using AIGC.

PricewaterhouseCoopers, a leading global consulting firm, predicts that AI will have more of a creative effect than a disruptive one in China, adding about 90 million new jobs over the next 20 years.

While AI is taking over the basic jobs, new jobs are also being created for "mentoring AI".

Many companies in BOSS have created the position of "AI designer". Responsibilities include proficiency in using AIGC design tools such as Stable Diffusion, in-depth application of controlnet and other plug-ins, and being responsible for lora model training.

Even with the popularity of ChatGPT, "ChatGPT Optimizer" has become a new type of position, responsible for optimizing and improving the performance and expression of ChatGPT in specific areas. It's easy to see that AI can be our right-hand man, but first, we need to choose a direction for it and give clear instructions.

When Young was hiring designers, basic drawing skills were definitely an important consideration. But now he will adjust the weight of this indicator and put more emphasis on the candidate's curiosity, responsibility and imagination.

According to "SOMEWHILE", the owner of B website, the emergence of photography technology has replaced the realistic function of painting, but to a certain extent, it has also contributed to the diversified development of painting: artists do not need to "roll" with the camera to restore the ability of realistic scenes, but can pay more attention to painting itself and their own emotional expression.

The same is true for the productivity revolution brought by AI. Instead of competing with GPT for rhetorical skills and Stable Difffusion for speed of production of visual spectacles, we need to focus on management, decision making, and creativity related skills.

As mentioned in Impromptu: Amplifying Our Humanity Through AI, there are three ways we can live with AI: 1. focus on asking good questions; 2. learn insights or skills that AI models don't have using "training data"; and 3. translate insights into action.

As Midjourney founder Holz says: AI is like water. It's dangerous, but people who know how to live with water live better. Translated with www.DeepL.com/Translator (free version)

"With immediate effect, the management decided to completely stop four types of outsourcing expenses related to creative design, solution writing, copywriting and short-term employees indefinitely."

Some time ago, an internal email from Blue Cursor was widely circulated online, which magnified the "unemployment anxiety" of people under the AI wave.

A report by investment bank Goldman Sachs suggests that 300 million jobs worldwide will be replaced by ChatGPT-generated AI, with lawyers and administration most likely to be laid off.

OpenAI, which created ChatGPT, also recently published an academic paper that systematically argues for the replacement of existing occupations by AI. The paper concludes that 80% of people will be affected, and 19% will be "greatly affected," and OpenAI says bluntly that blue collar jobs are relatively safe and white collar jobs are more at risk, with programming and writing being the most affected.

After interviewing three entrepreneurs who work in design and creative jobs, we found that the argument that AI is "taking jobs away" is perhaps too pessimistic.

But there is no denying that the nature and requirements of many jobs are changing, that various parts of the workflow are being reshaped, and that we are on the verge of an era of working with AI.

Game art: outsourcing basic execution to AI

Peter is a technical art team leader at a game startup.

According to him, the work related to game art is quite subdivided in the whole production line. The scenes and dynamic images of game characters are determined through 7 stages: "requirements - UI and original drawings - models - binding and modeling - animation - simulation - technical art".

Peter's point of view is that AI tools are already almost capable of mapping design, model part of the prop design, but at this stage they can't replace any of them completely.

For example, the model part is divided into character model and scene model. The character model, which needs to conform to the character's personality settings, technically involves the high-precision construction of various details such as hair and skin, which is currently unattainable with AI mapping. The same goes for the macro scene models, such as the global map of Song of Ice and Fire, which AI does not have such a huge imagination for the time being.

But the game props, environmental decoration of these models, more detailed, trivial, not so important in the scene; secondly, the shape and more basic, before Peter will be handed over to outsourcing or interns, now AI will be able to complete very well.

In other words, Peter's creative team's establishment has not been reduced, but AI did replace some of their previous needs for interns and outsourcing teams.

Then there's the mapping job, which involves unfolding a 3D model and mapping it on a 2D plane. Now you just need to give the texture you need in the Prompts, and AI will generate a texture that's pretty much the same, and the designer can modify it based on what AI generates.

For common generic mapping tasks, if the AI is well guided, it can output 70-80 points; then the original artist still needs to refine it to meet the delivery requirements.

In addition to some basic work assistance in the formal design process, AI can also make the preliminary internal communication more efficient. For example, in the concept design stage, AI can generate more image references in the same amount of time than manual drawing.

For small companies, this is a good thing," Peter said, "small companies have one carrot and one hole, and the previous demand is incomplete, with AIGC, the output is greatly improved, no need to outsource, two or three people can meet the concept needs of a project.

And in large companies with more than enough staff, the original artist will be more saturated. There may even be a surplus of personnel, and this time the ability to master AI graphics is very important competitiveness.

Many large factories in the game department have been a large number of use of AI to assist in drawing, and even set up AI art center. The threshold of AI painting is not as low as we think: AI is an advanced tool, and people are still needed to master it. It is the consensus of game art practitioners that "you can't not use AI, you can't rely on AI".

To use Stable Diffusion well, it requires a lot of effort to learn. For example, in the locally deployed Stable Diffusion, it is not enough to use only natural language to make a better picture, you need to use some command line and code to do it.

But Peter also admits that the speed of progress of AI tools gives him a sense of urgency. Previously, it was difficult to handle the details of the hands when drawing a figure with AIGC tools. Now, Stable Diffusion supports fine-tuning a certain part of the figure.

