Let's Dance...
The AI Wars, Pro AI Photoshoots, Cloud Hosting, Andrej Kaparthy and more!
👋 This is Deiniol - founder of AI Suite.
What. A. Year. Week.
You think it’s crazy now? It’s about to go completely bananas…
Here’s why 👁👁↓
📣 Highlights
Andrej Kaparthy, friend of Elon Musk and highly sought after talent in Silicon Valley is rejoining the already stellar rolecall at OpenAI. More on Adrej below.
Microsoft nailed it. Compared to BARD and Google’s presentation mishaps, The New Bing and the Edge Browser have made Google look weak and irrelevant. Relevance is everything. Especially in this red-hot market. That said, there are undoubtedly bigger moves still to come.
Google presentation - BARD's demo starts at ~13:18
Baidu, the search giant of China and an early adopter of AI integration have announced the impending release of their own ChatGPT equivalent, ERNIE (anyone else see that as a direct slap in the face to Google’s BERT?). Let’s remember, China has an immense amount of data and their Government is very proactive at providing support to AI startups. Their paper states Ernie has 260 Billion parameters, making GPT3 look inferior at ~3/4 the size.
RunwayML announces GEN-1: a video AI system that can generate video in any style.
Getty Images files to sue StabilityAI.
Google announces onboarding for their generative AI API will commence in March 2023. Great news for devs.
Cohere, a company that integrates large language models into businesses of any size, are in talks to raise funding at a +$6billion valuation.
There’s more although I’m assuming you don’t have time to digest absolutely everything. Time for dessert…
💬 Quote
💻🧑💻 Person
Andrej Karpathy
Andrej is a highly accomplished AI expert with a diverse background in both industry and academia. With experience in various technical and leadership roles, Andrej has made significant contributions to the field of AI. To put the icing on the cake, Andrej announced on Wednesday that he will be rejoining the OpenAI team.
These are just some of Andrej’s notable achievements:
Sr. Director of AI at Tesla:
Led the computer vision team of Tesla Autopilot
Oversaw in-house data labelling, neural network training, and deployment in production
Goal was to develop and deploy Full Self-Driving to the fleet of cars
Research Scientist and Founding Member at OpenAI
PhD focused on convolutional/recurrent neural networks:
Advisor was Fei-Fei Li at Stanford Vision Lab
Worked with Daphne Koller, Andrew Ng, Sebastian Thrun, and Vladlen Koltun
Designed and was the primary instructor for the first deep learning class at Stanford (CS 231n: Convolutional Neural Networks for Visual Recognition)
The class became one of the largest at Stanford
Enrollment grew from 150 in 2015 to 750 in 2017
Interned at Google Brain, Google Research, and DeepMind:
Worked on learning-scale unsupervised learning from videos, large-scale supervised learning on YouTube videos, and deep reinforcement learning
MSc at the University of British Columbia:
Worked with Michiel van de Panne on learning controllers for physically-simulated figures
BSc at the University of Toronto:
Double major in computer science and physics and a minor in math
First got into deep learning, attending Geoff Hinton's class and reading groups.
This is a huge win for OpenAI. It adds to the amazing pool of talent at an exceptional company. BIG things are on the horizon.
🤗 Know someone who would benefit from some of these resources? Fwd it on, reply back “sent” and I’ll send you an exclusive 50% discount on an annual Writesonic account (great for when ChatGPT is down!).
🎙Podcast
Dumb Money Live | This could destroy GoogleVery legitimate and real disclaimer - I am not a financial advisor nor do I pretend to be one. Do not take my endorsement of this podcast as financial advice.
As we’re well aware, Search is about to undergo a massive change.
Chris, Dave and Jordan go deep into the potential implications and hypothesise what may happen to Google and Microsoft in the very near future. This podcast was streamed live just before Microsoft’s presentation.
Some of the questions Dumb Money Live go into:
Can the old guard of the search engine industry keep up with the AI revolution without sacrificing their advertising revenue?
Will AI be the end of Google Search as we know it?
How will everyone monetise an AI chatbot in the short term?
And what percentage of people will actually click on search results from a chatbot?
Our verdict: 💰💰💰💰💰 out of 💰💰💰💰💰
Let’s dance.
⚙️ Tool 1
PhotoAI | by Pieter Levels (link below)
Say cheese! A new tool just hit the scene: PhotoAI. Created by the legendary indie hacker Pieter Levels.
PhotoAI lets you strike a pose in any location on earth, all with the help of AI. Just snap a few pics of yourself, train the model and voila! Professional-quality, highly realistic photos in no time at all (look at those fingers!).
Secondly, ever admire a photo but wish you were the star? Try the copycat feature - now you can step into your favourite snaps, no cloning required.
For a limited time, you get one free model training, which means 50 FREE photoshoot possibilities. Snap to it!
Pro tip: Close and medium shots work best for now when training your own model.
⚙️ Tool 2
Evoke
Have you ever found yourself in the middle of a project and thought, "Man, I wish hosting AI tools was as easy as ordering pizza"? Well, your wish has been granted! Cloud hosting your AI app is a piece of cake with Evoke.
Richard Gao, CEO of Evoke, recently chatted with me and let me in on some of their secret sauce.
Richard explained that the aim of Evoke is to provide a platform that simplifies the process of building AI tools for developers and businesses. The cloud hosting of open-source AI models means that developers can easily access the models they need for their projects, without having to worry about the technicalities of hosting them themselves. This saves time and resources, allowing developers to focus on building the best AI tools possible. Currently used by numerous companies, the API offered by Evoke streamlines the process and makes things super easy for developers.
Features:
Unlimited usage
Pay per use
Direct customer support via their Discord
Scale effortlessly
24/7 uptime, serverless for on-demand generations
Privacy and security ensured through access restrictions and limited permissions
AIs Evoke hosts:
Stable Diffusion (v1.5), All Stable Diffusion Models (coming soon), Dreambooth (coming soon), Other AI models (coming soon)
🧑🏫 Demo/Tutorial
Lenny's Newsletter | How to Build Your Own GPT-3 ChatbotThis is a great piece by two highly influential (and aspiring!) people, Dan Shipper and Lenny Rachitsky. The tutorial is highly detailed so rather than the usual breakdown, you can see it all here👇. Happy building.
🖼️ Images
@DiffusionPics | AI art of a real artist in her real art studio creating real art based on AI art 
📚 ELI10
HyperparameterA hyperparameter is a value that determines the overall design of a model or how an algorithm works. It's set before training and doesn't change during training. Different values are tested to find the best ones for the training process. For example, in a model like a building, the number of rooms is a hyperparameter (it doesn't change during building), while the furniture inside the rooms (like chairs and tables) are parameters (they can change during building).
I want you to influence the direction of this newsletter so please let me know if you have any feedback by replying to this email.
🏆 Sponsor Us
Would you be interested in sponsoring our next newsletter, which will be distributed to a diverse audience of AI enthusiasts, startup founders, aspiring entrepreneurs and artificial intelligence students?
Reply to this email and I’ll get back to you asap.