Meet the $4 Billion AI Superstars That Google Lost
bloomberg.com
bloomberg.com
I am absolutely sure there is hesitancy at Google of overturning their search model (I.e. money printing machine) which led to the loss of the scientists. But let’s wait until they actually turn some semblance of a profit before claiming Google “lost” $4 billion.
So that leaves that they weren't allowed to create these products, which is the same as saying that they were forced to work on other people's ideas ... ideas that mostly didn't pan out.
Or that Google didn't pay them what they were worth (by the ultracapitalist definition of worth: you're worth what someone is willing to pay).
Perhaps a combination of both: that they were forced to work on unsuccessful products and couldn't get raises/promotions as a result of that. That management had it's own ideas they really, really, really needed to succeed, and when it failed they dragged a lot of people down with them.
I must say, that sounds exactly like what you'd find happening in large companies.
I always thought that's what Alphabet was for, to be honest. Trying to keep a sort of ecosystem of acquisitions and Google spin-off startups with direct Google investment.
1) They exhibit an exceptionally risk-averse stance from a legal standpoint. Consequently, matters with even the slightest chance of legal repercussions (usually revolving around privacy) undergo an exhaustive examination that can take years. Often things get killed just because of the spectre of legal/privacy review.
2) Gaining unanimous agreement within the company for a particular course of action, especially necessary for projects demanding substantial resources, can truly be likened to an arduous Sisyphean endeavor.
3) Anything that doesn't immediately and unequivocally exhibit the potential to evolve into a multi-billion dollar industry is met with significant doubt.
a competitive system of graduate school where some number of people are assigned a ranking number upon completion of the studies != "best people"
the most US dollars per year with maximum total compensation != "best environment"
this is not only annoying, but creates an echo chamber among management, investors and lots of classes of "camp followers" that assume these numbers are actually the goal.
building companies to maximize Total Compensation and Total Return on Investment are exactly the root of unending race to the bottom consumer practices, addictive content and dark pattern customer services.
$0.02
When has that ever been the case?
Just because some company says "We do X", you shouldn't fall for the trap of believing them.
All of the FAANG companies say the same thing, as do quite a few non-FAANG companies.
Not the OP, but let me reframe the point here: I do not dare to say that Google hire the best; but what they goes quite close to get the top talent from some demographics and with certain skills and give the resources, some amount of freedom, and guidance and then collect the astonishing research results.
Personally I do not think that Google Brain has more talent than TU Berlin/TU Munchen, but the main difference it’s that the former has more access to resources and certain amount of intelectual freedom to spend a bit of time in something.
End result: The top conferences in AI, Machine Learning and so forth are dominated by Google and other big companies.
the best would also relate to the size culture, goal of the company, existing products etc etc
why would the best seek employment? i chat with a guy who claimed to charge 5000 per phone call, he would be much to busy studying to even consider employment.
You don't have to be a google apologist to acknowledge the fact that they did invent the transformer. People leaving google doesn't change that fact.
Pivotal moment was BERT, which demonstrated that with some additions transformer can achieve very significant improvements, and inventor is still at google, and transformer creators got lot of hype from that.
Also, most important part is attention, which also was not invented at google.
So there you go...
And AI already can do that. E.g. help reading a sales effort; help qualify an investor; help target a pitch. Doing just an adequate job at these things can make agents more productive, sooner. Do it for car sales or real estate agents or financial services, that's money in the bank.
No matter how much hardware you throw at GPT, it's not going to reason against its own ingested material, but it could be a foundational part of an actual GAI the way that a battery led to the creation of a laptop. (maybe not that far apart, but the analogy stands)
It's not about these people being irreplaceable or not.
The key word here is all of the paper's co-authors have left (8 in total). That's a sign of talent bleeding. and likely isn't limited to that team alone.
It's an indicator that Googlers are finding better and more attractive opportunities outside the nest.
Depending on the scale, that can be fatal.
That problem is not unique to Google. The bigger the company, the bigger this problem is. Big corporations are primarily concerned with keeping share holders happy. Share holders by and large are not very technical. Nor are most executives. And there are several layers of management in between them and the people actually coming up with new stuff. Communicating anything technical and the sense of urgency about doing something with that information to those people is very hard.
For any meaningful budget to be allocated to some new and exciting but unproven thing, a lot of very busy people in that org tree need to be in agreement that that's a good thing before anything happens. That happening is just not very likely. It's literally easier to get investment outside corporate hierarchies. That's why startups are a thing. Google files lots of patents that they will probably never do anything with. And that's just the ideas they've had that get patented. Most of what Googlers come up with never gets actioned in any way.
In this case what happened is that a bunch of smart Googlers invented something, wrote a paper about that, and then some external people read that paper and got inspired. Multiple startups got busy working on those ideas and Google ended up sleeping through the whole AI revolution until belatedly realizing they needed to get their shit together. Probably around the same time their shareholders noticed that something was up. Now they are in catch up mode. And they'll probably pull a thing or two together. But the point is they are reacting to the outside world and not acting on their own insights or inventions. They are one of many big companies that now know AI is a thing.
They'll likely miss the next big thing as well for the same reasons. Even if they invent it.