For an external observer, AI researchers are enviable. They are required by high -tech giants. They are bringing a remarkable salary home. And they are in the hottest industry at that moment.
However, this is all accompanied by strong pressure.
TechCrunch’s more than half -dazzers spoke, and some of them were afraid of retaliation and demanded anonymity, saying that the fierce pace of the AI industry had hit mental health. The fierce competition between AI Labs has promoted an isolated atmosphere, but the rising bets ratch a stress level.
“Everything has changed in effect overnight,” said a researcher, “Our work -positive results and negative results are measured by exposure to products and economic results. It has a big impact.
Just in December, Openai held 12 live streams, during which 12 new tools, models and services. Google responded using its own tools, models and services, with press releases, social media posts, and blogs that look at the eyes of blogs. Bearing between the two high -tech companies was worth noting at that speed. This is the speed that researchers come at a sudden cost.
Grind and hustle
Silicon Valley is not a stranger to hustle culture. However, with the AI boom, the general support of overwork has reached a troublesome height.
In Openai, it is not unusual for researchers to work six days a week. CEO Sam Altoman is said to promote the company team to turn break -through into public products on a strict timeline. Openai’s former chief researcher, Bob McGrew, is said to have quoted burnouts as one of the reasons for leaving last September.
You cannot find it in a competing lab. The Google DeepMind team is developing GEMINI, a flagship series of Google AI model. At one point, it rose to 120 hours from 100 hours a week to fix bugs in the system. ELON MUSK’S AI Company XAI engineers are posting regularly about the night of bleeding work early in the morning.
Why is there a relentless push? Today’s AI survey can have a significant impact on corporate profits. Google Parent Alphabet has lost a market value of about $ 90 billion in the aforementioned bug, creating a depiction of Google’s Gemini Chatbot in controversy of historical people.
“One of the biggest pressure is competitiveness,” said Kai ARALKUMARAN, a research leader of AI service provider ARAYA, “combined with a quick time scale.”
Above all, leader board
Some of this competition is very public.
Every month, and every week, AI companies place guns on guns to replace each other with leader boards such as ChatBot Arena, which ranks AI models in the entire category, such as mathematics and coding. Logan Kilpatrick, a leading product of some Google Gemini developer tools, said in X’s post, “AI has a trivial effect on the speed of AI development.”
Not all researchers are convinced that it is good. He says that the speed of the industry is finding their job as the risk of obsolete before they ship.
“This has a lot of questions about their work value,” said Zihan Wang, a robot engineering engineer who works for Stealth AI startups. “What is the meaning of what I am doing if someone is very likely to go faster than me?”
Other researchers lament that focusing on commercialization has sacrificed academic friendship.
“One of the fundamentals (the cause of stress) is the transition from providing product solutions and providing solutions because AI researchers pursue their own research agendas in the industry.” Arulkumaran said. “The industry has set the expectation that AI researchers can pursue academic research in the industry, but this is no longer likely.”
Other researchers are greatly about their surprises and pain -open collaboration and discussions on research are no longer standard in the industry, except for some AI labs that have accepted the openness as a release strategy. 。
“We are now more and more focused on commercialization, closing, and execution,” said the researcher, “without contributing to the scientific community.”
Execute Grad Gauntlet
Some researchers track anxiety into the AI graduate program.
Gowthami Sumepalli, a doctoral student student who studies AI at Maryland University, states that it is difficult for graduates to distinguish the development with meaningful Fad because their research has been issued very quickly. Ta. It is very important that SOMEPALLI has seen AI companies giving the candidates more and more priority in “very relevant experiences.”
“The Ph.D. is generally a very isolated and stressful experience, and the Dr. Machine Learning is particularly difficult for the rapid progress of the field and the mentality that” public or perish “,” said SOMEPALLI. “If many students in your laboratory have published four papers while publishing only one or two papers a year, they may be particularly stressful.”
SOMEPALLI said that after the first two years of the graduate program, she felt guilty to escape before publishing his research, so he stopped taking a vacation.
“I always suffered fraudulent syndrome during my doctorate and almost dropped out at the end of the first year,” she said.
Forward road
So, what kind of changes can be promoted a less punishment AI working environment? It is difficult to imagine that the development pace is decelerating.
SOMEPALLI emphasized a small but influential reform, as he talked about his challenges.
“One of the biggest problems is that no one has opened openly discuss his struggle. Everyone has a brave face,” she said. “I believe (people) may feel better if they can see what others are struggling.”
Professional Services Company’s AI consultant Bhaaskar Bhatt says that the industry should work to build a “robust support network” to fight isolation.
“It’s essential to promote a culture that values work -life balance that can truly separate your work,” said Bat. “The organization should develop a culture that values mental happiness as much as innovation with specific policies such as reasonable working hours, mental health days, and access to counseling services.”
OFIR PRESS, a Princeton postdoc student, has proposed a one -week “suspension” with Paper Submissions so that researchers can take a break from tracking new works. Raj Dabre, an AI researcher at the National Institute of National Information and Communications in Japan, said that researchers should be calmly remembered.
“We need to educate that AI is just a job,” said Dabre.