Jana Eggers is the CEO of Nara Logics.
While the most talked about possible impact of artificial intelligence is killer robots, jobs — specifically job elimination — is close behind it at #2. So, when asked to moderate a panel on the economic impact of deep learning at the (fantastic!) Deep Learning Summit by Re•Work, I knew jobs would be the key topic. What I didn’t know is how wide and deep the discussion would get in half an hour. Here are my top three learnings from the panel and conference attendees:
1 - Jobs will change.
Han Shu of Airbnb raised the importance of people seeing the job shifts in advance. For example, obviously, more machine learning experts will be needed. Also consider though what shifts happen when we provide a product or service more efficiently. Like say, providing places to sleep, Han? ;-) This increased efficiency is likely to create more rented beds in total, and therefore, while venues might change – hotels to homes – many services, e.g., cleaning and maintenance, will still be needed and likely at an increased rate… this rising tide raising those boats. There will of course be jobs that are automated. And I don’t want to downplay that. The consensus at this summit was that most nearer-term automation will be of the repetitive, clearly scoped tasks, leaving problems requiring a bit more creativity and judgement for us mere humans to solve. Typically, I’ve judged this on two scales:
- Individual level. Few people, across job level, type and industry, get everything done they need or want to get done. Automation will help. For example, robot scientist “Eve” will replace some jobs in the lab, but will robots alone be able to find a cure for cancer? No. We will use the time and information given by Eve to scale the next mountain in finding that cure and others.
- Societal level. As we shift to human creativity from human labor being needed for jobs, we still have work to do. Some jobs will be eliminated all together. Driving as a job, as one possibility. My biggest hope – my goal really! – is that we as a society change our values and value other problems, i.e., change the economic value to jobs for problems not solved yet. For example, what can the “drivers” do to solve homelessness or hunger? We have so many problems in our world, both big and small, I see an opportunity to shift what is valued as a job, as an opportunity for some profound, positive change in our society.
2 - …More slowly than you might think.
While we nerds are excited about the advances that deep learning has made in the past 5 years, we have a long way to go. This is why I asked Naveen Rao of Nervana Systems to give his thoughts on the hype versus reality. Naveen’s company helps enterprises reach new levels with hardware and software systems that accelerate deep learning progress, so he knows well the current state. He made this reality clear with the simple example that we are a long way from being able to build individual healthcare models for even millions of people versus the 7 billion on Earth now; and that’s one example in one industry. My summary: We have a few of the key aspects of this new “rocket science” understood, but we haven’t launched the rocket yet and we certainly haven’t figured out what’s needed for the space program we need to build.
3- And likely more targeted than you think.
Priya Vijayarajendran of SAP’s Innovation Center in Silicon Valley brought up the impact of adding cognitive abilities to many applications we use today. Her examples from the work SAP is doing now for customers hit exactly on the problem stated here:
“Our management data came in the form of a 700-page report of financial line items. We spent 90% of our time…trying to assemble the data and ask the right questions — and 10% on the issue. Those proportions needed to be reversed.” - CEO, global consumer products company
The AI solving this won’t be “thinking like humans”, but parsing, assembling, and calculating faster than humans. The tasks will be specific to each application, like marketing metrics, ad purchasing, financial analysis, sales assistance, or the AI favorite, image recognition. We are already seeing a wide variety of machine learning being applied in these areas. Expect this to continue, and continue to improve. Don’t expect a simple or quick cross over to artificial general intelligence from these targeted applications.
Hopefully the perspective here helps us see that (1) there is a job shift coming, but (2) there’s time to prepare and (3) the shift will happen in areas targeted for efficiency first. The question of how to prepare for this future is next: what should kids and transitioning adults study now? While computer science was definitely at the top of the list for this panel, several of us put in a plug for mathematics, for its ability to teach problem solving. (GO MATHS!) AT&T’s CEO Randall Stephenson is expecting employees to retool themselves with 5 to 10 hours of online training each week. (Full, fascinating story here.) Whatever your situation and approach, from the above discussion, I hope you see many opportunities as the landscape changes. The keys: look for the training or major that enhances your creativity and stay current on technology!
P.S. Coming soon is an interesting discussion that was also part of our panel: Will deep learning’s substantial data requirements increase or decrease the gap between the have’s and have not’s? Or change who falls in which bucket?