How OpenAI Came to Be

Welcome to the No Longer A Nincompoop with Nofil newsletter.

Here’s the tea 🍵

• The inception of OpenAI 🧬 

  • Founding 👨‍👨‍👦‍👦

  • Shaky relationship with MS 🧗‍♀️

  • For profit restructure 🤑

  • The most important contract ever ⚖️

  • Funding 💰

• Claude can see everything 👀

We (Avicenna) recently worked with an industry leader in Australia (Claimo) and 2X’d their output, saving them 100+ hours of work already. We’re going to 4X their revenue next year.

If you are interested in doing this for your business, reply to this email with one sentence / little bit about your business. Lets chat!

It’s been two years since the release of ChatGPT and the start of the AI revolution.

So, how did OpenAI, the creators of ChatGPT, come to exist?

Let’s take a look.

How it all began

OpenAI was founded by Elon Musk, Sam Altman, Greg Brockman and Ilya Sutskever.

But, why?

It all started with this email from Altman to Musk.

From this email, it seems Elon is slightly interested. Turns out he was incredibly invested.

But, why?

What caused him to provide tens of millions in funding to start OpenAI?

Apparently, it’s all because Larry Page called Musk a “speciest” for favouring humans over the inevitable digital “life-forms” of the future.

You might be wondering then, how on Earth has Elon started his own company, xAI, and why he’s no longer with OpenAI.

Simply put, OpenAI realised pretty quickly that they needed money.

A lot of money.

To compete with Google, they needed Google type of money.

There are only a handful of companies on the entire planet that can provide this kind of money, one of which is Tesla.

So, it makes sense that OpenAI would partner with Tesla to fund their work, right?

Even Karpathy thought so.

Turns out, not really…

Before there was even mention of Tesla, OpenAI considered a hosting deal with Microsoft that would provide them with compute on the basis that OpenAI would agree to promote their products.

Naturally, they declined.

Microsoft later provided $50M in funding in “good faith” which OpenAI accepted.

Where it went south for Elon and the rest was the idea that Elon needed to be CEO and in turn have all the power.

Brockman and Ilya sent this to Elon and Altman.

If you go through the emails, it really seems like there is terrible communication between all of them. They are never on the same page.

There is constant suspicion on other peoples motivations.

This suspicion does make sense considering they believe they’re building a superhuman artificial intelligence.

At some point, Altman planned to release a cryptocurrency to raise money which Elon shut down.

Besides this, its not too clear how things went so astray with Elon, and how Altman, Brockman and Ilya went on to run OpenAI for years and release ChatGPT.

I guess they never really trusted each other since they tried to oust Altman in 2023 and soon after that, Ilya left to found SSI.

What’s strange is that Brockman and Ilya seemed to be partners against Altman.

In the end, Altman got Brockman on his side and kicked Ilya out.

Tough luck.

You can read all of the emails that have been released on Lesswrong here [Link].

Things are as shaky as ever

Fast forward five years and now OpenAI has daddy Microsoft funding everything.

All is well and happy… right?

Absolutely not.

There is serious tension between MSFT and OAI right now. As of now, both parties have been looking elsewhere for the things they need.

Microsoft has been building in-house AI models. OpenAI has been building partnerships with other companies for compute.

And it’s not just compute.

OpenAI is building their own AI chips with Broadcom and TSMC [Link].

They’re serious about it too, considering they’ve hired the people who built Google’s TPUs - Richard Ho and Thomas Norrie.

But, why?

1. They both have their own agendas

2. They both don’t want to be completely dependant on the other

For one, Microsoft “acquihired” hired Mustafa Suleyman, former founder of AI startup Inflection, which did absolutely nothing besides buy some GPUs to increase NVIDIA’s share price (this is a joke but not really).

Apparently, people at OpenAI haven’t taken too kindly to Mr Suleyman screaming at them for not delivering tech to Microsoft fast enough.

This isn’t the first time he’s done this either, and, is apparently the reason why he left Google DeepMind many years ago [Link].

But there’s a much larger play here, especially for OpenAI.

Going for-profit

OpenAI is going full for-profit, to the surprise of absolutely nobody.

Just look at this filthy org chart.

If I was Elon Musk, I’d be pretty upset too, considering he funded the initial company and could have had a stake in what could be one of the most valuable companies on Earth.

Will Sam Altman take equity in OpenAI?

Yes, even if he doesn’t want to.

Investors are insisting that Altman get a stake in OpenAI.

Equity would tie Altman to OpenAI.

Remember, equity > salary.

