Misleading article and headline; a few celebrity researchers in high-up positions make this much. You can probably find some chemists who make $1M too if they have made it to executive positions.
Most AI researchers in industry make a lot of money, $200k+, but that is not so outrageous in the context of big tech companies.
And in fact the vast majority of AI researchers are making $20k-$30k a year, because they are graduate students.
Also, it's not unusual for star professionals in other fields to make that much at nonprofits. E.g. neurosurgeons can often make that much at nonprofit hospitals. And if one gets into executive comp at a large nonprofit, a million a year is no big deal.
But there is a large variance in the nonprofit world, and it's more common for a PhD in CS to earn something like $60K at research nonprofits. Other nonprofit employees - social workers, for instance - have it even worse, earning not much more than minimum wage with Masters degrees. Sadly, many nonprofits capitalize on the idealism of applicants to compensate for poor comp and grueling working conditions.
So the word 'nonprofit' really doesn't convey much information; it all depends on specifics.
Exactly. The problem with this field is that if you want to see what you'd have to do to be a star researcher in AI. You'd have to be top-10 amongst 20.000 or so machine learning PhDs. This is not a reasonable aspiration. Plus I would argue that although Ilya Sutskever did a ridiculous amount of effort and then got lucky, 2 two others were at best interested academics on the sidelines for most of their careers, then got lucky with a grad student. There's at least 1000 who did similar effort to Ilya (let's say half as much, which is still triple what any reasonable person would find reasonable), way more than the 2 other named examples, yet have no hope at all of ever achieving his level of recognition, including several of the students who really did most of the work the 2 others got recognition for.
You also have to do a PhD, and make sure you get a good adviser/good department, and get lucky enough for nothing bad to happen to your adviser and department, and be very careful with how you spend your time (if you spend a lot of time on something that doesn't end up being fruitful, that's opportunity cost that other top researchers might have turned into a breakout publication and now you are behind). And you make like $30k/year while in grad school and might experience some pretty bad mental health due to stress.
Of course, the average/above-average case isn't terrible. I think non-famous AI researchers are still getting hired into $400-500k/year at FAANG at "senior" level, it's just that you would be making roughly the same amount of money if you just went directly into software engineering at a FAANG out of undergrad and did above-average at promotions.
To me, that makes it not really make sense to go into AI research (especially now, when it's a hot topic and you have a 3-5 year wait to get your PhD) unless it's something you are genuinely interested in. For purely career reasons, it doesn't seem much better than just going right into software engineering.
The strangest thing is that knowing a lot of PhDs and PhD students (my wife is currently one), very few are motivated by financial rewards. Making money isn't always seen as a yardstick of success by their peers.
That's why I never recommend people to go into PhD programs for career advancement. I have a colleague who works 50% as an NLP PhD student and 50% in my data science team. To really get that PhD done you have to live and breathe for the subject, and only the subject, for however long it takes. He's been on this track for 2 years now and is thinking of quitting, his own words; "I often feel like if I really wanted PhD, I wouldn't be working so hard building models for industry all the time".
Star jugglers do not make a killing, unfortunately. A man with a not unreasonable claim to "best juggler in the world" gave it up and became a concrete contractor. Salaries at Cirque start at $50k.
In fairness, I suspect it depends on what you're juggling. I could envision someone juggling management of an electric car company and a space exploration technologies company making bank.
I think it's far more likely this person would realize juggling so much is harder than he thought, and it all comes crashing down in the end (including his net worth).
Actually you are incorrect. Moschen won the grant in 1990. Adjusting for inflation, that is just over a million dollars in 2019.
Even despite this, your comment strongly indicates taking a rather pedantic position. If you are suggesting that the mere amount of the grant somehow counters the spirit of the discussion, I’d say that’s extremely incorrect and a strict adherence to some threshold like “millions” would be totally missing the point.
It's a one time grant that seems to be given sort of randomly, by which I mean I don't believe people apply for it. You just kind of get surprised one day to learn that these folks think you are spiffy.
Good for him, but I really don't see that as a rebuttal to "wages are so low for this occupation that people really good at this choose to do other things as a job." It's a bit like saying "Well, just do what you love anyway! You could totes win the lottery as a way to keep a roof over your head!"
Well, kudos to the lottery winners! But most folks would like a paycheck they can kind of count on that covers their needs reasonably adequately without hoping for a windfall out of the blue as the solution to their problems.
The whole thread is specifically about lottery winners. This is an example of such a thing happening due to juggling skill. It refutes the idea there are no such examples that result from juggling skill.
Only in the sense that Paul Graham talks about successful startups "doing everything right and winning the lottery." There is a certain amount of luck in anything, but being a statistical outlier for income isn't the same as "winning the lottery," which is basically random chance.
And the reason it matters is because earned income encourages people to develop useful skills that the world desires (so as to presumably make the world a better place). Telling people they can actually make bank at X if they keep applying themselves is telling people "You don't have to be a pie-in-the-sky idealist to make the world a better place. You can do things the world actually values, be really good at them and get paid excellent money, so: win/win!"
The Nobel Prize and even MacArthur grant are intended to reward people after the fact for making the world a better place. Part of the point is to empower them to keep doing what they believe in and not give up and go do something else for pay that would be less beneficial to the world at large.
It may help encourage some folks to keep at the thing they believe in despite the low pay, but it's not a plan you can take to your accountant for how you will make your retirement work: "I'll just be amazingly good and then get a random windfall. It's Fine!"
I'm a freelance writer. Like a lot of writers, I work for peanuts.
I don't expect to be the next JK Rowling, but I can learn from her (she rewrote the first chapter of the first book like twenty times) and I've applied myself and gotten better. My pay has been gradually going up.
Because of how publishing works, JK Rowling's financial success is somewhere between salary of a million annually and random grant out of the blue. Plenty of people think writers must be nuts because most writing pays so poorly. (I have been repeatedly told to "get a real job" if I don't like being poor.)
But there's still a difference between that and your "win the lottery" framing for this discussion. It's really not the same thing.
