circlehat Intention How will AI affect jobs? After “Will AI destroy humanity?”, this is the most important question about technology and it remains one that is extremely difficult to pin down, even as the frontier moves from science fiction to reality.
At one extreme there is the somewhat optimistic assertion that new technologies will simply create new jobs. At the other extreme there are fears that companies will replace their entire workforce with AI tools. The debate is often about the speed of the transition rather than the end state. A cataclysmic change that is completed in a few years is devastating to those caught in the middle, whereas a cataclysmic change that takes 20 years may be survivable.
Even the parallels with the past are not as clear-cut as we would like: the internal combustion engine eventually put an end to horse labor, but the steam engine, on the other hand, had a much bigger impact. increase Number of draft animals employed in the UK. Why? The arrival of the railways increased freight traffic in the country, but deliveries could not be completed from warehouse to doorstep. Horses were needed to do the things that steam engines could not do.
Until it isn’t.
Steam power and the internal combustion engine are examples of general-purpose technologies, breakthrough technologies that revolutionize the entire structure of society. There are not many such technologies, even if you count from writing, or even before that, from fire itself. It is pure coincidence that the initial letters of the term “Generative Pretrained Transformer” are the same, which is why GPT looks like GPT.
That’s not a job, idiot
Humans are not horses [citation needed]It seems hard to believe that AI technology will be able to do everything humans can do. Becoming HumanThis is an inconveniently circular argument, but an important one: horses still race, because if you replace horses with cars, it’s no longer a horse race. [citation needed]people will still provide the services they want for one reason or another, and as culture warps around the rise of AI, some of those services will teeth You might be surprised. For example, AI in healthcare is underrated because for many people, the “human touch” is bad The problem is the doctor who worries they are judging your drinking, or the therapist who lies to you because they want you to like them.
As a result, many people like to think in terms of “tasks” rather than jobs: take a job, define it in terms of the tasks it contains, and ask whether an AI can do them. In doing so, we can identify some jobs that are at risk of being completely cannibalized, some jobs that are perfectly safe, and a large intermediate group of jobs that will be “impacted” by AI.
It’s worth pointing out an obvious fact: this approach results in a higher number of jobs that are mechanically “influenced” and a lower number of jobs that are “destroyed.” (Even the jobs most influenced by AI are likely to have some tasks that the AI finds difficult.) That may be why the technique was pioneered by OpenAI, who in a 2023 paper wrote: The researchers in the lab:“80% of workers are in occupations where at least 10% of the work requires a law degree, and 19% of workers are in occupations where more than half of the work requires a law degree.”
The report claimed between 15 and 86 professions were “completely at risk”, including mathematicians, legal secretaries and journalists.
I’m still here. But a year on, the idea is trending again, thanks to a paper from the Tony Blair Institute (TBI). The giant think tank, powerful and influential even before Labour’s landslide victory two weeks ago, is now seen as one of the architects of Starmerite thought. And it believes the public sector is ripe for disruption through AI. According to the TBI paper: The potential impact of AI on the public sector workforce (pdf):
More than 40% of the tasks performed by public sector workers could potentially be partially automated through a combination of AI-based software, such as machine learning models and large-scale language models, and AI-enabled hardware, ranging from AI-enabled sensors to advanced robotics.
Governments will need to invest in AI technology, upgrade data systems, train employees to use the new tools and cover the redundancy costs of early retirement – costs that are estimated to amount to £4 billion under ambitious implementation plans.That averages $1 billion a year for the term of this Congress.
Over the past few weeks TechScape has been keeping a close eye on the new Government’s approach to AI. Tomorrow, the King’s Speech is expected to announce the AI Bill, and we will hear more. The TBI paper makes one takeaway worth watching: Will investment in transformation approach £4 billion a year? There is a lot that can be done for free, but much more could be done with more money. The institute estimates that spending would return more than nine times, but a £20 billion bill would be hard to get through Parliament without question.
AI Geek
The report drew renewed attention over the weekend as critics took issue with its methodology. From 404 Media:
The problem with this prediction is that POLITICO, Technology
Breaking down work into tasks is already done by a huge database created by the US Department of Labor. But with 20,000 such tasks, describing which ones should be exposed to AI is a daunting task. In a similar paper from OpenAI, “the authors personally labeled a large sample of tasks and DWAs, and hired experienced human annotators who reviewed the output of GPT-3, GPT-3.5, and GPT-4 as part of OpenAI’s tuning efforts,” but they also had the then-new GPT-4 perform the same tasks and found a 60-80 percent match between robots and humans.
After newsletter promotion
Source: www.theguardian.com
