The direction of the AI ​​industry has determined the researchers

For outside observers, AI researchers are in a good position. They are wanted by the modern giants. They are taking home pay. And they are in the hottest industry at the moment.

But all of this comes with a lot of pressure.

More than a half-dozen researchers TechCrunch spoke to, some of whom asked not to be named for fear of backlash, said the company’s AI rush has taken a toll on their mental health. Fierce competition among AI labs has created an environment of isolation, he says, while the slow pace has added to the pressure.

One researcher told me: “Everything has changed suddenly, and it is our job – good and bad results – to have great results as measured by things like exposure, and financial results.

Last December, OpenAI did 12 livestreams where it announced more than a dozen new devices, models, and services. Google he answered and equipmentmodel, and work his that disrupts publications, social media posts, and blogs. The back-and-forth between the two tech giants was surprising because of its speed — speed that analysts say comes at a high price.

Grinding and enjoying

Silicon Valley is no stranger to multiculturalism. With the rise of AI, however, people’s acceptance of overwork has reached a critical point.

On OpenAI, that no unusual for researchers to work six days a week – and have passed the layoff. CEO Sam Altman says he is pushing the company’s teams change achievements to public goods in difficult times. Former CEO of OpenAI, Bob McGrew, he says He mentioned that one of the reasons he left last September was because he was too tired.

There is no relief available at competing clubs. The Google DeepMind team that develops Gemini, Google’s popular series of AI models, once dropped from working 100 hours a week to 120 hours to correct the error in the system. He is an engineer at xAI, Elon Musk’s AI company, frequently post of working days that bleed into the wee hours of the morning.

Why does he keep pushing? AI research today can have a huge impact on a company’s bottom line. Google Parent Alphabet lost about $90 billion in market value than the aforementioned error, which caused Google’s Gemini chatbot to generate conflicting images of ancient people.

“One of the biggest challenges is competition,” said Kai Arulkumaran, director of research at Araya, “combined with fast turnaround times.”

Blackboards above all

Some of these tournaments are played in public.

Every month – and sometimes every week – AI companies compete on leaderboards like Chatbot Arena, which ranks AI models in categories like math and writing. Logan Kilpatrick, who leads several Google Gemini development tools, he said in a post on X that the Chatbot Arena “has had an unwanted impact on the development of AI.”

Not all researchers believe that is a good thing. The company’s approach is that, they find out that their work is in danger of being out of date before they can send it.

“This makes many question the value of their work,” said Zihan Wang, a robotics engineer who works at an underground AI startup. “If there is a great chance that someone will go faster than me, what is the meaning of what I am doing?”

Some researchers complain that the focus on marketing has resulted from academic friendship.

“One of the causes (of stress) is the transition of AI researchers from their research activities in industry to working (in AI models) and providing product solutions,” Arulkumaran said. “Industry set the expectation that AI researchers could conduct academic research in industry, but that is no longer the case.”

Another researcher said that – to their concern and stress – open cooperation and research discussions are no longer the norm in the industry, outside of a few AI labs that have embraced openness as a means of translation.

“Now they are focusing on commercialization, promoting closed sources, and killing,” said the researcher, “without giving back to the scientific community.”

Running the grad gauntlet

Some researchers trace the seeds of their concern to their AI grad programs.

Gowthami Somepalli, Ph.D. student studying AI at the University of Maryland, said that research is being published so quickly, it has been difficult for grad students to distinguish between fads and meaningful development. That’s important, Somepalli said, because he’s seen AI companies prioritize people who have “the best knowledge.”

“A Ph.D. are often isolated and stressful, and the machine learning Ph.D. it’s especially difficult because of the rapid progress and the ‘publish or perish’ mentality,” said Somepalli. “It can be very stressful when many students in your lab are publishing 4 papers while you’re only publishing 1 or 2 papers a year.”

Somepalli said that, after the first two years of his studies, he stopped going on vacation because he felt guilty for leaving school before graduation.

“I always suffered from hypocrisy during my Ph.D. and I almost dropped out at the end of my first year,” Somepalli said.

The way forward

So what changes, if any, would make for a less punishing AI workplace? It’s hard to imagine how development is slowing down – not with so much money at stake.

Somepalli emphasized small but effective changes, such as explaining your problems.

“One of the biggest problems (…) is that no one talks freely about their problems; everyone wears a brave face,” he said. “I hope (people) can feel better if they can see that others are suffering.”

Bhaskar Bhatt, an AI consultant at technology firm EY, says companies should work to create “support networks” to combat isolationist attitudes.

“Fostering a culture that respects work life, where people can leave their jobs, is important,” Bhatt said. “Organizations need to foster a culture that values ​​mental health as it is new, with tangible policies such as adequate working time, mental health days, and access to counseling.”

Ofir Press, a postdoctoral fellow at Princeton, organized smaller AI conferences and week-long “pauses” at the paper so researchers could take a break from pursuing new work. And Raj Dabre, an AI researcher at Japan’s National Institute of Information and Communications Technology, said researchers need to be gently reminded of what’s important.

“We need to teach people from the beginning that AI is just a job,” Dabre said, “and we need to focus on family, friends, and the finer things in life.”