Nonprofits are more like businesses than people realize: https://seliger.com/2012/09/02/why-nonprofits-are-more-like-... and they pay market rates. Maybe some people will accept someone less for the sake of working at a nonprofit but very few people will accept substantial paycuts over the long term
I've generally found nonprofits pay well below market in exchange for doing social good. Universities are the most obvious example, Chan Zuckerberg Initiative being another
Off-topic: Reminds me of "Half a mil in twenties like a billion where I'm from."
On-topic: that's misleading; once you pay fixed costs, you have a lot more latitude for discretionary income. Not everything is the same ratio more expensive in the Bay Area.
What happens when you have a family? A MacBook Pro costs the same but does childcare? How about a home with a reasonable commute, good schools, and is large enough for a single family?
If you have a family it’s not possible. Home prices average close to $1m. You’d have to save an entire years salary PRE tax to afford the down payment. That will take years and years to save.
If you’re raising a family you’ll likely need two $200k incomes to afford a house, and even then, it’ll take at least 2 years to save for the down payment.
The Bay Area is for college kids and millionaires. That’s it.
It's certainly possible. You just don't own a home, you rent. It's much like NYC in that regard - you don't own a detached home in NYC regardless of how rich you are, you rent a swanky loft or penthouse. At most you might buy a condo, but you can still find plenty of hedge fund managers or Google engineers that rent their permanent residence (and sometimes have over $1M in liquid assets).
In my wife's mom's group (which had a number of dual-tech-income couples working for places like Apple, Google, Palantir, Facebook, etc.), not a single parent owned their own home. I live in a small 13-unit townhome complex, and about 8 of the units have families with young kids.
Very interesting how affluent people still rent. I noticed the same in LA in a high end building. What is the reason? Having more money to invest in different asset classes instead of the biggest chunk being in real estate?
On an economic level, it makes sense to own other assets and rent your primary residence if the discounted cash flow from the investments available for the price of your home exceeds your rent. Typical returns on real-estate are about 3-4% (you pay the bank about that much to borrow the cash to buy a place, and if you own the place outright, you can make about that much in rent), but there's also the mortgage-interest tax-deduction, so it's equivalent to maybe 5.5% returns. The stock market typically gives about 7% returns, but with higher risk: if your risk-tolerance is high it makes sense to not spend money on a house but instead to invest it in stocks and use the cash spun off by that to pay your rent. (If your risk tolerance is very high, you can take out that mortgage, borrow the money for the house, and invest your savings in the stock market anyway.)
On a pragmatic level, high-income people who rent often have jobs that require a large level of geographic flexibility to get their maximum salary. Think of your professional skillset & network as an asset that various companies can "rent" for different rates, depending on how much value it adds to them. Different companies will pay wildly different amounts for that; it doesn't make sense to buy if you get a 50% raise by moving to a different city (or just a different area of your metro region) but costs of selling your house eat up several months salaries.
20% down is a rule of thumb, and a requirement to avoid mortgage insurance, but it is not a requirement for lending.
My credit union will do loans of up to 95% loan to value for $1.5M to first time homeowners or 90% loan to value up to $1.5M if you don't meet the first time qualifications. [1] I'm sure other volume lenders in the bay area have similar lending standards and programs.
It's a financial mistake to not put 20% down. Not only are you hit with $500+/month in PMI, you have almost negative equity into the house when you buy it. Meaning if you have to sell on short notice you're either completely underwater having to pay back cap gains, or you're just breaking even. That's not an ideal situation and if you don't understand the mechanics going in, you'll find yourself in a world of hurt in an economic downturn.
Sure, PMI costs money. If it gets you off of the housing inflation train years earlier than waiting, it might be worth it. If housing prices go down after you finance with PMI, it might not have been worthwhile, but it's hard to know what's going to happen.
It's worth at least looking down the path, and considering the details of your current situation, and not just not considering purchasing at all because you only have 10%, 15%, 18%, or whatever that isn't 20%.
I don't understand how you can be 'underwater' but yet have capital gains. Certainly, selling a house is expensive, and if you put down only 5%, the loan balance could be more than your net sales price.
If you truly don't have the money for the down payment, the math of mortgage+insurance+property taxes+pmi+deductions might still work out to be in your favor over renting
It's pretty unlikely in the Bay Area. There are fairly large distortions in the local market (because of Prop 13, zoning, wealth inequality, and small home supply in general) that usually make renting significantly more affordable than buying in general.
See eg. this $1.9M home (estimated payment $9000/month, not including PMI) vs. this $4900/month equivalent:
(Also note that the $1.9M home is currently assessed at $170K, so the current owners are paying < 10% the roughly $15K/year in taxes that the buyer would be.)
I will say though, might not make the most sense to pick a house in Mountain View as an example. For a couple just starting out in the South Bay that don't have two high earners and want to own a house, Fremont/Union City/Milpitas/San Jose are more realistic.
Also because of prop 13, it can make sense to lose money on buying vs. renting in the short term if you anticipate staying somewhere for a long time, because of the potentially huge property tax savings. And for basically the same reason it can make sense to lock in a mortgage payment that doesn't look like a great deal on paper in the short term.
Yes but then you need to commute an hour instead of 15 mins. Meanwhile that $5k rental oftentimes is in a brand new luxury building with full amenities and is very nice, and also has the 15 min commute
Wow the Bay area is crazier and crazier, 2M for that dump... yeah 4 bedroom in Mountain View, but it's a 1964 and looks like it's going to fall down at the first gust of wind. Everything inside the house is to be remade too and in 2012 the property was worth 700K. That area is literally unliveable now.
It's a different lifestyle. You're just not in your house a whole lot in the Bay Area. My weekend involved a trip to a science museum & zoo; picnic lunch by the Bay; friend's kid's 2nd birthday party at a park; gymnastics class for my toddler; mom's group at the park; meeting up with mother-in-law; and a carnival. During the weekdays you're usually behind a computer screen most of the time, or you go for a hike, bike home, or take part in an organized activity (I took Krav Maga lessons after work from the former head of the IDF's counterterrorism training for a couple years). It doesn't make a whole lot of sense to spend a lot of money for a house or a lot of time to fix it up nicely when it's basically just a place to sleep. (Sometimes I wonder if folks who buy these $2M homes are just folks who have bought into the conventional idea of the American dream and absolutely must own a home.)
