Wow, I'm amazed at Andrew's list of accomplishments. I mean any one of these would be a career capstone for most people
- Stanford Professor
- Founding lead of the Google Brain project
- Author of one of the most famous and loved MOOC's
- Head of AI for Baidu, built up the AI team in both China and the US.
- Founder of Coursera
And I see from wikipedia that he and I are roughly the same age :(
Engineers are often seen as a cost center for most businesses, which means you'll eventually hit a compensation ceiling If you want to elevate yourself into one of the engineers you hear about that is able to break through the compensation ceiling then the below is one of the best ways to do so.....
> My team birthed one new business unit per year each of the last two years
If you can directly tie yourself to a Pnl then you'll always have more options than someone who is considered a cost center.
I hope that what ever he does, he takes some time of first if he needs it. I'd hate for someone like this to get burnt out.
...and all that (excluding the professor part) since 2011! Boy, talk about working fast!
In case you can't tell, there's a great big sarcasm tag there. There is absolutely no way that the guy has contributed substantially to all of those things in less than six years.
Sometimes, people get a reputation for something, and then leverage that into a slingshot of career advancement. Good for him, but the people who are doing the actual work of developing this stuff can easily spend six years on a single problem, so it's helpful to maintain perspective.
Or, perhaps a little less cynically, that it takes a long time for your work to reach a critical mass, but past that point many factors come into play that allow you to ship stuff at a much increased pace: your work has a much higher visibility, more people are willing to collaborate with you, you have easier acccess to grants, etc
This guy is an executive. He isn't "shipping" anything, and it isn't cynical to say that.
Engineers tend to think about these things as if the world is a meritocracy based on your individual contributions. But for a lot of famous people, 80% of their contribution comes from attaching their name to the management page.
I agree with you that none of this would be possible without the years of work he did to build his reputation. But people shouldn't feel bad just because a well-known person has lots of flash on their resume. This stuff accumulates at an unnatural rate.
Become Stanford professor > Get prominent job > show you're cool > Get invited to do things!
But I doubt his work rate and energy was any less before, just as you say, it's visibility. And very cool.
Maybe we should all be on the lookout for at least one career move, that gains us a relative boost in visibility, even at the expense of other desiderata.
But assuming we each perfectly knew how to set our attainable goals - full awareness of human ability to succeed beyond imagined limitations, counted - what a CV like this tells me, is that the benefits to oneself, but also the world at large, really only begin to accrue when it seems to my experience (my bro becoming prof, unusually late in career, but for good reasons in pursuit his own interests, before) should absolutelynot be considered a ultimate, terminus, goal.
What I say, could be said as "rest not on your laurels", but it is i think more profound than that. At l;east I think so: because one's value to others can massively multiply, if you seek beyond what you may first think you are working towards.
And, if your goal is for instance to be appointed a Stanford Professor, does your perspective on career, pace of work, the work you pursue in fact along your way to that goal, all of that actually change so much, when you look beyond that, and see the utility you should instead aim to be, afterwards?
When my dad taught me Squash (he was a well known, widely published, coach) he first taught me to swing through the stroke, much as golfers know how also. Not only is there a huge uptick in stroke confidence, when you connect cleanly in this way, and power, too, but he was teaching me to calculate the angle of incidence, through my body, and that gave me instantly greater understanding of the shot. Of the future arising from my objective. Sure i was very young, this was a first lesson almost, so it is hardly great insight into Squash technique. But is it not, applied, not great insight into technique in playing Life?
He happens to be a very good teacher -- I've seen this first-hand, and anecdotally a great builder of teams and mentor.
So ... these are real skills that supersede individual-contributor skills.
In the same way that software offers leverage, really good people-finding -training and -inspiring skills allow you to turn yourself into a group intelligence.
The smartest thing Andrew has done is to hire a PR agent right after he did the ml-class MOOC, and never breathed a day without consulting him/her ever since.
Just as electricity transformed many industries roughly 100 years ago, AI will also now change nearly every major industry
This is an underrated point, and something I don't think most people outside of the high level Machine Learning/AI world take seriously.
It's also one of the biggest challenges for the industry going forward because of natural monopolies. I say that because if AI is electricity then data is the coal/oil that drives it.
The big technology players have a massive advantage in their ability to build and deploy tools that collect the data, and then bring it back to be turned into "electricity." If we aren't careful they will be the only groups who can make progress and show actual real world ML driven capabilities - making the barriers to entry even higher.
If you just look at the computer vision space, to do really good Machine Vision you need a LOT of novel image data and the primary platforms creating image content are largely owned by the top 5 players in the form of collection (smartphones) and warehousing (cloud servers).
I'm not sure if there are solutions here that make it possible for a lot of companies to do really well - everyone will just be bought up or out competed by the bigs once the big ones notice a threat on the horizon.
