OpenAI employees did not want to go work for Microsoft • Hacker News Board

Re: OpenAI employees did not want to go work for Microsoft

#261

Post by vitorgrs » Tue, Dec 12, 2023, 1:52 AM UTC

VirusNewbie wrote: post #227

hughesjj wrote: post #136

Earlier quoted context omitted.

IMHO, today, Google takes the cake for slowest. Meta is still probably the most agile, and Amazon is super hit or miss (so most variance) Idk where Microsoft would fit in that hierarchy, like Amazon it's kind of hit or miss but with less variance and extrema. From what I've seen and heard, both Gaming and Azure are pretty darn competent these days from an engineering and product perspective. Not perfect of course, bu…

Do you think Microsoft would have a competitive foundational model if not for OpenAI? Cause Google and Meta seem like they’re top of the pack there, with AWS at the very least releasing some.

Microsoft already had competitive models, especially Megatron-Turing. They were even using on Bing Chat before... I'm pretty sure they still use their model on Bing Chat today (Balanced mode).

It was a competitive model for the time, it was made in 2021 if I recall in cooperation with Nvidia. The largest model at the time (and still remains one of the largest models ever made).

https://developer.nvidia.com/blog/using-deepspeed-and-megatr...

The thing here is that OpenAI managed to create GPT4 - that no one matched yet.

Re: OpenAI employees did not want to go work for Microsoft

#262

Post by hughesjj » Wed, Dec 13, 2023, 10:08 PM UTC

vitorgrs wrote: post #261

VirusNewbie wrote: post #227

Earlier quoted context omitted.

Do you think Microsoft would have a competitive foundational model if not for OpenAI? Cause Google and Meta seem like they’re top of the pack there, with AWS at the very least releasing some.

Microsoft already had competitive models, especially Megatron-Turing. They were even using on Bing Chat before... I'm pretty sure they still use their model on Bing Chat today (Balanced mode). It was a competitive model for the time, it was made in 2021 if I recall in cooperation with Nvidia. The largest model at the time (and still remains one of the largest models ever made). https://developer.nvidia.com/blog/using…

Yup, I took remember taydolf swiftler from way back when. Hey even if it didn't pan out, it still shows they'vd been trying and innovating for a while

Similar for Cortana.

Now, I don't like any of those products, but, still, effort is seen.

Re: OpenAI employees did not want to go work for Microsoft

#263

Post by hughesjj » Wed, Dec 13, 2023, 10:11 PM UTC

a_wild_dandan wrote: post #210

hughesjj wrote: post #136

Earlier quoted context omitted.

IMHO, today, Google takes the cake for slowest. Meta is still probably the most agile, and Amazon is super hit or miss (so most variance) Idk where Microsoft would fit in that hierarchy, like Amazon it's kind of hit or miss but with less variance and extrema. From what I've seen and heard, both Gaming and Azure are pretty darn competent these days from an engineering and product perspective. Not perfect of course, bu…

Huh, I've heard several people refer to Microsoft as a "retirement home." Maybe that's just an ageism thing.

I've heard the same about Google. It's definitely not untrue at MSFT today either.

My hot take is that 2018 Google was late 90s MSFT more or less (juggernaut resting on their laurels and cashing in on their monopoly). Today, they're more like mid 2000s MSFT. Here's hoping 'tomorrow' they'll have a similar comeback.

Re: OpenAI employees did not want to go work for Microsoft

#264

Post by ElevenLathe » Wed, Dec 20, 2023, 2:18 AM UTC

scarface_74 wrote: post #176

ElevenLathe wrote: post #74

Earlier quoted context omitted.

Even from a compensation point of view, they presumably have a better chance for a big exit of some kind at a startup vs as a Microsoft employee.

A startup that was going to implode? And if you think random “some kind of startup” will have a realistic chance of some type of exit, I got a startup idea where you can be an unpaid employee for “equity”. In what world in today’s market do you really think open AI would ever see its value at an exit based on their current revenue model?

The choice we're talking about is between:

1. Bail to Microsoft. Your salary is matched but your equity (in whatever OpenAI entity) is forfeit. You have to move to Washington potentially, and definitely have to deal with job switching stress. If you're lucky, you will get the same corporate RSU grant that the Office team got for shipping on time.

2. Stay at OpenAI. You don't have to move or have the stress of starting a new job. Maybe it implodes and you have to take your very-marketable skills elsewhere. Maybe it all works out and your equity is worth millions.

I know which I would pick.