Panel on Jobs and AI at MIT
John Werner
It’s important to think about the eventual impact of what’s happening right now with AI. There’s actually a lot of discussion about this within the MIT community, and elsewhere, as spring comes on this year, and we see model progress blossoming along with the flowers and trees.
One example is the recent AI and Society Forum, a product of the MIT Schwarzman College of Computing, which I was privileged to be present at, to listen to a panel on AI and jobs, with some of the best voices on artificial intelligence and its ramifications for human societies.
Rob Loughlin interviewed David Mindell, an MIT professor of the history of engineering and manufacturing, Schwartzman economics professor Sendhil Mullainathan, and Daniela Rus, director of MIT’s CSAIL Lab, about a topic many people are talking about: AI and jobs.
Loughlin asked the panel about when each of them became aware of the massive shift toward AI power.
“I think it's a little bit like boiling a frog at some level, because it's hard to know the moment at which it became transparent,” Mullainathan said, noting that he remembers hearing about early model advances from his father.
“I guess the thing that really persuaded me that we're in the presence of something genuinely new was so many colleagues at MIT beginning to apply these tools to their own research,” Mindell said.
MORE FOR YOU
For Rus, it was when someone from the business world came to CSAIL in 2023, to demo GPT4.
“It was stunning,” she said. “We were all amazed by the kinds of capabilities and experiments that he showed in his talk. He actually had access to GPT-4 six months before the rest of the world, and he shared with us extraordinary capabilities in coding, in game playing, in language interactions. We were all amazed.”
As for benchmarks, Mindell mentioned recursive intelligence and robotic creation as a watershed moment.
“Something to watch out for is, when are the humanoid robots being made by humanoid robots?” he said.
He also spoke about the idea of task-evaluation, versus a broader analysis of embodied AI.
“The physical AI world is still very focused on these mechanical tasks,” Mindell said. “But work is broader than tasks. It's this complicated linkage between tasks. There are many relationships on a shop floor. It's a very complicated social environment. There are authority relationships, there are economic relationships, there are shifting alliances.”
In terms of human response, Mindell suggested there’s a growing awareness of what AI robotic systems can do.
“People have not only a visceral, but a very practical sense of what these tools do for their daily lives,” he said.
“I think it's very much worth differentiating productivity gains from things that actually drive long-term growth,” Mullainathan added. “In that sense, for me, I think of there being quite a big difference between technology in whatever form factor it takes, whether it's an LLM or the technologies that expand the frontier of what we know, versus technologies that kind of let us do what we know more efficiently.”
Here’s how Mullainathan described the appeal of robots on an emotional level, for humans.
“You have this thing that talks like a person, that does things a person does,” he said. “It's just begging you to project everything about a person onto it.”
Mindell mentioned something similar, in a way, about the early train engine.
“When locomotives came out, people called them ‘iron horses,’ because it felt like they were going to do the horse's job a little better,” he said.
That kind of characterization, of cold digital AI, is probably going to factor into what things look like in a few years. The panelists seemed to agree about that.
“We also see a lot of productivity consequences of what we have right now,” Rus added, observing what’s happening at MIT CSAIL. “And this is very tangible. We can really feel it. My students tell me they write code faster, and that's actually a good thing, because we are getting to the scientific discovery, and to the advancement of knowledge.”
Part of the value, she suggested, is speed.
“Those standardized experiments no longer take weeks,” Rus said. “They take hours, because some of the routine code can be written much faster.”
The panel was on AI and jobs, so the panelists also covered the idea of job displacement.
“Jobs that synthesize knowledge, or that depend on doing knowledge synthesis, might have a question mark next to them,” Rus said.
Mindell posited the death of “office work in general.”
“9-5 white collar office work, which is only 150 years old to begin with, was kind of invented as the human-driven AI of the corporate world,” he observed, noting that the same type of automation process will take away quite a few human jobs.
Mullainathan thought knowledge work might get the axe, while noting that there’s a broader impact spectrum coming along the pike.
“There's such variety in the nature of tasks,” he said.
Asked about projections, Mullainathan said he doesn’t like to do too much prognostication, but offered what he called an “anodyne prediction.”
“If you said exactly how organizations restructure, I don't know, but is there going to be a lot of restructuring? It's hard to believe there isn't going to be a lot of restructuring,” he said.
Mindell mentioned the need to hand over the keys to a new generation of humans, who might, due to their AI-native experiences, be better able to make the right calls.
“It's absolutely imperative that we give the tool to the young people, and let them do what they find creative with all of this, and show us what the new work is going to be, because we're too late for us to figure that out, and so I very much agree with fostering, creating, celebrating new work and supporting them,” he said.
Rus agreed.
“The students absolutely have to learn the tools, because the jobs of the future, of the not so distant future, the jobs of the now, really require the tools,” she said.
Advice for the New Grad
It’s springtime, and the panel had some words for this year’s crop of emerging career professionals.
“You're less likely than generations before, to go into a single job and stay in that job, maybe even into a single profession, and stay in that profession over the course of your career,” Mindell said, urging youngsters to learn more about a broader set of systems.
“I just honestly feel very bad for young people today, genuinely,” Mullainathan added, speaking to the changes in media that impact our attention spans and much more. “When I was young, I had a TV that had, I think, eight television stations; two of them were PBS, and most of them had Gilligan's Island and things like that, which aren't particularly interesting shows,” he said. “Flash forward: imagine being young and having this box in front of me, with endless distraction and entertainment. That's the thing that has been unleashed on people, and it’s made people's psychological lives incredibly hard.”
And then, he added, there’s AI.
“We've kind of created this incredible rugged environment for young people, where they are left in a psychological environment, where they have to figure it out for themselves,” he said. “I think we need to do something about that.”
In the end, panelists mentioned being prepared for change, with a diversity of skills, and a proactive response to the biggest technological changes any of us have ever known. This was a ground breaking conversation.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。