And then on the flip side, most of those folks own financial assets elsewhere - either they get stock options in their employer, or they sell those stock options and diversify among other stocks, or they buy 5 homes in the Midwest and rent them out as an absentee landlord.
The system is driven by the massive amount of money flowing into the region from all over the globe as software eats the world - this fills local municipalities with tax dollars that they can use to provide public goods, and it inflates the stock compensation that companies pay their employees, which gives them money to afford the exorbitant rents while still saving up a million or so. It's ultimately unsustainable - bad things are going to happen to Silicon Valley once software is not the engine of growth throughout the world - but that's probably a few decades off, and in the meantime people who partake in the system can afford to buy up whole city blocks in Detroit.
That does sound like a wonderful weekend. We are relatively new to the Bay and still trying to find out way and people. How did you get connected to everything / find things to do and people to hang out with? We have a few month old baby
I've been here a decade and my wife grew up here - I end up piggybacking on a lot of her friends and social activities, and I picked up a bunch more from coworkers and such. But for a whirlwind HN guide to the Bay Area:
Parks: check Google Maps for anything that's a splotch of green and just go to it. Local parks are free, county & state parks often charge a $5 admission (frequently not enforced). Most of them are pretty good. BAHiker.com has a good guide to state & county parks, Google Maps has pictures and reviews on local parks.
Events: Google [$city events] or check out EventBrite, Johnny Funcheap, or SFGate & the Mercury News's events sections. Also look for flyers when you visit your local downtown area, or check out the local library's website (they often sponsor a lot of fun stuff like storytime for kids or Star Wars Day). Many of the peninsula towns sponsor things like classes, swim lessons, outdoor movie nights.
Socializing: the Bay Area has a thriving Meetup scene, but tbh I've never made a friend that stuck through public Meetup groups. Had much better luck with staying friends with former coworkers, and with paid activities (gyms, courses, music lessons, kid stuff, etc.) Google is your friend here; Google [$activity near me] and you're almost certainly going to find something. Because the Bay Area is both wealthy and densely populated, it's a magnet for really good instructors: aside from the aforementioned Krav Maga gym (KravZone in Sunnyvale), there are 3 world-class gymnastics gyms (West Valley, San Mateo, Airborne); a mandolin orchestra (SF Mandolin); some hardcore cyclists (have seen them riding over the mountains in a group occasionally); and many other activities.
For mom's groups: there're some that are just organized over Facebook (oftentimes by invitation - if you get plugged into the parent networks you often find out all sorts groups), some that are public, and some that are organized through local non-profits (eg. Blossom Birth or El Camino Hospital). I'd probably start with a for-pay class to meet people (folks are often more committed in the for-pay classes, and it draws from a social class that is more tightly networked) and then ask them if they know of other resources.
Museums & amusement parks: the big science ones are the Exploratorium and Cal Academy in SF, the Tech in San Jose. Smaller science ones include Chabot Space & Science Center in Oakland, Monterey Bay Aquarium, SF Aquarium, and Seymour Marine Discovery Center in Santa Cruz. Kids ones are CuriOdyssey in San Mateo, Happy Hollow in San Jose, Junior Museum in Palo Alto, the Discovery Museum in San Jose, and Gilroy Gardens in Gilroy. Amusement parks: Great America in Santa Clara, Raging Waters in San Jose, Children's Fairyland in Oakland, Santa Cruz Beach Boardwalk, and Six Flags in Vallejo. History & general interest: Egyptian Museum in San Jose, Hiller Aviation Museum in San Carlos, Computer History Museum in Mountain View, Maritime Museum in SF, the Hornet in Alameda, and probably a few others. There are also 3 railroads that my kid loves (Billy Jones @ Vasona Park, which also has a carousel; Roaring Camp @ Henry Cowell Park; and the seasonal Train of Lights between Fremont and Sunol), and most of the local parks have splash pads that he really enjoys.
I have no idea how that is possible. I made $100k in the bay area and it was just my wife and I.
Here's how it worked out:
$100k salary = $74k after tax, so that's about $6k per month take home. $3000 per month for rent leaves $3000 for everything else.
My wife and I have $1000 per month on our student loans. So we're down to $2000 per month. Health insurance and medication was $600 a month for the two of us, down to $1400. Groceries worked out to about $600 a month as well, down to $800. We got around via muni, so that takes off $200 per month for passes, down to $600. Electricity, water, internet, and cell came to about $200 per month. That left $400 for entertainment, incidentals, and savings.
And that was for 2 of us. I can't imagine 5, but kudos for doing it.
I think your issue is SF. In most of the Bay Area (including desirable parts of the South Bay like Mountain View & Sunnyvale) you can get 1BRs for $2000/month or 2BRs for about $2600. Less in the East Bay. My family of 3 spends about $300/month on groceries - knowing the local immigrant grocers can shave off a lot on fresh produce, our vegetable bill is usually about $13/week for about 20 lbs. of fresh vegetables. We get meats at Costco for pretty cheap.
I would also highly recommend paying off the student loans before having kids. My parents borrowed close to $100K to send my sister and I to college, and 4 years after graduation it was paid off completely, after saving 80% of my income and contributing it back to them. Because of compound interest, a few extra contributions towards the principal shorten the term of the loan dramatically, because it means that the loan balance shrinks, the amount of interest owed each month shrinks, and so a greater percentage of each payment is applied to the principal, which speeds up repayment even more.
Haven't gone too much into financials (given that it was my dad who was making the salary).
But I know it was right around 100k and rent is exactly 3k(live in Cupertino) a month. Not sure what other strings had to be pulled, but I'm glad I lived a decent life in SV.
Kind of ridiculous part is we don't even qualify for financial aid for some colleges/application fees.
