Facebook Runs On AI - But 70% of Its Engineers Who Use AI Aren't Experts (wsj.com) 91
An anonymous reader shares an excerpt from a WSJ report: AI algorithms are inherently black boxes whose workings can be next to impossible to understand -- even by many Facebook engineers. "If you look at all the engineers at Facebook, more than one in four are users of our AI platform," says Mr. Candela. "But more than 70% [of those] aren't experts." How so many Facebook engineers can use its AI algorithms without necessarily knowing how to build them, Mr. Joaquin Candela, Facebook's head of applied machine learning says, is that the system is "a very modular layered cake where you can plug in at any level you want." He adds, "The power of this is just hard to describe." Pieces of that platform are performing all kinds of "domain-specific" tasks across Facebook's properties, from translation to speech recognition.
It's called "specialization" (Score:5, Insightful)
You might as well say that Facebook's AI runs on electricity and (generously) 99% of Facebook's engineers aren't experts in electricity generation and distribution either.
Re:It's called "specialization" (Score:5, Insightful)
Agreed, this is nonsense. When I was programming I used compilers, I sure as heck was not a compiler expert. At best I could be an expert in using compilers and that would be fine.
Even with teams as large as they are now in common environments, you can afford to have one expert at compilers creating your optimum build packages
It should even be uncommon to have the expert utilizing the technology they are an expert at building. Those roles are often separated out for good reason.
Re:And anyway...FTFY (Score:2)
The real headline would be, then, something like "Facebook engineers work with Russian intelligence...
Re: (Score:2)
In a nascent field, most of the users are also experts. It comes with the territory.
The specialization into designers and expert users indicates maturation of the field. This is what happens when people take your technology and build something new on top of it.
In fact, this specialization may be the only universal metric of maturity---anything else I can imagine does not apply historically.
Re:It's called "specialization" (Score:4, Insightful)
And I'll bet not a one of them is an expert car designer. Or even capable of designing a basic internal combustion engine....
Re: (Score:1)
Experts (Score:5, Interesting)
Are there answers to these questions behind the paywall? I'm guessing no.
Remember one thing when talking about experts (Score:1)
Expert these days has come to mean someone with a paper credential.
Even their head of machine learning is calling it an AI platform. He's clearly one of those "experts."
libc (Score:2)
Isn't this as if I were using libc or, god forbid, libc++, boost even, while not being an "expert" there? I'm pretty certain it would take me an obscene amount of effort to even replicate some of the stuff in boost, for example.
Isn't this all that modern development has been trying to achieve since forever?
Re: libc (Score:1)
AI? (Score:2)
Congrats (Score:5, Insightful)
Re: (Score:3)
Re: (Score:1)
I've come to the conclusion that Facebook draws in mostly sub 100 IQ people - which is not coincidentally everyone that votes either Democrat or Republican.
The most intelligent people are not part of that system because they know Facebook, like our two party system, is inherently untrustworthy.
Re: (Score:2)
define meaningful
Re: (Score:2)
I won't argue about whether those jobs themselves are useful or good.
I'm only saying that people who are doing those jobs are doing it with the help of a system they don't understand nor do they need to understand it.
Creating AI Models is Easy (Score:3)
It is prepping the data that is hard. The Machine Learning Algorithms have been established for a long time. The big limiters on it have been processing power and decent data sets.
Re: (Score:2)
If more things move to AI done by people who don't understand how AI works, well, the equifax breech will seem like nothing in comparison.
I view the process of collecting and selecting data to be a very different activity than programming to the point that conceivably you could have non-programmers doing a better job than seasoned software engineers. It's a different skill set in my experience.
Re: (Score:2)
It is prepping the data that is hard.
The users do that. FB just asks them to "tag" their friends in photos.
Security (Score:5, Insightful)
Re: (Score:3, Informative)
How many developers understand encryption algorithms that they use for security... this is the point of libraries?
