This week was absolute insanity for AI news. Just when I thought things were slowing down, it seemed every major AI company launched new products. And even some unexpected companies like Wendy’s! Let’s take a look at all the news!
Google IO Announcements - All About AI
Google came out strong with a series of AI announcements during their annual Google IO conference. It’s clear that Sundar Pichai, CEO of Google, is transforming Google into an AI-first company. Here are the major announcements:
Bard Upgrades - Dark mode, support for 20+ programming languages, Bard moves to PaLM 2, Bard is available to everyone
Gmail gets the “help me write” feature, and Docs, Sheets, and Slides all get AI features
Google Search will summarize results in quick snippets
Partnership with Adobe Firefly - create incredible generative AI art directly from Bard, powered by Adobe Firefly
MusicLM goes open to the public
Google Maps gets an “immersive mode”
PaLM 2 Announced
Adobe and Google partner on bringing AR to life
Wendy’s x Google Partnership
In another piece of huge news for Google, they’ve partnered with Wendy’s to bring AI to the drive-thru. Trained on Wendy’s food terminology, this LLM will take orders when customers in Wendy’s test location in Ohio go through their drive-thru. Read more about it here. This is yet another example of GPTs starting to change industries.
HuggingFace Launches Transformers Agent
Transformers Agent is an application of the HuggingGPT research paper HuggingFace put out a few weeks ago. It can also be described as a competitor to LangChain. Transformers Agent can leverage thousands of other LLMs and ML models that have been fine-tuned on specific tasks and “programmed” by talking to it in natural language.
Anthropic Launches a 100k context model!
Claude, by Anthropic, just released a new model that takes context sizes to another level, 100k tokens! This is a huge achievement, and the race for context size continues. Right now, you can apply to use Claude or use it directly in Slack immediately. It’s comparable to GPT4, which is saying a lot! I’ll probably make a video comparing it to ChatGPT 4 and some local models.
Meta Releases Their Multi-modal Modal
Meta has been on an absolute tear lately with AI. This week they open-sourced ImageBind: The new ImageBind model combines text, audio, visual, movement, thermal, and depth data. Right now, it’s only a research project, but the fact that they are leaning into open-source is appreciated.
StabilityAI Launches Stable Animation
StabilityAI, the company behind StableDiffusion, launched a new text-to-video product competing directly with Gen2. And it’s open-source, which I appreciate very much. Here’s a video demo of it:
How to reach me:
Announcements:
I launched a Discord server where we talk about everything AI. Come join us. It’s completely free and friendly to all skill levels!
I also launched a Patreon, where you can help support all the work I put into making as much AI content as possible. As a Patron, you will get access to exclusive channels in Discord, influence over future video topics, and Q&A sessions with me.
New Videos From This Week:
If you have any feedback or suggestions for topics, I’m always open to feedback! Let me know here, on Twitter, Discord, or anywhere else. I love hearing from you.
See you next week!
Matthew Berman