> AI is electricity then data is the coal/oil that drives it. The big technology players have a massive advantage in their ability to build and deploy tools that collect the data, and then bring it back to be turned into "electricity."
Data is important now, but when we have solved vision, speech, text and robotics to a decent degree, data won't matter as much. The great thing about AI is that we can cheaply copy already trained models or already labeled datasets. There aren't so many datasets needed to solve the most interesting and financially profitable few problems. Of course, there will always be fringe projects where more data is needed, but the main applications will be in the commons. You can copy an AI model if you can talk to it (use it to produce sample outputs). Any model could be copied in a dataset and transferred into another model. The great thing about machine learning is that it learns directly from data, so it's cheap to copy by tracing the inputs and outputs of other public AIs, just as current AIs are taught by tracing the inputs and outputs of people (supervision).
The distinction should be made between training data and [actionable] data.
In the former case we are taking data, labeling it, then using it to build our nets and models. You are correct to an extent that it's a usable model once trained and that data is less important.
However, equally if not more important is the data that is being put into the net to come out as a result/action. Arguably this data comes through the same pipe as training data - and the pipes are similarly limited. So its ALWAYS important because you can't take an action or classify or otherwise without it.
When you add in the reinforcement mechanism, or later unsupervised techniques then those data mechanisms blur between training and action data so the point is moot. It's not a one run process in the long run, it's iterative and always evolving based on the user.
The key point is that it's about data friction. If you already have uploaded all of your photos to google cloud, then any new tool or capability google comes up with using a CNN, will be immediately applicable without you having to do anything.
Was about to say the same thing. The incumbents are probably even at a disadvantage because as a young upstart you can always get in on the action by building a more efficient and simpler model and then querying the incumbents to bootstrap yourself.
you can always get in on the action by building a more efficient and simpler model
Except the big 5 have hoovered up a good portion of the ML talent from the universities and are themselves leading the pack in iterative improvement on ML capabilities.
If you are trying to bootstrap a ML company with the trickle of raw data that the big 5 puts out, you're not going to ever get to their size.
This is not about being able to create a 300k/yr company off the back of some table scraps. It's about major companies having too much influence over one market.
so did Kevin Kelly, in "The Inevitable". He stretches the electrical metaphor for about a chapter.
"The AI on the horizon looks more like Amazon Web Services — ... This common utility will serve you as much IQ as you want but no more than you need. You’ll simply plug into the grid and get AI as if it was electricity. It will enliven objects, much as electricity did more than a century past."
It's not just AI. Here's Jeff Bezos talking about the internet as electrification
Maybe open source crowdsourced data? e.g. If various people need thousands of examples of people's handwriting, someone makes an open-source repository and hopefully lots of people upload for the good of everyone.
So the big trick is to convince humanity that we all need to be the teachers of AI as an intentional practice. AGI will be our offspring so why shouldn't we all have a hand it teaching it and seeing the fruits of our labor?
I didn't find a single reference other than this as to what he is going to do post resigning.
>I will also explore new ways to support all of you in the global AI community, so that we can all work together to bring this AI-powered society to fruition.
It is true that AI is the new electricity which will change practically everything in our lives and it is good to see that alliances like OpenAI are forming to democratize the knowledge, this is because giant companies hold a monopoly, they are the only ones who have the sufficient data to do any meaningful research at all.
It's similar sentiment that Jeremy Howard presented. It sounds that they both see AI such a big societal change that working on a single startup / company isn't enough for them.
Ng's wife's company, Drive.ai, has made more progress in less time and with fewer resources than anybody in the autonomous driving space. So it's probably a billion dollar company. I wonder if that has anything to do with it. There's a lot of money in autonomous driving startups that can deliver results.
Googling around, I haven't seen any substantial evidence of any of your claims. It's hard to separate the hype from the reality when hearing anything about the self-driving car space.
Baidu is now one of the few companies with world-class expertise
in every major AI area: speech, NLP, computer vision, machine learning,
knowledge graph.
Just idly, I find it interesting that practical applications of this technology seem to often funnel down to just this subset.
There's a lot of room to apply machine learning to solve actual problems that many people have, but often its unclear that doing so would end up with results that are significantly better than traditional approaches; or how to achieve those results, tangibly.
I'm sure we haven't heard the last of Andrew Ng; there are a lot of people who want those sorts of skills.
it was only after your comment that I'm noticing: talking, listening, understanding, seeing and remembering. I wonder if we're just too focused on doing things that people do.
I haven't seen any comment or quotes yet from Baidu (please correct me if I'm wring!). Amicable high profile departures like this are almost always coordinated between the person leaving and the company, with joint statements. Not saying it's not amicable, but I do find that, and the fact he's not sharing his next move, a little odd.
I also saw a FB thread that seemed to suggest this may have been a surprise to some of his colleagues. Take that with a grain of salt though - I don't know the people on the thread and they were vague. I just got the sense it was news to them.