The rumour is that they want Altman to have a 7% stake.

For reference, OpenAI is valued at $157B.

A game of semantics

Microsoft has invested over $15 Billion in OpenAI. It’s their biggest bet on AI and getting to AGI.

OpenAI has been quite clear that we will get AGI sometime this decade. Sam Altman even released a blog saying we are only “a few thousand days” from AGI.

A few thousand days does not sound like that much, until you really think about it. It could be a few years to, well, over a decade, which is definitely not what OpenAI has been alluding to.

There’s one particular reason why the coming of AGI is so important, both for OpenAI and Microsoft.

You see, in their contract, it states that if OpenAI ever actually reaches AGI, then they have the power to drop any responsibility to Microsoft and are not required to owe them anything.

Yes, you read that right.

In this case, Microsoft also loses any and all access to OpenAI technologies.

There is only one single question that matters at this point.

What on Earth is AGI.

Soon, there will be a court case.

OpenAI v Microsoft for AGI.

For the most powerful AI on the planet.

A judge, who perhaps has no idea what AI is, will have to determine what AGI is.

It’s going to be hilarious, and I can’t wait for it to happen.

The mother of all court cases.

Although, there is another way this goes.

They get AGI and simply never announce it. Considering OpenAI is tied to the US Government and the NSA, I wouldn’t be surprised if this happens.

Update:
I wrote this a while back and it’s being published now.

Thinking about this now, it does seem like building on Azure and using OpenAI through Microsoft might actually be a bad idea?

If their relationship goes sour and courts rule in favour of OpenAI, Microsoft customers are screwed. Would love to hear other people’s thoughts on this.

Funding

OpenAI has raised $6.5 Billion in a new funding round.

That’s a lot of money.

You might think then, that very few companies would have been able to invest in the round, or that they may have struggled to raise such a large number.

You’d be wrong.

Turns out they were oversubscribed, meaning they have a lot of people willing to give them a lot more money than what they want to take on.

In fact, not only do they have to decide who to cut out of the funding round, because there was so much interest from so many different parties, they had to set a minimum amount:

$250,000,000.

A quarter of a billion dollars.

It is reported that so many investors were ready to fund them that the excess demand was in the billions of dollars.

They left billions on the table.

In fact, OAI has so much leverage right now that they’re toying with VCs. OpenAI has openly told VCs to not back their rivals Anthropic, xAI or SSI [Link].

If you do, you’re out.

Give Claude access to your data

Anthropic has released the Model Context Protocol (MCP). It is a standardised framework that lets an AI, like Claude, access information from another system.

This allows Claude to access essentially anything; a website, a database, even your local files.

Simply put, think of MCPs as tools.

Need to access a website?

Use the “website searcher” tool.

Need to access a database?

Use the “database reader” tool.

There’s only problem with using MCPs - it is bloody complicated to setup.

But don’t worry, I got you covered.

If you use Claude, which I recommend you do, then you should download the desktop app. Then, install this MCP.

This MCP lets you install other MCPs easily. This is how you can install it.

1. Go to your Claude Desktop App

2. Go to Settings

3. Find the Developer Tab

4. Click the “Edit Config” Button

Edit the file and add this code in it.

  "mcpServers": {
    "mcp-installer": {
      "command": "npx",
      "args": [
        "@anaisbetts/mcp-installer"
      ]
    }
  }

Save. Done.

Now, you can use this “mcp-installer” to download other MCPs.

There are a number of different MCPs you can install.

The best thing about the MCP is that I can create an instance and anyone else can use it. These are community instances people have created you can download and use.

Let’s take the “Sequential Thinking” example. Just tell Claude its name and to download it. You can find the name of the mcp by clicking into it and scrolling down to the configuration.

Just remember, some MCPs have arguments you need to provide them in order to work, like the Slack one.

Without providing these, it won’t work.

With this, you can do something like search the web for the latest architectural styles, generate AI images based on these styles, and then create a website with these images [Link].

The power here lies in being able to connect so many different functionalities together.

If you’re working on a project locally on your computer, Claude can now access your files and edit them directly.

There’s no need to copy and paste from Claude anymore. It can do it all for you.

See where this is going?

The AI is doing all the work for you. All you have to do is guide it.

I wrote a lot more for this newsletter but emails can’t be that long :/. You’ll have to wait for the next newsletter 😉.

If you enjoyed this newsletter, please consider subscribing or contributing, it helps me keep writing more newsletters 🙏.

As always, Thanks for Reading ❤️

^{Written by a human named Nofil}