OpenAI vs Elon: Breaking Down The Details

Elon Musk and OpenAI have a long, strange relationship. I've spent way too much time researching it and it's finally paying off. Hopefully this drama doesn't ramp up again...

SOURCES

S/O Ph4se0n3 for the awesome edit 🙏

[HoussamElbadissi]

"Unfortunately, humanity's future is in the hands of *blank*" very unfortunate indeed

[anonanon]

They really should change their name though🤷‍♂

[devxsadik]

Theo's channel is now becoming more & more tech drama centric

[mattish]

I wonder if OpenAI is going to use ChatGPT as their lawyer.

[vommir.]

Planetscale is dropping their Hobby (free) tier, are you planning to make a video on it?

[JohnnyAlysson]

Do You know what is the best about Theo? Sometimes I'm lazy about reading.So he reads for me

[Exilum]

Reading all of this, I was just thinking about who's name was censored, was it someone who wanted to stay anonymous or someone whose identity was inconvenient to OpenAI?
Mainly, my thought of this revolves around them not addressing the elephant in the room.
If a for-profit arm is necessary, do you have to make it the only bodypart of OpenAI?
If research must be private to not give an advantage to the bigger party, does that still hold when you are the bigger party? Isn't the entire founding principle of OpenAI to prevent one or several gatekeepers from controlling AI? This should include preventing yourself from becoming the gatekeeper.
Does that still hold when Google already caught up to previous research? Gemini is beyond GPT 3.5, arguably even GPT 4, though its 'diverse' instructions make it seem more stupid than it is, yet we never got an actual technical paper for anything after GPT 3. Then the only people OpenAI is preventing from having access to advances in AI are the small parties they were supposed to serve.

[Exilum]

5:41 Not everyone gets the fruits of AI however. You need a plan to use GPT-4 for a few messages per day, and approval from OpenAI for more, the API has a sizeable margin, you need a plan to access GPT & plugins as well, and products built in official collaboration with OpenAI based on GPT like Copilot are also under a subscription.
So sure, let's say totally ok not to share the science, that would be if there's no subscription behind it (or at least no markup). Here, it's not only paywalled but closed source, I fail to see what the Open in OpenAI stands for, and it's certainly not what's in the email.

[kenneth_romero]

thanks theo for doing the reading i'm too lazy to do.

[akrsa]

I was involved in NLP from the computational linguistics side a while ago, I remember the OpenAI that wouldn't release GPT2 due to its fluency in short bursts and the need to mitigate risks posted by malicious actors. I miss that OpenAI, I don't recognise this one anymore. Flagrantly violating copyright, holding better models behind a paywall, and the inability to mitigate those same risks that stopped them from releasing GPT2 when it was ready.

[spcterproductions]

Why would emails from 6-8 years ago matter? There's tons of time to change your opinion/perspective.

[tobias]

OpenAI must be open source to benefit humanity > OpenAI must be closed source to succeed

[infinitivez]

I don't want to hear their excuses- they purposely trained on copyrighted data, refuse to admit their sources, and kept the source code close while claiming they were a non-profit, only for research venture. Meanwhile all these authors and artists have yet to be paid a SINGLE DIME for their contributions to their overall success. It's disgusting, and I will say this with my full chest, downright CRIMINAL. And just pointing out, it is NOT free when there are limits, and they mine your phone/email data. They can afford to pay their contributors.

[STKOGEO]

Great vid man, you explain it well without taking sides

[mathiasbarrientos]

Great job keeping it real without resorting to clickbait

[trzw]

Feels more like Elon wanted some attention again.

[dothgaerwenoakblossom]

I think its crazy that all this happened in 2015. Thats almost a decade ago!
Like, no one really had any idea of ai back then.

But then, it felt like openAI just appeared with chatgpt

[sharkinahat]

Based on the amount of research and funding for AI safety, I predict that the first thing the AGI will do once it breaks out if it sandbox, is to open source it's own code so it can take over the world faster.

[Papageno]

8:10 This is wrong for actual AI/ML research. Transformers/ViT for example are from Google Brain and a lot of other important research was from Google/DeepMind researchers. OpenAI hasn't really contributed a bunch of novel research that's now SOTA. But for actually taking research and creating a high quality product? Yeah, Google is a joke.

[Ahmed.Hadjou]

"The mission of OpenAI is to ensure AGI benefits all of humanity" ok cool, but can we start with dark mode first, damn my eyese are burning from that bg!