Assuming 4k/month on rent, your leftover pay after taxes and rent is more than most senior engineers gross in the rest of the US/Western Europe. 200k is a lot.
> A third big name in the field, the roboticist Pieter Abbeel, made $425,000, though he did not join until June 2016, after taking a leave from his job as a professor at the University of California, Berkeley. Those figures all include signing bonuses.
425k after signing bonus is probably closer to 300-350k base + annual bonus, which feels low for someone in such a distinguished position (close to a typical specialized physician or big tech software engineer with 5-10 years of experience). After finishing my masters degree in machine learning in the mid-2000s, I immediately got job offers higher than my much more talented professors. To me, the real story is that academia has set the compensation bar low for people adding so much value.
Edit: to be clear I know these salaries are high and people in my field are lucky to be making so much. But I also don’t think it’s accurate that people in the field are overpaid, which titles like OP may make it seem.
Yeah its really not that high. I was reading the Ian Goodfellow, creator of GANs, was only making 800k at google. Which is honestly, really low for someone who has done ground breaking research in neural networks.
Yeah that’s pretty insulting, honestly. Black or Sholes were making >10m as thought leaders in banks. It’s surprising to me that finance is so passe as a career despite the fact, as Matt Levine said, “they are run as employee run co-ops.” In otherwords, big big money. Entry level analysts at my old firm were making over 250k total comp.
Scholes & Merton also joined a hedge fund that traded based on their models and cased a financial crash after the Russian bond default in 1998 which their models didn't take into account. Great example of a black swan event.
Context is always important. At the end of the day it is about what value you bring to the table. If the work fits your experience and capabilities well, charging anything between $150 and $250+ per hour on a long term contract isn't out of the question, I've done it many time, with multi-year engagements. The best deal I did was some 25 years ago was substantially above that range.
Of course, part of it is being a good negotiator --and I am not the best by far.
Again, it is all about what one side needs vs. what the other side can deliver. You then add a context that includes such things as urgency, uniqueness of the problem to be solved, supply of practitioners with the required expertise as well as other dynamics (does the company need to meet a milestone ASAP?) and you have the makings of doing very well if you are in the right place, at the right time and with adequate negotiating skill.
Of course, if close such deals you have to be able to deliver the goods or you better start thinking about an alternative career.
Contract numbers shouldn't be directly compared to employee compensation numbers. A contractor has additional expenses and taxes and can have periods between contracts with nothing coming in, so needs to be making more money to have equivalent net income after taxes and expenses.
No difference in my experience. It's all part of the negotiation. I used "contract" as a general term rather than to mean just contract work.
Of course, I am talking about outliers here, that should be understood. In most of these cases it is, as I said, a right-place, right-time, right-skills circumstance, an alignment of the planets, if you will.
I've seen even crazier stuff. Back during the most heated phase of the internet bubble savvy guys were making outrageous deals. I remember one guy who wanted to charge $500 per hour if he worked from home and $800 if he had to come to the office...to complete a Visual Basic project. No, he was not hired. Wrong place/time/skills.
I think people are supposed to report their salary with the stock price at grant, not after appreciation (at least on levels.fyi). But yea it's a lot more because google stock is half their comp and it keeps doubling every 3-4 years.
I’m from Levels.fyi. We actually request users to report compensation with appreciation. This is because the appreciation can actually be used to negotiate compensation elsewhere and provides a more accurate picture.
7-figure salaries have been a thing for quite a while at Facebook, Google and definitely at hedge funds.
Right now, you can make a half million a year or more if you're willing to optimize the low-level code in any of the major AI frameworks.
If you publish a major conference paper, 7 figures is pretty easy to attain. You don't have to be in the 0.001% just the top 5%. That's hard work but it's doable.
The only thing that's remotely fungible in AI is python-based data scientists who can do little more than port new data to existing models open sourced on GitHub. And even then they do pretty well because of the huge demand.
I don't see that changing anytime soon. I believe expecting automated systems to replace data scientists is the same level of naivete as believing we'd reach L5 autonomy in a few years in 2014.
Anthony Lewandowski was making 120 million dollars a year when he left Waymo. If you're an AI expert in the valley and you're not making at least $500k, you are one job hop away from doing so if you play your cards remotely right.
The first thing you need to do is convince yourself you're worth that kind of money. Then when they ask you how much money you want during a job interview start talking about your peers in the industry making that kind of money.
Finally make sure you're talking to multiple potential employers so you can play them against each other. Elon Musk gave that money to Ilya Sutskever because Google was probably paying him at least $1.5 million. Even if not, his share from the $60M DNNResearch acquhire (that's what got the whole ball rolling here) was probably worth far more than that.
Sadly, you're really only worth what you can convince another person to believe you're worth. But then the USA is suffering through someone convincing an electoral majority of Americans he was the best and only choice to lead the country and that's just working out splendidly right?
I did not know about these conferences. I googled the first one and found a headline about how tickets sold out in minutes. Is this why? Because everyone attending hopes to get a 6 figures job?
I'm being facetious, this is not at all true. At my company, we typically look for multiple major publications and you certainly don't get a million dollar job out of it.
Then the people who come to your company could probably do better elsewhere if money is all they care about. You work at Cruise am I right? I know someone personally who worked at Cruise who was making 7 figures. That sort of compensation is just not that unusual anymore.
But I've learned that money isn't everything the hard way. However, if you're whining about being underpaid, there's something you can do about it. There are plenty of solid reasons not to however especially if you enjoy your current gig and they're paying you sufficiently that you have a promising future. Money truly isn't everything.
I don’t doubt that either people could do better elsewhere or that someone makes a million dollars, but it’s disingenuous to claim that “If you publish a major conference paper, 7 figures is pretty easy to attain” when this is not true by most measures. It might happen to some people but “pretty easy” implies most fresh PhDs could do this.