Peter even had an idea: "Can I record all the data about what I've mastered, how I've adjusted and controlled a model, and so on, and then feed it to the AI to learn and generate a digital doppelganger of my own? That way, as long as there are enough graphics cards, my capacity is unlimited."

Sounds, a bit crazy, but who knows if it will come true in the future?

Creative Studio: AI doesn't guarantee delivery yet

Zhai Dong and Young, who both came from graphic design, set up their own creative studios respectively to provide design outsourcing services for large companies.

Because the design process is relatively independent, the output is better measured, and the work is mostly project-based, the "individual studio" model is quite popular in the design industry. It is a more common career choice for designers with outstanding personal ability to set up their own studio after accumulating certain client resources.

Designers, according to different directions can be subdivided into illustration, graphic design, space design.

Zhai Dong's feeling is that although the major design categories will be affected by AIGC, but at present, the biggest impact is still the illustrator, graphic design and spatial design are relatively less affected.

He doesn't plan to hire full-time illustrators in the next few years unless the candidates have more complex abilities - such as planning skills, business thinking, etc. "What illustrators need to present is often not abstract content, but figurative expression, the more like the better, which is what AIGC is good at.

The higher-order needs of graphic design require designers to have user thinking, copywriting thinking, and certain abstraction ability, which are beyond AIGC's reach.

Young's company mainly provides new media graphic design materials for commercial real estate, and he holds the same view: "If a graphic designer only has the technical ability to operate software, lacks aesthetics, and lacks understanding of customer needs, even without AI, it doesn't quite meet our needs.

Young's studio now has five graphic designers, each with the ability to serve clients independently, more like project managers who understand design.

Another important reason why graphic design is less affected by AI is that AI drawing is less variable and manipulative, and elements cannot be split up. "Usually the client's needs will be more customized, for a small element fine-tuning, such as changing the color, adjusting the size, although the operation of the basic, but the controllability of AI is still not as good as manual.

The spatial designer, who ultimately delivers a 3D model, is only able to do a certain viewpoint of a drawing.

But Young's experience is that AIGC can really improve productivity. For example, when you show a client the intention drawing, you can use AI to produce the drawing directly, so you don't have to spend time and effort to model and present for a client who hasn't paid yet. In addition, AI can also help space designers broaden the boundaries of their thinking and provide new inspiration.

Zhai Dong's studio is already using AI tools extensively in the production process of the project. This includes style drawings, intention drawings, process drawings for initial contact with clients, and inspirational references for internal discussions. "Some of them will refine multiple angles or modeling based on the AI drawings.

For studio owners like Zhai Dong and Young, AIGC can help them improve session efficiency, not delivery efficiency.

Old jobs disappear, new opportunities come

When we look at something, we always have two choices: optimistic, and pessimistic. You can choose one or the other, or you can choose both.

Peter, Young and Zhai Dong have all made the "optimistic" choice.

"I don't see any job type being completely replaced by AI, but rather that these tools will make our jobs easier," Peter says. Peter says.

Young added, "Overall, AIGC is good for the creative industry, avoiding a lot of redundancies.

Zhai Dong believes that both this round of AIGC and the previous automated SaaS tools can help the formation of "super individuals" and solve the problem of "dumplings in a teapot": "With AIGC, a planner with no art skills can render a complete set of plans.

In a letter to investors, the CEO of Blue Cursor expressed a similar view: "We will see a group of 'super individuals' stand out. The best people will quickly master new tools, use the wealth of knowledge and problem-solving skills embedded in the models, and be far more productive than others through human-machine collaboration with AI work partners."

Both Zhai Dong and Young's creative studios have set up planning positions. They both said that the planning posts will not be reduced, but will be replaced: "Employees who are not as efficient as AIGC, for example, may have to be replaced to find a planner who is more proficient in using AIGC.

PricewaterhouseCoopers, a leading global consulting firm, predicts that AI will have more of a creative effect than a disruptive one in China, adding about 90 million new jobs over the next 20 years.

While AI is taking over the basic jobs, new jobs are also being created for "mentoring AI".

Many companies in BOSS have created the position of "AI designer". Responsibilities include proficiency in using AIGC design tools such as Stable Diffusion, in-depth application of controlnet and other plug-ins, and being responsible for lora model training.

Even with the popularity of ChatGPT, "ChatGPT Optimizer" has become a new type of position, responsible for optimizing and improving the performance and expression of ChatGPT in specific areas. It's easy to see that AI can be our right-hand man, but first, we need to choose a direction for it and give clear instructions.

When Young was hiring designers, basic drawing skills were definitely an important consideration. But now he will adjust the weight of this indicator and put more emphasis on the candidate's curiosity, responsibility and imagination.

According to "SOMEWHILE", the owner of B website, the emergence of photography technology has replaced the realistic function of painting, but to a certain extent, it has also contributed to the diversified development of painting: artists do not need to "roll" with the camera to restore the ability of realistic scenes, but can pay more attention to painting itself and their own emotional expression.

The same is true for the productivity revolution brought by AI. Instead of competing with GPT for rhetorical skills and Stable Difffusion for speed of production of visual spectacles, we need to focus on management, decision making, and creativity related skills.

As mentioned in Impromptu: Amplifying Our Humanity Through AI, there are three ways we can live with AI: 1. focus on asking good questions; 2. learn insights or skills that AI models don't have using "training data"; and 3. translate insights into action.

As Midjourney founder Holz says: AI is like water. It's dangerous, but people who know how to live with water live better. Translated with www.DeepL.com/Translator (free version)