Andrew Ng on AI's Potential Effect on the Labor Force | WSJ

00:01 AI will lead to a massive productivity boost and create new job roles. 02:13 AI is disrupting job roles like call center operations and sales back office. 06:31 AI requires responsible and safe usage with training. 08:37 AI technology continues to improve, reducing error rates and increasing usability over time. 13:04 Disruptive technology is creating new powerful companies, particularly in the cloud business. 15:08 Open-sourcing AI provides free tools for innovation and independence from proprietary platforms. 18:53 Consider using multiple providers for AI implementation to optimize solutions. 20:25 Discussion about semiconductor shortage and progress of AMD and Intel 23:51 More intelligence in the world is better for progress and open source is a positive contribution. 25:31 The fear around AI's impact on humanity is diminishing, focusing on concrete applications and regulation. 29:16 Corporations are enthusiastic about AI, and non-technical understanding unlocks brainstorming. 31:11 AI automation can improve productivity and allow focus on other tasks.

I tutor Python online. Last year: $1200/month. So far this year: $200/month. Fortunately I’m not heavily dependent on that income, but I imagine some tutors are. Or were.

Consider carefully the implications of AI replacing tasks that a person does on a routine basis. This does not mean each person will keep their job and just do more interesting higher level work. It really means fewer people will be utilized to do the same amount of work or even more work. Productivity will mean a pool of 10 people will be reduced to 1 or 2 doing the same amount of work currently being done now, just differently. This reminds me very much of my early office experience and what I witnessed with the introduction of personal computing and later the dawn of the internet. When I started work, each executive in my company had a secretary. There were also file clerks supporting the C-suite. The support staff numbered 20 in a company with about 15 executives. This was 1989. Within a decade the File Clerks were gone and rather than each executive having a secretary, only the CEO had a dedicated individual, the rest of the VPs and Senior VPs shared individuals - usually 3 or 4 executives to a single secretary or AA. Today, the pool of assistants is even smaller. One AA still is dedicated to the CEO, but only 1 or 2 AAs service the remaining 14 executives. How did this happen? Word processing, scheduling software, online travel bookings, voice mail, the collapse of dictation and inter-office memos replaced by first email and then messaging, later zoom, etc.. So, bottom line -- YES, tasks will be eliminated just as they predict. But what they aren't saying is that JOBS will be eliminated as well. Using my Administrative Assistant example, you will end up with 1 person currently doing the job of 5 or 10 people. Those other jobs aren't being replaced. You don't need as many people when the AI can do so many tasks. So, where a factory once needed 500 people, it might need only 15 or 20 in future to monitor the equipment and machines. You can apply this across many other industries. The biggest problem up to this point has been the loss of 'entry-level' middle class work. By that I mean jobs that needed some education and provided a small toe-hold in the middle class with white collar work -- basically desk work in a nice office with modest pay, health benefits and maybe a 401K. Not a huge salary, but not physical labor. Now, however, we are looking at AI moving UP the food chain in skills, destroying not just the AA jobs but middle management and eventually the COO, the CFO and all the accountants, plus even the CIO and all his or her minions, leaving maybe a small team of say 2 or 3 executives (wild guess) plus 1 support person, and all other humans in the org being either replaced and reduced to a handful monitoring the machines. We are looking at massive income inequality, the final collapse of the lower, middle and even upper middle class and the emergence of a small hyper-rich cognitive elite. That's the short term. Long term?? Who knows. The final piece of the puzzle here is this conundrum. While technology has spawned a host of new jobs, those jobs have required an increasingly high level of cognitive ability and education. Unfortunately our population only has a limited pool of people with IQs that can continue to acquire skills at a rate fast enough to adjust, adapt and do those jobs. Bottom line: Society needs to find work for a lot of people (at least the middle 60% of the population) that have only upper middle, middle and lower middle IQs. What are these folks going to do? Sure, physical labor is possible for some, but not all, and the real hit will be to the group of people that are 'average or above average' in intelligence but not smart enough to adapt. These are the people that 50 years ago would have ended up in middle management or some respectable, but not senior role in an organization. They are going to be wiped out and impoverished.

