There is a point of view that the artificial intelligence development we see is just an extension of machine learning in the past 20 years.Except for ChatGPT, there are no other artificial intelligence products in the market.
But this time is different.Two important models have appeared at present-one is the image diffusion model, such as Midjourney and Dall-E 2, which have a powerful feature of text-generated images; the second is a large language that uses the "Transformer" architecture.Model (LARGE LANGUAGE MODEL (LLM), which can generate coherent and smooth texts, and put forward opinions based on the Chain of Thought and logical reasoning.
Ajay Agrawal, Avi Goldfarb and Joshua Gans' power and predictions (Power and Prediction), the application of artificial intelligence and the application of electricityAfter comparison, the application of artificial intelligence was revealed.Although Edison invented the first light bulb in 1879, the large -scale application spent many years -after 20 years, only 3%of families and few factories used electricity.However, once the electricity is popular, the economy has changed dramatically and greatly changes the organization's organizational methods.
At that time, the factory used steam engine technology. It was located in the center of the factory and drove a central axis connected to the artificial pulley.Because the space around the central axis is very limited, the factories are built vertically.When the electricity began, it was originally used as a point -type solution, that is, using an electric motor to replace the steam engine.However, changing power sources cannot reduce how much electricity costs can be reduced.The use of electricity can only be popularized after using electricity at the system level, to complete a single job with smaller electric motors, and reorganize the factory.Since the space around the central axis no longer has any restrictions, the factory can be expanded to horizontal.This has spawned the assembly line and promoted the large -scale manufacturing industry including Ford Motor.
Artificial intelligence will evolve along the same trajectory.Although the "converter" has appeared as early as 2017, LLM only received significant attention only after GPT was released seven months ago.We are at the starting point of using curves.At present, LLM is regarded as a point solution for replacing artificial tasks, but the real potential is to reintegrate its application and organize our system around artificial intelligence.We overestimated the short -term impact of artificial intelligence, and at the same time underestimated its long -term impact on our lives and economy.
The entrepreneurial process of an entrepreneur fully reflects this potential.In the aviation industry, which is important and strict in supervision, his startup uses LLM to simplify security reports, allowing pilots to report in the application in the application through voice, and overcome the difficulty of filling in the lengthy and difficult table.For example, when a pilot reports the bird's strike incident, the application will ask additional questions to ensure that the content of the report is comprehensive.
By reaching an agreement with the top regulatory service company, the startup signed a contract with more than half of the airline within six months.With these comprehensive data, it develops a new system solution into a new system solution to help airline predict accidents.
This startup initially only OPENAI's GPT3.5.This situation is likely to repeat the entire economy.The subversion will not only come from the startup, but also from a large company like Microsoft.Each software that Microsoft is selling is implanted with artificial intelligence.
For each CEO, the key is that each company needs artificial intelligence and data strategy.In the future where artificial intelligence models are continuously improved and public data used for training, proprietary data will become the key to standing out.Therefore, it is necessary to determine the company's strategy to be tightened and ensure that its safety is crucial.We must understand the data collected and sent.
We will usher in an exciting future, and everyone has powerful tools available.The programming language translates the assembly code into a more natural language, so that more people can write programs.LLM is the end point of the evolution of programming language. Anyone who can communicate with mother tongue can let the machine work for us.
Two weeks ago, I used PAIR (a public service tool based on GPT3.5) to generate a preliminary draft of speech; using OpenAI code interpreter to analyze statistics to enhance my dream football team; use Gamma using GammaAI generates slides for classrooms; use ChatGPT to understand more courses and timetables about product engineering knowledge; use Claude to plan my will soon start the Bali Island trip.To flourish in this new world, we need to use the right brain more.
Daniel Pink (Daniel Pink) lists the six emotional capabilities we need to develop in the book of the new thinking: not only functions, but also re -design; not only the arguments, but also stories; not only talkProfessional, you must also integrate; not only talk about logic, but also care; you can not only be serious, but also play; not only care about making money, but also significant.In other words, we need to be more humane.Microsoft CEO Satya Nadella believes: "The next generation of artificial intelligence will start a wave of new rounds of productivity growth. The powerful Copilot will eliminate people's daily tasks and hardships in work."
In a world where a machine can generate answers based on past questions, how can we learn to ask the question of letting us discover new answers?
The author is the Ministry of Communications and News (Development) Smart Country and Digital Government Work Team (Development) and the President of the President of the Prime Minister's Office (Development, Network Security).