I think you are embodying the notion "You miss all the shots you don't take." You're right that most fresh PhDs cannot do this (I did say top 5%). But I think we disagree as to why. You seem to think these positions do not exist and I know otherwise because I've been on both sides of the very same hiring process for which they do exist. The reason most new PhDs won't get them is because they lack the career experience to navigate the interview process to such a position. I'd love to see a world where someone teaches them those skills while they're still in academia. I sure could have used that.
I also know because the day I left academia myself long ago was a major payday for me. Finally, I know because people like Andrei Karpathy pulled it off because of his excellent communication skills, intelligence and the right background. And people like Mu Li pulled it off by associating themselves with an academic who who helped found AWS AI and in his case he landed a principal engineering position straight out of school. That's the sort of position one usually only gets after one to two decades of job experience.
Interesting that you mention about Andrej Karpathy. I've long wondered about his meteoric rise. No doubt he's intelligent, but there's definitely more to the story. What do you think it is? Is it majorly because of the right timing and being under the right profs? His blog? His class? I'm currently a grad student. So if I want to replicate even an iota of it, how do I do it?
I think the turning point for him was his excellent presentation in the first youtube broadcast of cs231. If I remember correctly he was getting seven-figure offers before he even graduated from Stanford. I do not know him personally and I doubt I will ever meet him but we did compete for the same position at Tesla and I lost to a very worthy opponent. my only regret is that the final interview would have been with Elon Musk himself and I would have loved to meet him.
Fat chance, there’s hundreds of papers and tens of thousands of attendees at NeurIPS every year. Less than 5% of these people will make 7 figures yearly.
Saying he earned $120 MM a year when really that was a payout after multiple years of hitting agreed upon milestones - that's sort of like saying everybody makes millions a year at a company that just IPOed. He obviously made a lot of money, but you're exaggerating by almost an order of magnitude.
The $680 MM that was also contingent on milestones he never hit and thus was never paid? Not to mention that was distributed across other people as well, although he was supposed to get the vast majority.
Also he's likely going to end up in jail because of what happened with that Uber fiasco.
That he turns out to be a scumbag seems orthogonal to his compensation. He would have gotten that money if he had delivered, no?
That said, my exiting salary at some companies was dramatically higher than my salary when I joined because the stock had appreciated and so had my compensation. And if someone wants to hire me away that's the compensation they have to deal with, not my initial comp.
Except in this case the main contingency was stealing Waymo's IP, and it's fairly well-established at this point that both parties were aware that the transfer of knowledge would be illegal. So he couldn't have delivered it without being a scumbag, since only a scumbag would try to steal it.
Anyway this is all a digression away from the fact that you are wildly misrepresenting salaries. You're also ignoring the two companies he sold to Google as part of his compensation. If you're interested you should google a bit about 510 Systems because the history is actually interesting, especially with the added knowledge of the founder becoming a total scumbag.
I think this article just reeks of jealousy. What is wrong with paying people doing top research a top salary? Their skills are extremely rare. And I bet executives get way more anyway.
What's amazing to me is the same people who write articles like this don't blink at the salaries in Hollywood. A top-tier performer can get over a hundred million dollars per movie. A bottom-tier character actor or below the line talent barely gets by.
The difference in compensation here is nowhere near as bad as what happens in Hollywood.
Without commenting on quality of the journalism here nor making any judgements about the normative economics [1] of the situation described:
Hollywood actors salaries (or those of sports stars, or CEOs, or authors of best sellers or pop stars) are a known thing so you don't write a 'news' story about the magnitude they have obtained. AI researchers making large sums is not widely known so is 'news' and of interest to readers.
How replicable is this thing? Looking from a monetary perspective, is it worth getting a Ph.D. in AI? Assuming you like both academic research and working in the industry, if you were to make a choice whether to do a Ph.D. or not, how would you make it?
Remember that Hinton and other deep learning pioneers labored in obscurity for over a decade, probably closer to two.
Techniques like deep learning were out of fashion for a long time. Google was making tons of money on non-AI techniques; I remember Sergey Brin saying he was surprised by the deep learning breakthroughs in 2012 because the conventional wisdom was that "AI doesn't work".
And people who did Ph.D's under Hinton and others 10+ years ago also were not following trends.
You have to take risk to get reward. The way to do that now would also be to labor on an obscure subfield of AI that isn't certain to even work, let alone become commercially viable. (And there are plenty of people who did that on fields that looked more promising than deep learning 10+ years ago.)
There's definitely a need for new AI techniques, see this other current story:
Not to also mention that much of this would still be in the stone age if it were not for MUCH faster and accessible hardware (GPU/TPU/QPU/XPU), what I think makes up for the high salaries is that these people are quite aware of how to allocate for hardware on the ML/RL/AI training side and how to allocate time for the people responsible for tagging and optimizing models. They know how to reach the objective(s) in training/testing/optimizing already.
No, it's not. Im an AI researcher at a big tech company. It would have been far wiser financially to start as a software engineer after my Bachelor's.
Also, having a few NEURIPS or ICML papers is not a hiring guarantee any more. It's decidedly not 2015-17 any longer. In particular, I feel bad for the people who started their PhDs with $$$ in their eyes around that time.
Don't get me wrong, they will all be employable, but the field moves so fast, I have my doubts there will be cushy FAANG jobs for everyone capable of playing around with network architectures in a few years. It's a terrible idea to start one based on the hype few years ago.
Isn't this always the case though? A PhD is a long term investment and the field that you go into is probably not the one that will be hot when you finish your PhD. So driving your decisions on $$$ seems foolish at best.
Yep. If you're empty enough to follow a PhD solely for the money, there are far safer financial bets. Go into medicine or something. Not that I'd want to experience that bedside manner...
It is not fair to call someone "empty" because their motivation stems from something you don't share. It does not make their work and contribution any less difficult or necessary.
Though harsh and (of course) generalized, I think the word "empty" is a useful description here. Think of these meanings of empty:
- "containing nothing; not filled or occupied"
- "lacking meaning or sincerity"
- "having no value or purpose"
In the context of academia, there is an expectation of a certain fundamental motivation; namely, to pursue knowledge in a particular field and share it back via writing and/or teaching.
