Are there any signs of the development of AI

The development of artificial intelligence (AI) is getting faster and faster, and the impact on us is getting deeper and deeper. Stanford University doctoral student Michael Webb has a clever way to predict how likely your job will be replaced by AI in the future. ..

The development of AI is getting faster and faster, and the impact on us is getting deeper and deeper. Maybe our job will not be replaced immediately, but in recent years, many research institutions, media, and experts have predicted that AI will gradually replace us, and how many jobs will disappear after many years, it sounds shocking, but most of these predictions are from a few Individual cases, or estimates from what AI can do, cannot prove how accurate. Michael Webb, a PhD student at Stanford University, used a clever approach to develop statistical data that looks more solid than previous predictions.


The method he uses is the most common text comparison. Write what the AI can do. On the other hand, the professional work content is also described in words. The two texts are compared. If there are overlapping texts, then this This kind of professional work is connected with AI, and there is also a chance to be replaced by AI.

Are there any signs of the development of AI

Where can I find these two types of text data? There are some achievements in AI development, most of which will apply for patents, so AI data can be obtained from the patent database, on the other hand, from the government labor database, detailed occupation classification and job description can be obtained.

There are many types of patents. To obtain AI-related patent data, you need to use keyword screening, such as "neural network", and then match "verb-noun" composed of patent names, such as (diagnosis, disease) (identification, aircraft), etc. Etc., each patent name may have several groups of gerunds, and the name of all AI patents can be combined to create a huge gerund pairing database.

Similarly, from the occupation classification and work content, a huge gerund matching database is also created. The two matching databases are compared to find the overlapping pairs. After statistical weighting, the AI's Exposure Score is calculated.

The higher the exposure value of AI, the greater the chance that the job will be replaced by AI. The conclusions developed are similar to previous predictions. For example, affected by AI robots, the most exposed jobs include forklift driving, boom crane operation, elevator installation and maintenance, cleaning work, and train locomotive driving; while the least exposed jobs include wage timing, art and entertainment performances, priests, communications and communications Order services, word processing for government affairs, etc.

Under the influence of AI software, the most exposed jobs include broadcasting equipment operations, water treatment planning, parking lot management, manual packing and packaging, and train locomotive driving; the least exposed jobs include combing hair, treating foot problems, university lecturers and coaches, Arts and entertainment, postman and postal services.

Previous estimates said that robots will replace manual labor and software will replace repeated routine data processing, but there are many jobs in each occupation. These jobs are not affected by AI in the same way, so AI exposure is not based on professional estimates. Full, but calculate different exposure values for different jobs in the profession.

Taking production as an example, about one-third of the work is low-exposure, one-third is moderate-exposure, the other one-third is highly-exposed, and some high-exposure jobs are four times the median. In addition, sales work is focused on low and medium exposures.

On the whole, it is inconsistent with previous ideas. Most people think that people with high education, high salaries, and majors suffer the least, and at least the slowest. Work associated with production has high AI exposure. Graduated from graduate school is 4 times more exposed than graduated from high school, and a bachelor's degree is 5 times higher than graduated from high school. From the salary point of view, more than 80% of salary levels, the exposure value is as high as 40%; less than 25% of salary levels, the exposure is 0.

Men, major working ages, Caucasians, Asian Americans, are most affected by AI; women are more involved in social welfare, education, and health care. These service jobs have blocked AI attacks. The exposure value of men is 22%, Women are only -0.06%. In terms of age, 25-54 years old has the highest exposure, 55-64 second, and under 25 years of age and over 65 years of age have negative exposures. Asian Americans have the highest average salaries, second among whites, and relatively high exposure, with 19% for Asians and 10% for Whites.

In terms of location, big cities, high-tech, and highly involved manufacturing communities are all affected by AI. On the contrary, small and remote communities are minimally affected by AI. The economic types of information, technology, and professional management in big cities are oriented toward analysis, forecasting, and strategy, and are extremely vulnerable to AI.

The above are the statistical results from the comparison of AI patents and occupational content. Although quantification is still a prediction, not the results of quoting AI for jobs in various occupations. Although the development of AI is enthusiastic, it is still in its infancy. Whether it will completely replace existing jobs or partially replace or even develop new jobs that require manpower is still difficult to predict. The development of AI has not shown signs of slowing down, and various aspects of predictions will continue to push new. If your work repeatedly appears in different predictions, you must pay attention to it!

Readers interested in Michael Webb's research can download this report, The Impact of Artificial Intelligence on the Labor Market.