A simple Google search would show how much FB has contributed to the AI/ML community.
I hate seeing these comments in HN. No substance. Just saying whatever to jump into the FB hate bus.
I hate seeing these comments in HN. No substance. Just saying whatever to jump into the FB hate bus.
There's a difference between running a research lab and a product organization. I used to work in a research lab (Nokia Research). Lots of cool research. But product divisions routinely ignored what we were doing. Not invented here, not a priority, we're busy moving deck chairs around on our Titanic, etc. Very little industrial research translates into customer/product/share holder value.
Facebook, like several other big companies, pays a lot to keep some very smart people on staff. They constantly come up with new stuff that they publish about and that also result in patents. That's not the issue.
However, from there to product features is where it goes wrong. Most of Facebook's product portfolio hasn't changed substantially for quite some time. And they haven't really launched any noteworthy new stuff either. Publishing a paper doesn't really count. It's nice and it will no doubt help other companies to succeed where Facebook is clearly struggling to succeed.
Most large companies struggle with this. A lot of former industrial research labs have either disappeared entirely or became a shadow of their former selves. IBM, HP, etc. used to do lots of blue sky research. Xerox Parc was a big deal and they obviously failed to productize. But of course inspired Apple and quite a few others. It's a tough topic. I don't think many companies have figured this out.
There are positive examples of course but mostly the return on investment for open ended research in a corporate setting is not as high as it should be considering the valid research that gets published and then fails to get monetized by the companies doing the research.
OpenAI, is the smart counter move. Spin off a company, fund it, and then license back the technology. MS seems to have learned from it's past efforts in MS Research.
It’s also a fair criticism of most of big tech wrt AI, and a lot of AI/ML: there’s a lot of cool-sounding R&D, but so far there are few (arguably no) resulting products.
Maybe we’re starting to see that change, maybe not. Bing/ChatGPT is promising, but ... nascent. A lot of this has echoes of the self-driving revolution.
Whenever I read techno-optimists on this site, I cannot help but think back to 2014-ish when they all predicted that there would be no truckers in 5 years and dramatic changes to infrastructure to support a wave of self-driving vehicles.
The inevitable nature of large corporations is to become bureaucratic and slow, yet many of them exist for decades without being innovative. It's sufficient to just buy out whoever is innovative in the market - Microsoft just bought into OpenAI big time. Other than incrementally developing its 40yo operating system Microsoft copies AWS, Netflix for games, Sublime Text and Macbooks while throwing some random cash at garage projects that amount to effectively nothing that can exist without subsidies (Hololens).
They are talking about actions in the last year or so.
Meta's open source projects that they own are maintained now by eager volunteers, not staff.
Anybody could see how Facebook is floundering, Instagram can see the iceberg approaching, and changing anything major in Whatsapp will piss off literally billions of people.
They've cornered themselves, so now they're reaching for literally anything that could save them. That seems reasonable as a failing company, except when every action they make affects entire markets and legitimizes trends on a very shaky basis.