Not enough developers understand encryption algorithms (and it shows), and libraries don't help because they still allow the misuse of encryption.
Re: Security (Score:2)
Well, drat! I guess we should all go back to using telnet...
Re: (Score:2)
... Well, drat! I guess we should all go back to using telnet... ...
On the contrary... the solution is to use encryption in a better manner. Education, not reversion, is the answer.
I drive a car... (Score:4, Insightful)
but I'm not a fucking mechanic.
It doesn't work for me... (Score:3)
Re: (Score:1)
It's working exactly as intended. Furthermore, from my understanding, Recent Posts first may not show all newer posts. The user isn't in control, Facebook is. Presumably, much of Facebook's AI is designed to keep users on site long as possible, keeping them coming back, and bringing in new users.
Re: (Score:2)
It's working exactly as intended....
By annoying me to the point that I leave the Facebook site quickly and frustrated that the site doesn't do what I want it to do? Shouldn't a good AI learn what I want and give it to me?
Re: (Score:2)
It's working exactly as intended. Furthermore, from my understanding, Recent Posts first may not show all newer posts. The user isn't in control, Facebook is. Presumably, much of Facebook's AI is designed to keep users on site long as possible, keeping them coming back, and bringing in new users.
You mean that users have to look through all older posts to ensure that they have seen all newer posts? Then yes, FB' AI successfully forces users to be on their site for a long time. Not sure that has to do with coming back and bringing new users part though...
Can't be an expert in somebody else's Intelligence (Score:1)
Re: (Score:2)
Re: (Score:1)
Re: (Score:1)
Normal, expected, and a problem (Score:4, Interesting)
The real problem with modern, "deep learning" AI is that usually not even the experts can tell you how such systems work.
The most they can tell you is:
* The model makes the choices we labeled on our training data set
* We add stuff to the training set as it makes detected mistakes
The weights in the neural network after training become an opaque fuzzy partition of the training set.
Does this inspire confidence in you? Me neither.
Well yeah, that's the point (Score:2)
The entire point of expert systems [wikipedia.org] is to distill the reasoning process of experts so that you don't have to have one of those available to you at all times.
In artificial intelligence, an expert system is a computer system that emulates the decision-making ability of a human expert
Honestly, having as much as 30% of the users being experts kind of sounds like a waste to me.
Comment Subject: (Score:2)
Did AI solve the fake news problem? (Score:1)
99% of developers use OSes or compilers (Score:2)
As a machine learning developer (Score:2)
It's about having good tools. You don't need to know the details of AI to go through the process of building a data set and making practical use of machine learning. It's somewhat like how many programmers don't know digital electronics or assembler but are still able to write software. We're far enough along with machine learning that you aren't starting from scratch for each project, it's more of a modular system and much of it can be setup and configured with GUI tools now.
The original quote in context (Score:2)
More of the article:
'If you look at all the engineers at Facebook, more than one in four are users of our AI platform,' says Mr. Candela [head of applied AI]. 'But more than 70% [of those] aren't experts.'
How so many Facebook engineers can use its AI algorithms without necessarily knowing how to build them, Mr. Candela says, is that the system is 'a very modular layered cake where you can plug in at any level you want.' He adds, 'The power of this is just hard to describe.' Pieces of that platform are performing all kinds of 'domain-specific' tasks across Facebook's properties, from translation to speech recognition.
This implies of the 25% of FB's engineers who use company AI services, 70% invoke it via a simple API without delving into the infrastructure or tuning it themselves.
Therefore only 7.5% of FB's AI users (30% of 25%) pass the Turing Test.
Most software (Score:2)
Also in today's non-news:
Most software runs on an operating system, but 90% of the software engineers who write applications aren't OS experts.
AI is a myth (Score:1)
How much AI is needed... (Score:2)
Give me a break, FaceBook software is idiotically trivial. The single hardest task FB engineers face is how to distribute the data, and that is a problem that is mostly solved by hardware.