Everything I learned about AI Gen just got blown out of the water
March mAIdness 2023, is a month when everything changed as far as text AI Generation goes. Everything I knew to comment about even last month looks and feels old and redundant.
Honestly, I’m afraid to write this article for fear it will seem tired in a matter of days. Do I have this right? Microsoft just gave Google a run for its money? People are vying to get their hands on Bing Copilot (Bing!?) and Google’s Bard seems like it was pushed out before it was ready to compete.
Let’s break down THE most dizzying 10 days in March that would appear to make tech history if not in the top tech tipping points in a lifetime. Thanks to Matt Wolf’s YouTube channel at least I can hold on to something to catch my breath before the next round of announcements.
March 13th - 17th
AI's Most Insane Week - Things Will Never Be The Same (14min)
On Monday, Stanford introduced the Alpaca 7B model, which is much more lightweight than similar models and can be used on a local computer.
On Tuesday, Google announced AI functionality for its workspace tools
as well as the release of the Palm API to select developers.
That same day, GPT-4 was released
… and Microsoft confirmed that it had been using an early version of the model in Bing for the last five weeks.
Wednesday saw the launch of Midjourney version 5, which can generate more realistic images.
On Thursday, Microsoft announced its 365 co-pilot and business chat.
Microsoft's AI Future of Work Event in 8 Minutes (8min version of the full one below)
The Future of Work With AI - Microsoft March 2023 Event (36min, but worth every sizzle reel moment)
In addition, Baidu released its chatbot, Ernie, but the presentation was underwhelming.
More significant announcements are expected at next week's Nvidia GTC event.
March 20th - 24th
Another Massive Week in AI! (Summed Up in 10-Min)
A Massive Upgrade To ChatGPT! (This is Crazy) (15min)
On Tuesday Google opened the Bard waitlist, giving quick access to those who signed up.
Also on Tuesday, Nvidia CEO Jensen Huang kicked off the NVIDIA GPU Technology Conference (GTC) with his keynote presentation unveiling the company's latest AI innovations
a new GPU, H100 NVL, for large-language-model inference, which can reduce processing costs by up to 10 times.
Huang also unveiled NVIDIA DGX Cloud, a service that will bring NVIDIA DGX AI supercomputers to every company
… as well as NVIDIA AI Foundations, a cloud service for customers needing to build, refine and operate custom LLMs and generative AI.
this new technology will allow businesses to create their own large language models using Nvidia's cloud computers and GPUs.
This means that companies can train their own AI chatbots and develop their own models without needing a supercomputer, as they can run them on Nvidia's computers.
NVIDIA is also partnering with Microsoft to bring NVIDIA Omniverse Cloud, a fully managed cloud service, to hundreds of millions of Microsoft 365 and Azure users.
On Tuesday, Adobe demoed Firefly which brings text-to-image generation to Photoshop and is currently in the beta stage. Adobe is treading carefully in this space as far as copyright infringement and is using the beta process to engage with the creative community and customers.
Also on Tuesday, Microsoft released text-to-image generation within Bing chat using DALLE, leaving people wondering if it's the next-gen version of DALLE. This feature is accessible to anyone inside of Bing chat and can generate images by simply typing a description.
As of Wednesday the Opera browser has integrated AI features such as ChatGPT and ChatSonic that allows users to summarize, rewrite and change the tone of voice for a selection of output text.
Also on Wednesday Microsoft announces Microsoft Loop, an AI-powered tool that is similar to Notion, with AI features built-in. Loop will integrate with other Microsoft tools like Excel, Word, and PowerPoint.
On Wednesday Canva rolled a ton of AI into their platform like a new Brand Hub, which provides tools to help users remain consistent with their organizations' visual identity. The Canva Visual Worksuite will also include an AI-powered "Magic Design" tool, a copywriting assistant, a translation feature, and a tool that generates entire branded presentations - basically all the AI things.
GitHub announced Copilot X on Wednesday, which is an AI-powered code helper that uses GPT-4 to help coders solve their problems and fix their code. It even has a voice interface to make it easier for coders to communicate with it.
Also on Wednesday, Ubisoft uses AI to auto-generate dialogue for non-playable game characters, speeding up game development process.
Now to Thursday, when Unreal Engine demoed the new version of metahuman at GDC, which enables realistic character creation from iPhone videos.
On Thursday OpenAI announced support for plugins in Chat GPT, enabling users to extend the functionality of the language model by adding additional tools like browsing the internet, calculating complex calculations, editing images and videos, and connecting to other tools like Zapier.
OpenAI has also open-sourced the code for the knowledge base retrieval plugin, allowing others to create their own plugins for their knowledge bases. This move could pose a challenge for other GPT-3-based companies that lack web connectivity and highlights Chat GPT's accessibility and versatility.