If one lacks that primary motivation, they certainly "lack" it and are "not filled" with the academic ethos. So "empty" is a fair term.
Now, I understand the realities as well -- seeking publications, recognition, tenure, and funding are also political activities. But this does not contradict the underlying community norm that I mention above. In fact, it supports it -- it explains why so many people endure a tough, grueling, political process despite it not being their wheelhouse.
Indeed, I should have expanded a bit, it was a bit disingenuous from me. But I don't think that academia's ethos is the primary motivation. It wasn't for me. The primary motivation for my phd was to get the doctor title and money afterward. It was a transaction between academia and me where they get a qualified engineer for a low price, and I get educated about research. No one ever talked to me about academia ethos. The closest thing about educating and sharing others my supervisor told me was "We might have a job opening for teaching and research after your phd if you are interested to further expand in Academia, but don't count on it. Such openings are rare".
So basically your conclusion doesn't go well with my experience. I only saw a contract with low salary and a lot of work on my end. It was worth it for me (I think?), so I accepted.
Thanks for telling us your story -- you make good points.
To what degree did you enjoy the process of learning, collaborating, teaching, writing, experimenting, and so on?
I'd wager you did enjoy some or many of these... otherwise, it might have been a long slog. :/
Maybe the following story can convey part of my message. You might have seen movies about a wily protagonist villain (or a flawed, tenuous partnership between several) who meticulously plan to steal some priceless artifact from some nearly-impregnable facility. What drives such people? I don't think it is purely money -- there would be alternatives that would, rationally speaking, generate more income, on a risk-adjusted basis. In the case of the ninja-suit wearing infiltrator(s), I'd argue they fundamentally enjoy the process (the preparation, the planning, the deferred gratification, the meaning). Perhaps the same is true for people that pursue and complete a Ph.D. -- some get a decent financial payout, but on average, I don't think the degree made them better off financially compared to other alternatives (e.g. holding together some rotting infrastructure with bailing wire). They value the title, the activities, the identity, the community, the kind of work they do.
Seeking a job only for money doesn't really endow much meaning -- (Please, don't take this as an endorsement to go off and work for some harebrained startup when you have better options. :P) -- though I think there is plenty of meaning even in the mundane (e.g. rearranging JSON) to be found if you open yourself to experience (e.g. books with dragons about parsers).
I would like to share my views around fairness and judgement. My apologies if the numbering makes them seem formal; my intention is only to give them a rough ordering.
1. One should not be eager to criticize others.
2. One should seek to understand others.
3. However, one should be willing, intellectually, to differentiate between aspects and assess those differences.
4. It requires some care to balance 1, 2, and 3.
5. One should be honest with oneself, at least, about your conclusions.
6. One should be comfortable with your assessments, particularly if you've thought them through.
7. One should be willing to share these thoughts with others, because debate will improve your thinking, scope, and articulation.
8. One should accept the consequences of what you say.
9. One should learn from what you say.
10. One should not refrain from making assessments only out of a fear that someone will label you as "judgmental".
11. Some people criticize others because they dislike the other person judging others. This is somewhat ironic, because in some cases this criticism is premature. If one judges another without engaging to develop an understanding, I think that is unfortunate. Doing so would be acting in a way inconsistent with one's own values.
All of these "should" statements should be adjusted to the situation. For example, repeated experience, if reflected on fairly, may warrant that some particular people do not deserve the same degree, say, of "benefit of the doubt".
What about domain-specific applications of the technology? I've been in the information security industry for ~25 years and while 'AI' has been increasing in use but is still woefully short of its potential in a few different areas. It seems that someone with deep domain expertise that also has PhD level chops in ML would be a profit machine.
>I have my doubts there will be cushy FAANG jobs for everyone capable of playing around with network architectures in a few years.
Presumably because technical needs of FAANG might be moving in other directions.
Could someone comment why this might be the case, and what other fields might look relevant
I think that the other force is that the skills that are being created by ML Ph.D programs get commodified.
I don't think this will happen, because typing "import tensorflow.keras.*" isn't the skill that an ML Ph.D develops, and it is the part of the skill set that is (and will be) commodity along with the automl stuff.
Constructing a problem, handling the data and running a proper process is harder, and it's the value that will put processes that use ML at risk, and deliver differentiating value for the ones where it works.
> Would you say that you'd have got an AI researcher position without a Ph.D.?
It's difficult to get any true researcher position without a PhD. It doesn't mean that PhD has to be in AI. Research involves a lot of reading and writing papers, which a PhD is supposedly training you how to do.
That said most places will say "equivalent practical experience" and it's entirely possible to be competent in AI/ML without a PhD.
I did a PhD in space science, I now do machine learning in ecology and spent the summer working on machine learning for disaster management. The interesting jobs (to me) are where domains cross, and it's also (hint hint) much easier to get a job doing AI for X than it is doing "fundamental AI". In any case, you're often doing stuff that nobody has done before anyway, but you don't need to spend your life hunting for the new ResNet.
> Also, why is NEURIPS or ICML papers is not a hiring guarantee?
What the OP probably might be implying is that everyone has a publication in NeurIPS nowadays.
I think it goes deeper than that though, publishing in machine learning is broken. Having 10k people at one conference is not an efficient way to distribute research. You have to submit a full paper in November for a conference next Summer - pretty much only computer science does this madness.
What's interesting is how unique this attitude is. In astronomy, for example, conferences are a fun place to catch up with folks in your niche. There might be a few hundred people and probably it'll be single-track. We publish whatever journal is the most relevant and they're generally all considered equivalent. Nobody cares if you publish in ApJ vs A&A vs MNRAS, if your research is good.
There are also concerns that the quality of these venues is decreasing because the pressure to publish in them is so high.
>I did a PhD in space science, I now do machine learning in ecology and spent the summer working on machine learning for disaster management.
Do you think it is possible to that without any background in anything? I mean could someone apply black box frameworks without understanding them. How would they be caught?
> Do you think it is possible to that without any background in anything?
