Which jobs will AI hit hardest, and which will stay safe the longest?

OpenAI’s ChatGPT is a slow, buggy, boring-looking chat bot that frequently hoaxes and cuts off mid-response. However, they can communicate as well or better than most humans, and they also write code. It does both in a variety of languages ​​and at times shows some extraordinary insight and creativity. It can already perform a staggering (and rapidly growing) range of tasks without These tasks can be performed in a fraction of the time it takes humans and almost for free.

This is an ominous preview of the coming months and years of AI systems, each smarter, faster, more capable, and more integrated with our lives and devices. . GPT-4’s prodigious ability as a “machine that can do anything” will look incredibly crude in a year or two.

We all knew AI automation was coming to work. Most people never imagined that work involving communication and creativity would take the first hit. So, who pays design firms to come up with business logos?

A friend of mine who runs a decent-sized company spent a weekend tinkering with GPT-4 to develop a personality type test specifically designed for his industry that is as scientifically accurate as Myers-Briggs. told me (i.e. not at all). He directed him to design a management plan for a specific personality his type and a handbook to help people understand their type and how best to work with other types. bottom. He wrote code for his website that tests employees with multiple-choice questions, grades them, and assigns them to groups. Entire companies are built around this kind of IP.he created something that he was happy to use in about 30 minutes at almost zero cost. A tectonic shift is underway.

A Goldman Sachs report is an attempt to predict what this dramatic shift will mean for economic growth. Thank you for doing the best you can when you see technology changing so much, barely being understood, and evolving at a frighteningly rapid rate.

“The labor market could face significant disruption if generative AI delivers its promised capabilities,” reads the report. “Using data on occupational tasks in both the United States and Europe, approximately two-thirds of current jobs are exposed to some degree of AI automation, and generative AI is expected to cover up to one-quarter of current jobs. , suggesting globally that generative AI could expose the equivalent of 300 million full-time jobs to automation.”

On the other hand, new jobs will be created here as well, with high productivity and low labor costs, “increasing the likelihood of a productivity boom that will significantly boost economic growth.” Goldman makes very conservative estimates for this. “AI could increase annual labor productivity growth in the US by just under 1.5 percentage points in the decade since it was widely adopted,” he said. What is the projected net economic impact? “AI could ultimately increase global GDP by 7% annually.”

7% of global GDP is indeed a staggering amount. But think of it this way. You, me, and everyone else will have instant access to something akin to an armada of minions in a box, able to perform tasks almost as fast as you can direct them. The amount of bullshit hectic work we do should plummet. Many big ideas take little time or effort to execute. From that perspective, 7% seems comically low, but it’s worth noting that zero-effort execution also drives output product prices to near zero.

Goldman is also trying to figure out which jobs will be hit hardest, at least given our current understanding of technology. makes for an interesting read.

The report found that “higher exposures were found in managerial (46%) and legal (44%) positions, and lower exposures in physically intensive occupations such as construction (6%) and maintenance (4%). I predict. The impact will be felt more pronounced in developed countries than in emerging markets, he said. The latter relies heavily on manual labor, and affordable general-purpose robots seem to be a long way off.

According to the report, the category with the highest percentage of jobs that are more or less completely unaffected by AI automation is building and grounds cleaning and maintenance (~95%), followed by installation, maintenance and repair (~85%). %) is. , construction and extraction (~75%), production (~72%), transport and material movement (~65%), and food preparation and serving related (~50%).

On the other hand, almost everyone on the right side of the chart can expect AI to radically change their workflows. This includes workers in healthcare, agriculture, creative fields, management, sales, social services, finance, education and computing. However, Goldman does not see a lot of manpower lost in these areas.

Behind management and legal, where nearly half of all jobs are expected to be under threat, Goldman said the categories most at risk of large-scale job displacement include architecture and engineering (about 10% are expected to disappear), which we estimate to include life, physical and social. Science (~8%), Food Preparation and Serving (~7%), Production (~7%), Business and Financial Operations (~4%), Arts, Design, Entertainment, Sports, Media (~3%).

Overall, the report predicts that about 7% of U.S. workers will have been displaced by AI, most of whom will find new jobs “only marginally less productive.” I predict it will.

Overall, this report looks pretty rosy from where we stand. Probably too rosy. The next wave of AI will be expert communicators that can instantly learn almost any process and perform complex tasks. Equipped with insight and quickly learned expertise, it is based on casual or verbal commands written in your preferred language. If they don’t like what they’ve made, they don’t complain and quickly tweak it or start over. Industry-specific models are specially trained and refined to tackle specific jobs.

Moreover, GPT-4 is already multimodal. We are beginning to demonstrate an uncanny ability to interpret images, audio and video and infer vast amounts of information from what we see and hear.

Moreover, while humanoid robots are certainly advancing, they are slowing down. It’s hard to imagine that in ten years there won’t be androids capable of versatile work duties. Share, watch, learn, and refine.

GPT-4 can already see pictures of cafeterias and tell blind people where the restrooms are and where the empty tables are. Embedded in a robot capable of effecting physical changes in the world around it, it’s hard to believe that even such rudimentary intelligence would struggle for so long with most physical tasks.

No one has yet attempted to put forward a general superintelligence timetable, the point at which AI becomes more capable than humans at most of all tasks. But looking at the steep rise in GPT, it’s hard to conclude that humans are so special that AI and robots can’t imitate them or do better. In the medium term, I think it’s well worth working on new problems and edge cases, but at least edge cases tend to diminish over time.

As a tech writer, writing is definitely a wall for my colleagues and me. This article took me all afternoon, but ChatGPT could have given him a summary of this report in less than a minute. At this stage, we choose not to use it. what does your profession look like?

Source: Goldman Sachs