Anthropic Research Shows How AI Affects Jobs in the Real World
Anthropic put out new research that looks at how people are using AI tools like Claude at work these days. The study introduces a new metric called “observed exposure,” which measures how often AI does tasks in real jobs. This method looks at how AI is actually used in the workplace instead of what people think it could do.
Researchers merged the O*NET database, which contains information on about 800 jobs, with real-world usage data from Claude. They also used a 2023 academic framework to see if AI can cut down on the time it takes to finish tasks by a lot. These datasets work together to give us a better idea of which jobs are at real risk of being automated by AI.
Source: The Financial Express
Observed Exposure Shows Gap Between AI Capability And Reality
The study showed that there is a big difference between what AI can theoretically do and what it actually does. The research framework shows that up to 90% of office and administrative jobs could be automated. But the current observed workplace usage only includes about 1/3 of jobs that involve computers and math.
This gap shows that just because a technology is capable doesn’t mean it will immediately replace jobs in all fields. Companies often take their time adopting new tools because they need to train their employees, integrate them into their workflows, and follow the rules. This means that millions of people still have jobs, even in fields that are thought to be very easy to automate.
Coding And Data Entry Roles Show Highest AI Task Coverage
Anthropic’s exposure measurements show that computer programmers are now one of the most exposed jobs. About 75% of the tasks that programmers do now show up in how AI systems are actually used. A lot of businesses already use AI coding assistants to write scripts, fix bugs in software, and do other boring development tasks.
As businesses add AI chatbots and automated response systems, customer service representatives also seem very vulnerable. Data entry jobs come in 3rd, with about 67% of tasks covered. This is because machine learning works well with document processing. More and more businesses are using AI to read documents and enter structured information faster than people can.
Recommended Article: US Military Used Anthropic’s Claude AI in Venezuela Raid
Financial And Administrative Jobs Also Face Rising Exposure
Financial analysts and office managers are also jobs that are getting more and more exposure to artificial intelligence. AI can do these jobs well because they often involve analyzing data, making spreadsheets, and writing reports. Even if adoption is only partial, the technology still cuts down on the time it takes to do routine analytical tasks.
Observed exposure for computer and math jobs is currently around 33%, but it is slowly rising. As companies use AI more in their digital workflows, analysts think these numbers will go up. More automation may not get rid of all jobs, but it could change the duties of many white-collar jobs.
Physical And Hands On Jobs Remain Resistant To Automation
Anthropic’s study found that about 30% of workers have jobs that don’t use AI in any way that can be measured. A lot of these jobs depend on being able to interact with people, be aware of your surroundings, and make decisions in real time. AI systems today don’t have bodies, hands, or the ability to sense things, which are all things that are needed to do this kind of work well.
Some examples are cooks making food, mechanics checking engines by hand, and lifeguards keeping an eye on swimmers. Bartenders also stay away because they need emotional intelligence to read social situations and deal with unexpected events. Farm workers who run machines and prune crops also need to be physically present and aware of their surroundings.
Labor Data Shows Automation Threats Differ From Past Waves
Researchers found that workers who are most affected by artificial intelligence automation are different from workers who were affected by other technological changes in the past. Workers with high exposure are usually older, have more education, and make about 47% more money than workers with low exposure. In the past, waves of automation affected jobs in manufacturing or manual labor that paid less before they reached professional jobs.
The change suggests that AI could have a bigger impact on white-collar jobs than previous cycles of industrial automation. But the study did not find any measurable rise in unemployment among professions that use ChatGPT. Instead, the first signs of trouble show up in the way younger workers are being hired to work in these fields.
Hiring Trends Suggest Early Workforce Changes Emerging
The rate at which workers aged 22 to 25 find jobs in high-exposure fields has dropped by about 14% in the past few years. The drop is still only slightly statistically significant, but it is similar to other signals that have been seen in independent payroll data research. Young graduates going into fields that are open to the public seem to have fewer job opportunities than in previous years.
Investors and policymakers are already making changes to deal with these new trends in the workforce that are being caused by the use of artificial intelligence. More institutional money is going into healthcare, utilities, and physical service sectors because there is less risk of automation. Governments are also expanding retraining programs that encourage people to work in trades, care services, and other jobs that involve helping people.