To do machine learning research? Or work in some random domain?
> I mean could someone apply black box frameworks without understanding them. How would they be caught?
Machine learning is rapidly becoming commoditised, but lots of people still don't understand just how much effort it is to get a good dataset and to prep
Domain experts scoff at machine learning people who are trying to solve Big Problems using unrepresentative toy datasets, but also tend to have much higher expectations of what ML can do. Machine learning people scoff at domain experts for using outdated techniques and bad data science, but then propose ridiculous solutions that would never work in the real world (e.g. use our model, it takes a week on 8xV100s to train and you can only run it on a computer the size of a bus).
There are also a lot of people (and companies) touting machine learning as a solution to problems that don't exist.
Overfitting models is probably the most rampant crime that researchers commit.
My question is whether someone could fake it and not be caught/fired. (So yes, I meant: "Or work in some random domain?")
From the second half of your comment it seems that the answer is yes?
Maybe a comparison would help: someone pretending to be an experienced iOS/Android developer without any qualifications or ability would quickly be caught. Since they couldn't produce any working app or use a compiler, and anyone can judge an app for themselves. You can't really just make it up out of whole cloth, people judge the results. You would have to start actually doing that, and if you couldn't or didn't want to, then unless you outsourced your own job or something the jig would be up pretty much instantly. (Unless you caught up.)
So, how about machine learning? Do you think a fraud could land and keep such a job, without any knowledge, qualifications, ability, or even interest in getting up to speed? Just, a pure, simple fraud.
Fake it til you make it isn't a terrible strategy. But pure fraud? If you didn't even make an attempt to learn on the job? You'd get caught pretty fast as soon as someone started asking any kind of in depth questions about the models you were supposed to be training.
I'm not sure you could land a job knowing nothing. Maybe. Depends how hard you get interviewed and whether they know about machine learning. If you could fake a portfolio and nobody questioned it perhaps? I can see that happening in academia for sure.
There are a few problem classes where you could throw stuff into a black box and get great results out. Image classification for example. Fast.ai have made that three lines of code.
So maybe there are a bunch of applications where you could fake it, especially if you were willing to Google your way round the answers.
Would be harder in industry I think, but you find incompetent people everywhere.
>But pure fraud? If you didn't even make an attempt to learn on the job? You'd get caught pretty fast as soon as someone started asking any kind of in depth questions about the models you were supposed to be training.
That's just what I mean. It would depend on someone asking you about it, right? (As opposed to being an iOS or Android developer or running microservices on the backend: in those domains nobody has to ask you anything, it's instantly obvious if you're not building and can't build anything.)
For machine learning, who is asking these questions?
If you throw data into a black box (3 lines of code) and are incompetent, can you please tell me a bit more about where you would get found out?
Let's use your example, ecology.
I show up, I get a dataset, and I put it into tensorflow using three lines of code I copy from stackoverflow.
I lie and bullshit about the details of what I'm doing, by referencing papers from arxiv.org that I don't read, understand, or actually apply. It's just the same 3 lines of code I copied on day 1. I don't do anything on the job.
How long could I last? An hour? A day? A week? A month?
Assuming I am outputting 0 useful work. I'm not doing any machine learning. Just 0 competence, or I make something up by hand or in excel.
As much as I'd like to say you'd get caught quickly, you could probably get away with it for a while in any group that didn't have ML expertise already.
If you really wanted to you could fabricate results and in lots of cases nobody would be any the wiser unless you were supposed to be releasing software. Despite emphasis on peer review and repeatability, science relies heavily on etiquette. If you don't release code or a dataset a lot of times it's extremely difficult to repeat paper results, and that also means it's hard to disprove the work.
It's quite hard to get rid of incompetent people in academia, so I imagine you could get away with at least a year or two.
> Also, why is NEURIPS or ICML papers is not a hiring guarantee? I thought they're highly sought after.
They're sought after, but the conferences have also grown huge. NeurIPS 2018 accepted around 1,000 papers! Based on a query of the DBLP [1] dataset, there were 4,409 distinct authors who had a paper at either NeurIPS 2018 or ICML 2018 (or both). If you add in a few of the other big AI and ML conferences (AAAI, IJCAI, ICLR), the number grows to 10,995 distinct authors, again solely for the year 2018. The field is hot, but is it hot enough for ten thousand people to be automatically hired because of one paper?
There's also decreasing confidence in the big conferences' review processes I think. NeurIPS 2014 actually did a study to estimate how random acceptance was by assigning some papers to two different sets of reviewers and checking how similar the decisions were [2], and found there was a much higher degree of luck in acceptance/rejection decisions than they had expected. I personally have more confidence in the review processes of smaller and more focused conferences (and journals!), though they don't have the same level of name recognition.
A Ph.D.'s worth of first authored NeurIPS/ICML papers will get you a very good job pretty easily still. But AI slices papers very thin and author lists are inflated relative to other subfields of CS. A single paper in one of the major conferences is a pretty marginal contribution, especially if you're in the middle of a long author list.
Also, NeurIPS reviewing has gone to absolute hell. I mean, peer review everywhere has problems. But I've never seen something quite this bad. At this point I think it's safe to say that most reviewers wouldn't even make it to an on-site interview for a faculty position at a research university. That's definitely nowhere near normal. You can't really blame anyone, I guess; the community is growing way too quickly for any real quality control.
Frankly, I think those conferences have outlived their usefulness as anything except marquee marketing events. I'm now mostly attending smaller and more specialized conferences.
I've gone in a similar direction. Only at smaller conferences can you have any kind of confidence that your reviewers are people with actual expertise in the field. That's pretty useful, not only because it makes it less likely you'll get reviews that are very annoying, but also because a review by a knowledgeable person can be genuinely valuable. The big conferences are full of reviews written by 2nd-year grad students, because with this many submissions, any warm body with anything approaching credentials is needed.
