Microsoft will marry OpenAI

Hey there,

before we get started, my team at AI Academy is doing research to develop new educational programs on AI 🤖. If you’re interested in AI, fill up this short form and I’ll make sure you get a nice coupon code when the programs come out 😉.

This week’s news:

• Microsoft to invest $10B in OpenAI
• Pizza Bytes 🍕: Microsoft releases VALL-E, ChatGPT banned from school
• Tiramisù 😋

Let’s get started 🕺

Microsoft to invest $10B in OpenAI

OpenAI is raising money at a $29B evaluation, and Microsoft seems interested in giving them $10B. That’s roughly enough money to buy Lufthansa, Coinbase, or Zalando, for reference. It seems like Microsoft is working on a complicated deal in which they’d get 75% of OpenAI’s profits until it recoups its investment. After they get their money back, Microsoft would own 49% of OpenAI.

Why does it matter? We shouldn’t be surprised that Microsoft is investing $10B in AI, and we shouldn’t think they’re the only ones.

Let me be clear: EVERY big tech company is making similar investments. Some examples:

• Google has poured billions into Deepmind and Google Brain. They have their own GPT-3 and already talked about domain-specific implementations of ChatGPT (like their AI doctor we talked about last week).
• Meta has a whole AI lab focused on foundational AI research. They released a ChatGPT for scientific research too but were not as careful as OpenAI in making it politically correct and it faced a lot of backlashes (its name was Galactica, we also talked about it here).
• Amazon announced AlexaTM back in August: it’s almost 10x smaller than GPT-3, but performs better.
• As usual, we don’t know much about Apple, but I’ll bet my Mac and iPhone are working on this too.

What’s interesting to me is not that Microsoft IS investing in AI, but HOW Microsoft is doing it.

Everyone else is doing that internally (well, technically Google is doing it through Deepmind, but the two companies are seen as the same thing in the public eye). This is making them much more careful about their risks, and slower in releasing them to the public. People are also biased towards big tech and tend to hold them more accountable than a relatively “new” company like OpenAI (look at what happened to Meta with Galactica).

Microsoft instead is allowing an external company to do the work for them. It’s providing them with the capital and computing power, keeping a preferential way of communication to integrate their technology into their product.

This is great for many reasons. I’ll share three:

• Microsoft doesn’t risk its reputation. Think about all the artists pissed at OpenAI’s DALL-E2. Microsoft doesn’t have to deal with them.
• OpenAI is free to move faster and take bigger risks than the internal AI teams at Meta, Google, or Amazon.
• It’s easier to attract talent. Microsoft isn’t seen as a “cool company to work for”, openAI totally is. BTW, the median compensation for an OpenAI engineer is

Now: More than 1B people use Microsoft’s office product. 345M of them pay for an Office365 subscription. This means that whenever Microsoft will be ready to pull the trigger, they will be able to integrate the most powerful technology humans ever had available into the workflow of 1B+ people.

Suddenly the good old boring Microsoft seems to have a strong seat to be the key player in this AI revolution.

Smart, isn’t it?

Pizza bytes 🍕

Other tech news in 5 lines or less.

🤖 AI 🤖

🎨 Artists are accusing Adobe of tracking their design process to power its AI. This fear started from a viral Tweet that showed how their privacy policy gives them the right to “analyze your content using techniques such as machine learning to develop and improve products and services”. This is how tech loses people’s trust. Not cool.

🗣️ Microsoft releases VALL-E: a model that can clone your voice from just 3 seconds of audio. You can hear the demo here. As the name says, the architecture is inspired by OpenAI’s DALL-E. I TOLD YOU MICROSOFT IS SERIOUS

🧑‍💻 OpenAI is working on a pro version of ChatGPT. You can subscribe to the beta here.

🚫 ChatGPT was banned from New York City public schools’ devices and networks. Their Department of Education was concerned about "negative impacts on student learning." Can/should you stop progress?

🚨 OpenAI says they have a solution to detect AI-assisted plagiarism. It’s pretty cool: basically, ChatGPT will use a statistical distribution to pick its words that are known just by OpenAI. In this way, they can analyze a sentence and estimate the likelihood that that sentence was generated by its technology.

🔐 Crypto 🔐

💸 Binance sees $12B in withdrawals in 60 days. That’s a lot, and some people are scared they may be insolvent.

🥊 The Winklevoss twins (the people who sued Mark Zuckerberg claiming he stole the idea for Facebook from them) are in a $900M crypto chaos with the crypto exchange Genesis. Basically, they own another crypto platform called Gemini that lands money from Genesis to give financial services to its users (why crypto names are so weird). Genesis doesn’t give Gemini their money, they can’t give money to their users who got pissed, so they are suing them. A classic crypto crisis shitshow.

Tiramisù 😋

The sweet part of the newsletter: fun news from the crazy world of tech.

1. Have fun creating sick beats using AI.
2. Ryan Reynolds uses ChatGPT to create an ad for his company Mint Mobile. All the cool people use AI 😎
3. Super cool video about how Roman concrete was made. This thing got almost 8M views on TikTok. You gotta love TikTok sometimes.

‍

I’ll talk to you next week.

Ciao 👋

‍