I have a lot of respect for Andrew Ng, but I've been seeing a lot of doublespeak from people in the field recently, with the phrasing of "it can replace humans in certain kinds of work" swapped with "it just does tasks". It sounds more diplomatic. Lots of AI people don't want to outright say that they would think it would make everyone happier in the long run if we replaced almost the whole shebang. I mean I think we would be happier.

Andrew is so reasonable, but it seems like he's really doubling down on normalcy bias.

Some technologies do create far more jobs than they eliminate but I think AI is a different beast. I am seeing more negatives than positives when it comes to the job market.

(content loudness -22.7dB) please normalize the audio volume to closer to 0 db

Ai is inherently pointed at removing jobs. We need to think about the permanent loss of jobs and how to redustribute wealth and benefits accumumated from AI when working class humans become unemployable.

A robot that's as smart as a human can be taught to do any job that a human can do. How does that NOT lead to the massive displacement of humans from the labor force?

Of course he's gonna say AI isn't harmful he's one of the biggest supporters of it

"efficiency" and "productivity" are just the gentle terms management uses to justify layoffs

Feels like he's is desperately trying not to scare anyone....It is up to ours leaders to make sure we manage the transition? God help us all....

Not interested in those who are bullish on AI explaining how it won’t effect regular people. Their interests are not our interests.

Good coursera plug

new technology will always make us more productive, create jobs, destroy others - but I think the no. of jobs lost will outnumber the no. of jobs created.

Consider all the workers globally who offer their skills on freelancing platforms for things like copywriting, graphic design , research etc. AI can do many of these roles at a junior to mid level now. I’m sure I heard an interview with Andrew where he suggests that a universal global wage should’ve considered by governments. I wonder if his policy advisors and media people have asked him to role back on some of his statements

All these VC and C-suite level talks seem to miss the bigger picture of job "loss" which to me is about economic disenfranchisement rather than the relatively shallow question of "will jobs exist/will I still have mine?" There's a lot of talk of whether we'll still have things for people to do but not what will the power dynamic between employer and employee be in a post-ai enhanced workforce. Personally I have a very hard time seeing it as anything other than a cheapening of labor and an even faster rate of growing wealth inequality. Much of the trend I'm seeing with this LLM "accelerated" work seems to be about "lowering the barrier" and "democratizing" any sort of work. These are positive spins on it but if you dig deeper this can also be phrased as "removing professional moats". By removing the need to hire a professional artist to produce an ad you've lowered the cost of the ad, economically disenfranchised the artist, and rendered any artistic training an AI product can produce net-economically negative (and this could be applied anywhere from tech to law to medicine). Certainly this won't be all of the work, but it's by in large targeting professional white collared workers moats and replaces it with jobs that have a much lower one. Imagine a scenario where say 4x productivity in each knowledge domain reduces expert staffing need to 25% of previous levels and to replace the previous expert jobs the company hires with just as many people as "entry-level prompt engineers". Will the moat for such a position be high enough for the prompt engineers to have any bargaining power? Will the now flooded market of the disenfranchised professional be able to sustain the same type of life? Certainly they could leave their position go home and apply to several thousand newly minted 'ai-enhanced jobs' but why would a corporation bother to compensate anyone for popular work with a low barrier to entry? There's a lot of talk about this reskilling but not what a worker's moat would look like in this scenario. It's most executives' dream come true to be able to free their company from all highly compensated folks and rehire at minimum wage and pocket the difference. I don't think we'll see anything to this extreme but I think these tools by and large moves the needle in that direction. As we "democratize" all skilled work the economic employer/employee relations sustained by scarcity of skill will dissipate and leave behind a richer CEO and a poorer city.

Excellent conversation!

Thank you for sharing this insightful discussion!

Very pacifying, but I think in reality we will see new companies spring up in each sector which completely obliterate 'human heavy' conventional businesses. For example, a handful of film makers using AI will bring out blockbuster quality films which overnight send Disney out of business. Not entirely an exaggeration.