It is of great significance. This is when AI enters the productivity stage.
On Monday, the artificial intelligence drawing tool firefly demonstrated by Adobe and Nvidia detonated the circle of friends.
This is a major upgrade of Firefly. Although the ability to edit videos has not yet been opened, this tool can already realize drawings and sketches. , modify the screen content with one click, and can also modify, layer, and output ultra-high resolution automatically generated images, which has initially reached productivity standards.
The products launched by Adobe are the software giant’s answer to the impact of AI drawing. Since the proliferation of model series shocked the field of painting last year, people are looking forward to what artificial intelligence technology can do. We have finished some work at hand, and the demos of some products have made us eager to try them.
In addition to images, the natural language processing direction of artificial intelligence has experienced an unprecedented breakthrough. Since the end of November last year, ChatGPT launched by OpenAI has quickly become a hotly discussed topic. Facing the new situation, domestic and foreign technology companies have followed up at an unprecedented speed. Technology giants are trying to build their own versions of ChatGPT and hope to achieve true general intelligence (AGI) through larger-scale models.
But at an event at the Massachusetts Institute of Technology (MIT) last week, OpenAI CEO Sam Altman made another statement. He warned that the research strategy that gave birth to ChatGPT has ended, and it is unclear where future progress will occur. OpenAI has not yet trained GPT-5 and will not train it in the short term.
Is the era of large models over?
In addition to adding more parameters to the model, there are many other ways to make large language models (LLM) better and more useful. On the other hand, many companies and institutions are exploring the application of large models. From the perspective of practical technology, our imagination may not have kept pace with AI innovation.
ChatGPT is here, what will the world look like? After the launch of GPT-4, OpenAI also conducted serious research. A paper it submitted stated that ChatGPT may affect 80% of jobs, and jobs with higher salaries will face more potential impacts from large AI models.
In other words, after the breakthrough of AI technology, there is a lot of room for improving efficiency in many jobs. You give orders and AI is responsible for doing things. , will be a completely new form of work. Are the good days of being a worker coming? We can see signs of change in some things.
First of all, the computer field itself. Based on the large language model GPT-4, GitHub Copilot has built a chat mode centered on software program code, covering all processes of the development life cycle. By integrating chat and voice capabilities, developers can naturally communicate with AI. In addition to automating code writing, Copilot is also integrated into pull requests, command lines, and documentation to provide instant answers to questions about your project.
According to statistics, AI writes 46% of the code currently using GitHub Copilot, which can help developers increase work speed by 55%.
The Office series, which is designed for most office scenarios, is also undergoing AI transformation. On March 17, Microsoft officially announced Microsoft 365 Copilot, which covers all components of the Office family.
Copilot is based on GPT-4. It can call data from other software in the user's Word, directly create Excel tables and PPT based on the Word content, and realize the conversion of content forms. In addition, it can change the writing style according to user needs, realize personalized writing, and automatically generate Excel analysis reports.
Microsoft CEO Satya Nadella said that the emergence of Copilot is a milestone, which means that the way humans interact with computers has entered a new stage. From now on, people’s The way we work will change, ushering in a new round of productivity explosion.
As more and more productivity tools based on large model technology gradually come online, we may be entering a new era. The outbreak of artificial intelligence has been described as the fourth industrial revolution - the first three have greatly improved productivity, but also provided more advanced production tools, but the fourth industrial revolution is a little different. "Finished products" can appear automatically, and many What you want to do, you only need to think about it and no longer need to do it.
But just like code generation tools are to programming, if you want to achieve automation in many vertical fields, you must have specifically optimized AI tools. After all, as Ali CEO Zhang Yong said , "All products are worth redoing with large AI models." What other businesses have been automated? Recently, an automated tool based on large language models (LLM) has made a good start in the field of UI design.
On March 28, a generative AI product for UI design can allow AI to generate "real-time" editable UI design drafts. "AI" opened internal testing applications and quickly gained the attention of people in the industry. Tens of thousands of applications were received within 24 hours. As of now, more than 30,000 people have participated in the internal beta, and more users are waiting for the issuance of internal beta qualifications.