“Andrew joined Baidu because of our shared pursuit for the future of AI,” stated Baidu on its official Weibo account, China’s answer to Twitter. “We still have this goal, which is to push forward the development of AI and make life in the future more beautiful.”
“Despite our regrets, we send our thanks and blessings! We wish Andrew even greater success in the future, and hope all goes well!”
High-profile departures are hard to read because leadership teams are usually (and should be) paranoid about the effect on morale. Also, as a general rule, clean breaks with little to no warning are easier to digest for the rest of the company rather than long, drawn out affairs where gossip can be given life through dint of "overhearing" the subject (almost incorrectly or out of context) in the hallway.
If you are planning to take it, I would suggest a different approach which I have started.
1. Do Practical Deep Learning For Coders:-
http://course.fast.ai/ To take a plunge directly into deep learning AI, this course has rave reviews. This course will allow you to do practical industry level stuff first, then learn theory behind it, rather than other way.
Course description:
>This 7-week course is designed for anyone with at least a year of coding experience, and some memory of high-school math. You will start with step one—learning how to get a GPU server online suitable for deep learning—and go all the way through to creating state of the art, highly practical, models for computer vision, natural language processing, and recommendation systems. There are around 20 hours of lessons, and you should plan to spend around 10 hours a week for 7 weeks to complete the material. The course is based on lessons recorded during the first certificate course at The Data Institute at USF. Part 2 will be taught at the Data Institute from Feb 27, 2017, and will be available online around May 2017.
2. Read http://www.deeplearningbook.org/ to gain the relevant math behind it. If you don't know some of the math like calculus or linear algebra presented in the book, learn as you read it from sources like Khan academy.
Now we are up to date on practical side of things, especially deep learning part. We can move on to gain a more generalized and rigorous outlook on various machine learning techniques.
3. Do https://see.stanford.edu/Course/CS229/ - CS229 By Andrew NG , its more rigorous, and complete compared to coursera course. And coursera course is not
I think within a year (max) just this coursework plan would give a strong foundations on practical, theoretical side of things in AI.
I get distracted trying to learn so many stuff at once, (clojure , sicp, haskell, advanced algo) etc etc. So I made this lesson plan to follow as I am interested in AI the most.
I recommend this AI/ML learning method too. Practical first. Trust me, it'll keep you from getting bored/sleepy.
And for the practicals, I'll also suggest going over some of the hands-on examples at blog.algorithmia.com, especially if you have some Web dev experience.
The fundamental techniques of machine learning will always be there. For that reason, Ng's ML course will always be the prime entry point for people new to ML. As someone who has taken both the Ng course and has roughly watched the videos of the fast ai course and done some work, I would recommend first doing the Ng course. Things will be much more calm if you then approach the fast ai DL course. The fast ai course is amazing and hats off to Jeremy and Rachel for the great work. But without any idea of machine learning, I would say you will need to put at least 20 hours of work each week to work through it, and that is if you are also a really proficient programmer.
Yes. It may not cover some of the most recent advances (e.g. deep learning) but it lays the foundations required to understand them. Besides, it is really really accessible. And fun. I enjoyed taking the class.
Frankly, I wonder why Dr Ng stayed at Baidu so long.
Without question, Baidu's dominance over search plays two very painful roles: 1) they're the cutting edge of China's online civilian spy apparatus, and 2) they're the gatekeeper for news and dissemination of gov't propaganda. If you search the web or visit links sponsored by Baidu, either you will find what the government wants you to find or you will be identified as an outlier, and at best, labeled as a potential threat and tracked. It's impossible to imagine that Baidu does not place the names of web denizens matching certain profiles on threat databases thus altering the trajectory of their lives accordingly (and invisibly).
Personally, I can't imagine anyone of conscience working long for an employer so committed to diminish the freedoms of thought and speech. I applaud Dr Ng for his departure.
I wonder if Baidu hiring Qi Lu as COO in January has anything to do with it?
He has been called "a leading authority in the field of artificial intelligence".
Andrew left just after Qi Lu was hired as COO in Baidu. Hopefully, Qi Lu can bring some shining product out of those great techniques, encouraging the AI team left behind.
- Stanford Professor
- Founding lead of the Google Brain project
- Author of one of the most famous and loved MOOC's
- Head of AI for Baidu, built up the AI team in both China and the US.
- Founder of Coursera
And I see from wikipedia that he and I are roughly the same age :(
Engineers are often seen as a cost center for most businesses, which means you'll eventually hit a compensation ceiling If you want to elevate yourself into one of the engineers you hear about that is able to break through the compensation ceiling then the below is one of the best ways to do so.....
> My team birthed one new business unit per year each of the last two years
If you can directly tie yourself to a Pnl then you'll always have more options than someone who is considered a cost center.
I hope that what ever he does, he takes some time of first if he needs it. I'd hate for someone like this to get burnt out.