In office buildings in cities across Africa, tens of thousands of employees scroll through pages, checking countless images, reading transcripts, and watching videos. .
This little-known tech army in hubs like Accra in Ghana and Nairobi in Kenya represents the power of the brain that has helped create the supernatural human responses of AI tools like ChatGPT.
Thousands of jobs have been created across the continent as data markers — digital workers who spend hours checking whether an AI bot has correctly determined whether an image of a dog is a dog and not a cat.
They read millions of pages of digitally transcribed text for accuracy, listen to recordings of the voice assistant and double-check surveillance camera warnings, correct errors and improve answers.
One such Chinese data labeling company employed 300,000 people, according to The Economist. people, carefully annotating the curriculum from which AI algorithms can learn.
Data labeling professions are rapidly becoming part of the new machine learning economy around the world. By some estimates, this is already a multi-billion dollar component of the AI industry.
But at the same time, there are fears that these machines could make millions of roles redundant. Analysts at Goldman Sachs have calculated that thanks to a new wave of artificial intelligence, robots will soon be able to perform 300 million jobs.
OpenAI CEO Sam Altman predicts AI could replace customer service work. Photo: JASON REDMOND/AFP via Getty Images
Sam Altman, CEO of OpenAI, the Silicon Valley company that develops ChatGPT, warned last week: “Customer service is one category in which, frankly, I expect that many of these vacancies will simply disappear.”
While fears that robots could replace millions of jobs have long been the stuff of science fiction, tech bosses are increasingly betting that the latest generation of artificial intelligence algorithms will make this a reality.
Arvind Krishna, chief executive of IBM, went the furthest earlier this month when he said that up to 30% of back-office activities such as HR could be done with AI. He said that IBM will stop hiring employees for roles that a machine can fill instead, affecting up to 7,800 jobs.
Enterprise adoption of these tools has accelerated as ChatGPT grew in popularity and millions of Internet users tried to create chatbot with digital intelligence and experimented with it to speed up their work.
ChatGPT and similar AI models can provide seemingly human answers to questions in plain English. The bot can be asked to write natural-sounding emails, articles and even essays, or chat with users in a Q&A style.
OpenAI technology has been trained on a vast database of texts and books to provide realistic human responses. But behind the scenes, it was tuned and corrected by human processors.
Other artificial intelligence tools have been developed that use computer vision — machines that can «see» the world around them — necessary for self-driving cars or autonomous robots. Again, people helped train these algorithms by labeling the images and correcting errors, which is known as “reinforcement learning.” Sama, the artificial intelligence company that pioneered this work, at one point referred to these workers as the «soul of AI».
Even those who work in the field of machine learning cannot deny that this work can be monotonous. Andreas Heindl, machine learning product manager at computer vision company Encord, admits that working with data labels can be «very tedious and time consuming…hands and fingers start to hurt after an hour.»
But proponents of a new wave of AI tools insist that innovations like ChatGPT will enable people to do more interesting work.
0605 The impact of AI
Tech giants Microsoft and Google have added a suite of AI-based tools to their products for workplaces.
On Wednesday, Google demonstrated how Bard, its new artificial intelligence robot, can be used to write job listings with a simple one-line invitation. Microsoft has demonstrated how a tool based on ChatGPT can write entire articles in Word.
“There are a lot of sales reps out there who do a lot of mundane work of writing emails to prospects,” says Rob Seaman, senior vice president of workplace messaging company Slack, which is working with OpenAI to implement ChatGPT into its application as a a sort of digital counterpart.
New AI tools could take away some of the most tedious aspects of these roles. But based on past experience, technology also threatens to create an entirely new class of secondary roles.
Take the rise of social media. To keep up with the skyrocketing growth of Facebook, Instagram, and TikTok, the tech giants have been forced to outsource hundreds of thousands of jobs to content moderators, whose job is to weed out the worst images of abuse, violence, and terror to keep our feeds clean.
Meanwhile, data labeling has created jobs for tens of thousands of people, mostly in low-income countries. «Unsurprisingly, it's much more economically beneficial where wages are lower,» says James Neave, head of data processing at job site Adzuna.
But sometimes such work can sound like something out of Netflix's dystopian series Black Mirror. Last year, the MIT Review reported that Venezuelan data-tagging specialists at iRobot, a vacuum company, began posting images online from customer bots, including images of people in a toilet. A spokesperson for the company said that at the time, the company canceled a deal with a labeling company.
Of course, the development of AI has created many high-paying technical positions.
Along with the in-demand AI coders and As data scientists, the new role of «operations engineer» has led Google-backed AI company Anthropic to announce a salary of up to $335,000.
A A prompt engineer is a person trained in the art of instructing an AI, teaching it to give the correct answers to a given problem. Using plain language, the hint engineer assigns tasks to an AI such as ChatGPT, modifying the input slightly to achieve different results.
But other roles are quickly being supplemented by powerful AI bots.
Is AI moving too fast? Poll
Customer service jobs may be at greatest risk as chatbots take over. Those who are not replaced by automation may be augmented by technology. In some cases, companies are changing the focus of workers with AI software to make it sound like someone from the US rather than outsourcing overseas.
Paul Flemming, general secretary of Equity actors union, last week told MPs that voice actors were being replaced by radio advertising machines. “It affects employment and labor movement,” he said. «This opportunity to earn is shrinking.»
Journalists already fear the worst. CNET, a tech news site, quietly used automation technology to publish dozens of articles about credit cards and payments, although the program was later put on hold. BuzzFeed said it will use ChatGPT to create quizzes.
In the UK, local publisher Newsquest is looking for an «AI-powered reporter» who will use automation technology to «create national, local and hyper-local content for our news brands» .
Adzuna's Niv says there are a number of jobs that will be «gradually replaced». Robots can take over administrative and legal work.
The tech giants have also used their robots to master the art of programming. Bots can write code, with varying degrees of accuracy. Last week, Google named programming as a key benefit of its new Bard chatbot.
Neave believes they are unlikely to completely replace engineers, in part because of the huge demand for such skills. But bots can act “almost like a coder assistant who works with you.”
“This will basically raise employers’ performance expectations.”
The current wave of AI tools also often allows egregious errors and cannot distinguish fact from fiction. Even if AI can speed things up, it needs a human to test it out.
Efforts are underway to further optimize some of the more repetitive tasks that AI tool development has led to. Encord, for example, is developing software that can process a much smaller set of labeled data and expand it into millions of pieces of information, reducing the need for hours of human labor.
“Labeling everything by hand is no longer realistic,” says Heindle from Encord.
But with the explosive demand for new and even more magical machines with artificial intelligence, the need for an army of human processors is still almost not reduced.< /p>
Свежие комментарии