According to an inventory report released by the authoritative domestic user experience platform "UI China", there are currently three products in the world that can achieve "AI-generated editable UI" ” functions, namely “real-time AI”, Galileo AI and Uizard. Among them, two foreign products are still in the small-scale internal testing stage. Currently, "Instant AI" is the world's first product that allows users to experience the "AI-generated editable UI design draft" function on a large scale.
"Instant AI" is based on the online collaborative design platform "Instant Design" and uses a large language model to allow people to edit the UI design draft through natural language description on the web page. Quickly generate.
Use "real-time AI" and describe it in a simple sentence. AI can automatically generate a UI interface template for you:
If you describe the interface requirements in detail and write the requirements with qualifying statements like AI drawing, AI can complete the task on demand:
As a practical productivity tool, Instant AI has component recognition capabilities. It can not only identify a single layer, but also identify most components on the page, such as buttons, title bars, bottom bars, etc. With component-level identification capabilities, designers can use this tool to generate design drafts that comply with different design specifications.
The design draft generated by AI includes vector layers and icons, and also automatically organizes the hierarchical structure, supporting manual secondary editing.
Now, designers only need to enter text instructions, and they can quickly get a UI design that initially meets their needs, is editable, and has a clear layer structure. Drafted. In the conventional UI design process, a page takes designers about 30 minutes on average, but the time it takes to generate a page using "real-time AI" is just the time you write the prompt plus 1 minute waiting for the large model to respond. And the generation time will be compressed again as the version is updated. The introduction of AI technology can give you more opportunities to organize your ideas and explore different designs.
Of course, this set of tools is not yet intelligent enough to replace designers. The designs it generates still require a lot of modifications by people before they can be used, and the scope of generation is also limited. Many people use the evaluation that the level of production is similar to the high-fidelity prototype drawings of product managers or the level of junior designers. If you reduce the description and let the AI do what it wants, it's currently a great brainstorming tool.
Designers said that after the emergence of AI generative tools, the form of UI design will change. First, after people give their needs, designers can provide Party A with a variety of design solutions to choose from in a short period of time, and quickly provide feedback and adjustments to further needs. On the other hand, if you encounter a relatively simple need that has appeared in the past, you can quickly complete the task in a short period of time.
The tool was updated this week, increasing the number of pictures each time. Due to the large number of people who applied for internal testing, an additional 20,000 internal testing places were added.
In the vertical field of UI design, the most popular tool in the world is undoubtedly Figma. Last September, Figma became big news in the industry when it was acquired by Adobe for $20 billion.
In order to maintain its position as the industry leader, Adobe offered a price that people could not refuse.
What’s interesting is that Figma’s vision is to “eliminate the gap between imagination and reality”, and now this is realized to a certain extent by AI technology .
Since Instant Design was officially launched in September 2020, its user base has exceeded 2.3 million. The startup is currently exploring more productivity tools. The launch of “real-time AI” has also brought more room for imagination to the industry.
Referring to the evolution speed of mainstream AIGC tools, perhaps in the near future, we will see real-time AI generate UI pages that are close to the level of senior designers, and directly generate multiple sets of series of UI designs. draft, or even directly produce a usable final plan, and finally have a complete design system.
It must be noted that AI based on large models cannot help us solve design problems in one step. But designers can significantly save time and give full play to their ideas and creativity with the help of AI.
Think from this perspective and build a set of AI auxiliary tools that allow designers to quickly generate and get started using "prompts" of natural language text or a small part of the design as a guide. Making improvements may be the way to work in the future.
In short, in the future, "knowing what the demand is" will definitely be much more important than "knowing how to do it".
In the future, many work processes of every company will be automated by AI. From user research, requirements, prototypes, beautification, front-end to back-end, most links will be simplified by AI. possibility.
Are you prepared to face such a trend?
The above is the detailed content of Less than half a year after the advent of ChatGPT, the industry has been transformed by AI. For more information, please follow other related articles on the PHP Chinese website!