The real winner in AI

Hello and welcome to Ben's Bites, the #1 daily AI newsletter. Like your attention span, it's short and excitable.

Let's get to it.

p.s. I'm looking to hire a writer. Reply to this email and rewrite today's 'Ben's picks' section. Make it casual, witty and informative, like talking to a friend. Kind of a mixture of Milk Road, Mailbrew and The Hustle. (lol this is like a prompt). Looking for someone who loves this space and is always checking out new things. If your application is promising, I'll follow up 😊 !

Sponsor | Join Discord | Host or join a meetup

🤌 Ben's Picks

The real winner in AI so far? Microsoft. Microsoft eyes $10 billion bet on ChatGPT.

Whew, who saw that coming?! Not me. A little recap of how well they're doing:

• $1bn in OpenAI (2019)

• Owns GitHub which released Copilot

• Dalle powers Bing Image Creator

• Valle which produces an AI voice from 3s of input

• GPT in Word, PowerPoint, Outlook etc

• GPT-powered Bing

• $10bn, 49% stake in OpenAI + 75% profit

The last point is from the recent news reported by Semafor. Apparently, Microsoft is eyeing up a $10bn deal with OpenAI. They'd get 49% ownership (49% to investors and 2% to OpenAI's non-profit parent).

Also 75% of profits would go to Microsoft until it recoups it's investment.

Satya Nadella is a boss.

Oh and you may have overlooked VALL-E which I included a few days ago. Trained on 60k hours of English speech, it can generate speech in any voice from only 3s of that voice input. (link) A good post on it here.

Is this huge win for Microsoft somewhat of a loss for OpenAI and the rest of us? I guess we will wait to find out.

Oh, and don't feel bad for Google sitting in their code red bunker.

• Google paid ~$650M to acquire DeepMind in 2014

• OpenAI is currently "valued" $29 Bn

• OpenAI's revenue is "tens of millions"

• DeepMind's 2020 revenue was $1.13 Bn and was profitable (source)

This year will be a big one for Microsoft vs Google. Who you betting on?

Two relevant reads from Ben Thompson:

AI and the big five. (link)

More on Google and AI; OpenAI, Integration, and Microsoft. (link)

🛠️ Cool Tools

• Steno - Your favourite podcasts, fully transcribed. (link)

• restorePhotos - Restore old and blurry face photos for free with AI. (link)

• Talk With Wind - Record sounds of anything (birds, wind, fire, train station) and chat with it. (link)

• API for Tortoise TTS - A usage-based, pay-as-you-go API for Tortoise based on open-source code. (link)

• Pick a Pic - An app for collecting human feedback on AI-generated images for supporting academic research in AI. (link)

• Sentient - An AI chatbot for your company documents. (link)

🤓 Miscellaneous

• How AI will reshape politics, education, creativity, and pet ownership? (link)

• What is an optimum degree of LLM hallucination? (link)

• AI in war: How artificial intelligence is changing the battlefield. (link)

• Search is Gmail’s achilles heel & the wedge (with AI) to build a great next-generation email client. (link)

• Six OpenAI rivals Google and Microsoft are watching. (link)

• Why the Chaser is going to put up a paywall? Hint: AI. (link)

• Isomorphic Labs, an AI-first biology company announces new scientific advisory board. (link)

• Sketchy ChatGPT app soars up app store charts, charges $7.99 weekly subscription. (link)

• RETRO models are a giant capability unlock for LLM tech, and they're shockingly under the radar. (link)

• My lawyer, the robot. (link)

• Making beef stew with ChatGPT. (link)

• Can AI pass the bar exam? (link)

• Large language models like ChatGPT say the darndest things. (link)

🎓 Learn

• Build a GitHub support bot with GPT3, LangChain, and Python. (link)

• Reinforcement Learning - A gentle introduction to learning from decisions. (link)

• Use ChatGPT to do data science for you. (link)

🔬 Research

• Variance Reduction System - A new offline RL framework developed by Meta to help ensure a more equitable distribution of ads. (link)

• D3VAE - Bidirectional variational auto-encoder equipped with diffusion, denoise, and disentanglement for generative time series forecasting. (link)

• 3DAvatarGAN: Bridging domains for personalised editable avatars. (link)

• Diffused Heads: Diffusion models beat GANs on talking-face generation. (link)

🖼 AI images of the day

They grow up so fast!

🤗 Share Ben's Bites

Send this to 1 AI-curious friend and receive my AI project tracker database!

or copy/paste this link: https://bensbites.beehiiv.com/subscribe?ref=PLACEHOLDER

⭐️ How did we do?

⭐️ REAL REVIEWS