Besides just "quality" in the general sense, one thing this has really hurt, I think, is any sense of history or continuity. There are a ton of reviewers who have basically no familiarity with the pre-2010 ML literature, and it kind of shows in both the reviews and the papers that get published. I mean I get that deep learning beats a lot of older methods on major benchmarks, but it's still not the case that literally every problem, controversy, and technique was first studied post-2010.
I do R&D in computer vision, and everyone is shocked when I tell them that the roles don't pay that much better than software engineering. Articles like these make people assume you can make 2-3x what you would as a software engineer at the same level of experience (remember a fresh PhD has ~5 years research experience), and it's usually not the case. The actual multiplier is something like 1.3-1.5x. I had one offer come in under the equivalent software role at the same company.
I also suspect that other fields have similar patterns where a few experts make a ton of money, they're just not as publicized because the media is so hyped on AI. How much you think Jim Keller is making?
TL;DR do the PhD because you love the research. If you want the money do some internships, land a FAANG job, and try and get to senior as fast as possible.
If you want to earn money, create the most revenue.
The problem that most technical people seem to have is that they think a qualification is a licence to print money. It isn't. The majority of people who obtain qualifications have no idea how to make money (that is why they are doing the qualification rather than making money, and this effect increases with the complexity of the course of study).
So if you want to make money and your interest is AI, sure do a PhD...but at some point, you are going to have to work out how you can use those skills to sell something.
I think, at some point, there is going to be a rude awakening. Right now, we this situation where you can tap up some VC idiot to pay your salary for a few years whilst you fuck about on some bullshit. You can get acquired by some megacorp with a ton of cash and have no accountability from shareholders while you light huge stacks of their cash on fire...this always ends. Always.
You are either a bureaucrat chasing qualifications to get no-accountability jobs or you are someone that can create value for other people. I am not knocking any kind of advanced degree (I have one). But the point of it is not money.
As someone presently doing a PhD in machine learning and neuroscience, which includes submitting to AI conferences...
Frankly, I think this is a bubble, and don't expect to ever make anything like that much money, in academia or industry. Yes, deep learning has a lot of successes and industrial applications, but deep learning wizards are not nearly so impossible to replace or replicate. As the field settles more towards engineering than wizardry, and as expectations are recalibrated towards the realistic, it's going to stop raining cash.
I'm a grad student, so my current means are pretty modest, but, well, yeah. Not saving and paying down debts is insane in an economy that is either in a recession or about to be.
I wouldn't make a Ph.D program decision based solely or even primarily based on potential salary. Aptitude and, especially, personal interest should be a stronger guide. AI will likely turn out to be as data science was 7 or 8 years ago: In demand, for now. Then, the market will be flooded and salaries will drop to the average rates for qualified professionals of a given level of credentials & experience.
Please add (2018) to the title. Also, these are celebrities, with some of the most important contributions in the field. They are like the top 0.001% of the field or something equally ridiculous.
I don't think OpenAI is a nonprofit as they are planning to commercially license their technologies. Ok, they limit investor returns at one-hundred times their initial investment, but I'd hardly call that a non-profit in the sense of the word.
It's kind of a weird structure, but I think more like a company than like a nonprofit in the short/medium term. The parent is a nonprofit organization, but it's not the same arrangement as Mozilla Foundation vs. Mozilla Corporation, where all earnings from the subsidiary go to the parent nonprofit. With OpenAI, the parent nonprofit only gets earnings in excess of 100x return. The for-profit company raises investment and returns any earnings to the investors until/unless they hit the 100x return mark. They call this a "capped profit" structure.
If you wanted to command, let's say $5-8 megabucks per year conservatively, I would sell the idea of self-programming systems. Automating software development to not need as many humans and implement prototype systems more rapidly is a business holy grail because it could significantly reduce the need for as many developers and their relatively large salaries. It will happen eventually, it's just a matter of time and who's going to profit that's the question. Furthermore, as automation intensifies over time, there will be overall fewer jobs but their salaries will continue to increase if they are related to automation and further reduction of jobs. If I were an AI researcher, I would be in self-programming systems, insist on patent assignment and consider forming a startup to deliver it as a hybrid cloud/on-prem SaaS... this would be the most obvious path to get rich from it.
Serious question, is this value justified? I don't have any contact with AI development or AI researchers. Do they really bring that much value to the business? It makes sense that for Google, these positions would have almost unlimited clean data. But many companies are not Google.
How do you define "justified"? The field is so new, someone who really knows what they're doing could mean a difference between a large team spending years and getting nowhere remarkable, and improving state of the art every 6 months. With the same exact team. So would you rather get nothing for your money, or something valuable every 6-9 months? That's the question. These people are typically hired after trying pretty hard to get results out of generalists, and failing.
$2 million per year is very low for people at that level if you think about the potential profits that advanced AI can bring and compare the real world utility to that of, for example, a celebrity or professional sports player.
They are at least talking about AGI. If those people get anywhere close to AGI, they should start charging ten or hundred times more, because having that kind of technology could be worth hundreds of billions or more. So even though there is no indication they are close to it, if there is even a remote possibility of getting there, the salaries are questionable in the context of those potential astronomical profits.
This ($1m total comp) is becoming more and more common. I know of a guy making $650k base and $400k in RSUs first year. Just a Java individual contributor, but with a reputation and buddies high up.
Computer science researchers should form partnerships like lawyers and doctors. They could extract far, far more value out of mega corps than they get paid as salary employees. We are at a moment when AI tech is so black box that they could create a company that is itself a black box.
Is it worth putting the time into learning AI/ML at this point? Sure, crazy high salaries- but it seems inevitable that these are the very jobs that will be automated away first.
The crazy high salaries are "ideally" going to the ones who have both the soft skills and the technical prowess to automate away other people's jobs, not their own. This is one of the reasons STEM PhDs have become so attractive to industry--getting a PhD (especially from a good school) is a certification that the individual is capable of producing previously undocumented methods for replacing other people's jobs.
Most AI researchers in industry make a lot of money, $200k+, but that is not so outrageous in the context of big tech companies.
And in fact the vast majority of AI researchers are making $20k-$30k a year, because they are graduate students.