Been working with the open AI team for about 6 weeks now optimizing the parallel training in Azure. The culture there is amazing, they are moving so fast, are extremely determined, and are almost zealots about getting to AGI (throwing every idea at what could make that happen, trying it out, iterating, again in a very fast manor) I don’t think there’s any hope the big tech companies can catch up. The culture there is not bound by the corporate non sense those companies have in place. OpenAI -> Go as fast as possible, try anything, don’t let anything get in your way, make an AGI google, meta, etc -> make sure you get approvals, check with PMs, have some meetings about things to decide every aspect. Also don’t forget to do sprint planing and score things, run things through security, and safety, and AI ethics group and accessibility. Also check if it’s inclusive to all groups too. It’s lost cause, OpenAI will win.
Sounds like Meta (going by what people on Blind say). Do you guys write unit tests?
Not even like meta anymore since regulators make us jump through hoops to ship although yeah we don’t write unit tests, I prefer it that way vs the other extreme of TDD. The latter is so useless unless it’s a critical system
So non critical product teams dont really need to write unit tests? Unfamiliar to me since I think I'm on product that is a critical system
Adding more weights + data to a narrow AI doesn't make AGI. We're no closer to AGI than we are to teleportation. Its science fiction right now because the theories of mind necessary haven't been invented yet. Anybody saying otherwise doesn't know what they are talking about or is selling snake oil.
You don’t know that, and the best AI researches don’t know that. This stuff is very hard to prove (ML it’s hard to theoretically prove anything)
I think OpenAI should have better security teams based on data leaks that happen every now and then.. No shit
Innovation should not bend to anything. At the same time incorporating bleeding edge solutions to systems that impact everyday customers should have regulations in place. Those who work on a bleeding edge sometimes too bleed.
I've seen this before: move fast and break things. Good luck
Hasn't legality already started catching up to them?
As opposed to Google and amazon move slow and never ship useful stuff
lol Italy didn’t give a fk and I reckon other govts would be joining them soon
That exact flow worked the same as every company you listed in the beginning. After users scale and the stakes get higher, it’s not reasonable to afford those types of mistakes. In the beginning, it’s great. OpenAI seems like a really cool company though.
Cuz all our talented AI folks jumped ship to OpenAI lol
There you go. Google claiming wholesale talent business
What happened to Lamda?
All companies start that way...when they grow too big, that bullsht model you're referring to doesn't work our research and military division has the same culture and we build autonomous vehicles, design parts for space x, create self-healing materials, experimental jet engines, used augmented reality to prototype builds, enterprise scale 3d printing business, AI deployments etc but when it comes down to funding actual production builds for customers, we need approvals to avoid going bankrupt by everyone doing whatever they wanted spending money everywhere
Sounds slow. I would hate working there.
true