Artificial intelligence: understand the impact on jobs – 07/12/2023 – Tech

In June, Febraban started airing a commercial made entirely with artificial intelligence, from the script to the audio. A 30-second insert like this usually involves the work of eight people, but it was done with just two and the help of four AI platforms.

Recently, Volkswagen also used technology to recreate the image and voice of Elis Regina, who died 41 years ago, in a commercial.

These technologies are, in general, widely accessible and cost modest subscriptions for advertising producers. The paid version of ChatGPT costs US$ 20 per month and the Midjourney subscription varies between US$ 10 and US$ 60. Not to mention that the platforms deliver texts and images in seconds.

The US employment agency Challenger, Gray & Christmas calculated that artificial intelligence contributed to the elimination of 3,900 jobs in the US in the month of May alone. In the month, the United States accounted for 80,000 layoffs.

Administrative and legal tasks will be the most affected – more than 40% will be automated – and construction and maintenance jobs are the least vulnerable to AI, according to a study by US bank Goldman Sachs economists Joseph Briggs and Devesh Kodnani. They estimate that AI could wipe out 300 million jobs over the next 10 years.

Class groups concerned about the deterioration of their jobs have already mobilized. This is the case of the WGA (West of America Screenwriters Union), which has been on strike since May 3.

The entity claims that the use of AI is regulated in the production of series and films. The authors ask that the technology not be used for writing literary material or as a source of information. Also, that movie and series scripts are not used to train natural language models, such as ChatGPT and the like.

The director of the technology group at Abra (Brazilian Association of Screenwriters), Marta Nehring, says that the WGA’s claims go beyond salaries and concern freedom of creation. “If everything turns into the market and who delivers more, we will move towards more and more generic productions.”

Another concern of Abra, aligned with the WGA, is the protection of copyright. For the entity, AI should be treated as a tool available to the screenwriter’s creative choices. “All of this must be correctly credited”, says Abra’s vice-president, André Mielnik.

“If we accept productions that do not credit the sources, ask for authorization to reproduce and pay the royalties, we will be left with nothing. The name of this is plagiarism, which is theft”, says Nehring.

Experts say that the introduction of AI in production chains is unavoidable. One of them was Oxford researcher Callum Cant, who coordinates the Fair Work for AI project (decent work in artificial intelligence).

For him, it is necessary to have minimum conditions to prevent work from being treated purely as a commodity. In a large part of the world, workers do not have the right to receive explanations about the purpose of their work or the production chain in which they participate. They are also at the mercy of the employers’ choice of which technologies will be adopted in the work routine.

Cant cites the Scandinavian countries as an example, where employees decide which innovations make sense in the workplace since the post-war period, when automation in industry flourished.

The researcher says, however, that it is too early to assess what the effect of AI will be in terms of working hours and remuneration in the set of jobs. “Everything should vary, depending on the area in which the business operates and its national context.”

The implementation of artificial intelligence in the job market is still in its infancy, according to technology innovation analyst Nicole Sarvasi, from startup FinTalk, which sells customer service solutions.

When serving a customer, for example, a chatbot cannot hallucinate and make offers that are incompatible with the company it represents or offend a customer. Companies like FinTalk constrain AI behavior with reinforcement training and narrowing down the speaker’s possibilities.

The very chatbot praised by Bill Gates, Pi, from InflexionAI, claims that generative artificial intelligences are in their infancy and that expectations must be adjusted.

On the other hand, the New Edge Wealth investment fund estimates that new artificial intelligence technologies will open up a global market valued at US$ 1.3 trillion next year. Innovation, in general, generates new jobs.

Among current workers, 60% are in professions that did not exist before automation processes, says a report by Goldman Sachs. 85% of the jobs created in the last 80 years were in innovative positions.

USP professor Marcelo Finger, also director of C4AI (Center for Artificial Intelligence, in free translation) says that adding quality annotations to raw information will be crucial to developing more accurate computational models.

Such work can be done wholesale and retail. When a finer treatment is required, specialized professionals carry out careful analyses. In situations where quantity matters most, freelance platforms distribute millions of temporary jobs, especially in countries of the global south.

The largest of these sites, called Clickworker, assembles an army of more than 4.5 million data labelers. The SoyFreelancer website delivers various services, including those related to AI training, and has 23 million subscribers.

Brazil already has, for example, 54 platforms that offer services to record data used in artificial intelligence training, according to a study by the State University of Maringá (UEM), led by Professor Matheus Viana Braz. According to the professor, the problem is the current lack of protection of workers working in the area.

Ensuring these jobs withstand the momentum of technology is one of the concerns of Callum Cant’s Fair Work for AI project. The project also recommends that employees receive clear information about each task.

Professor Julian Posadas of Yale University says that decent work on data labeling is key to avoiding biases and biases in artificial intelligence models. “Platformized workers we interviewed did not receive proper instructions, which can lead to many errors in the final result.”

Posadas assesses that large companies will now have to increase the quality of existing data, as current models already work with almost all available information. “It is still too early to guarantee that this will guarantee better conditions for workers.”

The president of Intel in Brazil, Claudia Muchaluat, says that the company will focus on investments in intelligence in the country and in the rest of the world, but that it does not offer opportunities in the area at this time. The statement was given to the report at an AI training event by the multinational with teachers from the Paula Souza Center, which manages the Etecs and Fatecs in the state of São Paulo.

Intel’s Youth for AI program aims to train 30 million young people around the world, thousands of them in Brazil. The training, however, still serves technical students in the technology sector.

In the long term, the idea is that 60% of the benefited students are computer science and technology students and 30% from diverse areas, according to Emílio Loures, director of corporate affairs at Intel do Brasil. “There is no data economy without AI. If we don’t know how to program, what we will produce here will be exploited abroad by the same players as always.”