10X AI (Issue #35): Microsoft's AI Education, Helpful AI Videos, and "Pogo" Vaulting?
PLUS: Stability AI's duo of launches, TikTok's AI songs, AI in Adobe Premiere Pro, better Runway Motion Brush, Google search, and AI in consumer products.
Happy Sunday, friends!
Welcome back to 10X AI: a weekly look at generative AI news and developments affecting the average end user. Today I’m also test-driving the “AI Videos” segment with two handpicked recent videos.
Let’s get to it.
This post might get cut off in some email clients. Click here to read it online.
🗞️ AI news
Here are this week’s AI developments.
1. Microsoft brings AI to education (+Copilot Pro)
The potential of AI to help both educators and students is increasingly well-studied. (
and are must-follows in this space.)Now, Microsoft is stepping up its efforts to introduce AI into the Microsoft Education environment.
First, Copilot with commercial data protection is now available to faculty and higher education students. (In related news, Microsoft launched Copilot Pro across its entire product suite for $20 / month.)
Second, Microsoft is developing dedicated AI tools for education, starting with its free Reading Coach.
You can ask AI to create a story or select from a range of ready passages aimed at different grades and Lexile levels:
You record yourself reading, and the AI coach tells you how well you did.
I don’t mean to brag, but I absolutely crushed my reading of the “Cloudy and Sunny” kindergarten text:
Four-year-olds might draw better than I do, but I’ll demolish them at reading any day. Bring it on, toddlers!
2. Two small models from Stability AI
Stability AI released two new, small models:
Stable LM 2 1.6B: This tiny LLM beats other small models (e.g. Microsoft Phi-2) and even much larger ones (i.e. Falcon-40B-Instruct).
Source: Stability AI Stable Code 3B: Small code completion model trained on 18 programming languages, which outperforms every other similarly-sized coding LLM.
This fits with the general trend toward smaller, more specialized large language models.
3. Adobe adds AI to Premiere Pro
If you’re an Adobe Premiere Pro user, rejoice!
Adobe has just launched new workflows in Premiere Pro to make sound editing easier, including an AI-powered feature that automatically tags audio clips as dialogue, music, etc.
Here’s all the new stuff:
4. Runway Motion Brush upgrade
If you thought Runway’s Motion Brush was cool, you’ll think their new Multi Motion Brush is multi-cool at least!
The Multi Motion Brush lets you select separate areas of the screen to animate independently of one another. Like so:
Multi-cool indeed!
Runway is working hard to maintain its lead as the go-to, all-in-one AI video tool.
5. TikTok AI songs
Taking a page out of Suno’s playbook, TikTok is testing an experimental feature called “AI Song” that can create songs with lyrics out of a short text prompt.
Unlike Suno, the results are currently a mixed bag with out-of-tune singing and mediocre lyrics.
Then again, how can you even try to beat epic rhymes like this:
“Listen up, fam, let me lay it real neat
Why Try AI here, droppin’ knowledge so elite”
You can’t, that’s how!
6. Google’s smartphone search shenanigans
Google is working to make search “more natural and intuitive,” largely driven by AI. The company just announced two new features:
Circle to Search: Highlight, circle, or otherwise mark anything on your screen to instantly look it up, regardless of whether it’s a picture or text. Here, check out the demo from Google in the form of an upbeat video suspiciously reminiscent of Adobe Firefly’s launch:
Google Lens + AI: When using Google Lens or uploading a photo, you can now go beyond just “search for this image.” Supplement your search with a question for AI to answer. (One example from Google: snapping a picture of an unknown board game at a yard sale and asking “How do I play this?”)
Google teased more upcoming capabilities in this vein. The company is sticking to its primary “Helping people find stuff” mission while exploring new, AI-powered ways to do so.
7. AI in new consumer products
Lately, companies plugging “AI” into everything imaginable just to jump on the bandwagon is starting to remind me of this cartoon by Alex Krokus:
We’re moments away from AI-assisted toilet paper. (Please tell me that’s not an actual thing yet.)
But brands are gonna brand, so now we have:
Samsung Galaxy S24 Series boasts a whole range of built-in AI things:
Live Translate for two-way, real-time voice and text translations
Chat Assist for striking the right tone in your messages
Android Auto that summarizes incoming messages and suggests replies
Note Assist & Transcript Assist that summarize notes and recordings
Circle to Search in partnership with Google (see previous entry)
Some Amazon Fire TV devices will be able to create AI-generated images, powered by the company’s own Titan Image Generator. So you could e.g. get a unique TV background of a “fairy landscape” simply by asking:
So in case you didn’t already have enough AI in your life, there you go!
📽️ AI Videos
Here’s the promised “AI Videos” test segment where I handpick entertaining, useful, or otherwise noteworthy videos. This week, I’ve got two:
8. “Alpha Everywhere” by AI Explained
AI Explained is an excellent YouTube channel that often dives deep into research papers to extract the most interesting bits.
This week’s video covers two news items that aren’t typical “10X AI” material but are impactful enough to highlight:
Google DeepMind’s AlphaGeometry solves geometry problems at a near-human Olympiad gold-medalist level.
A new AlphaCodium code generation flow makes LLMs significantly better at solving coding problems.
9. “How David Perell Uses ChatGPT to Write for Millions” by Every
Dan Shipper of Every fame recently started an underrated interview series where he talks to real people about the way they use AI.
The latest installment features David Perell—someone I wasn’t familiar with up to this point—and his process of working with ChatGPT. David’s life philosophy might not be everyone’s cup of tea, but he shares practical use cases that should be broadly applicable:
🤦♂️ 10. AI fail of the week
Wow, Olympic pole vaulting isn’t at all how I remember it
-Zooming out a bit, these changes all seem pretty incremental and not revolutionary
Yea, I feel we’ve reached a kind of plateau. On one hand I’m glad we’re not in a yudkowskian doom loop. On the other hand, I think AI will have at least as much influence on our lives as the internet itself. From now on, I expect only incremental developments. But I look forward to reading about every single increment along the way. Cheers to the future!
Zooming out a bit, these changes all seem pretty incremental and not revolutionary (apart from maaaaaybe the "alpha everywhere" concept). It seems like we're due for a few months of consolidation and small improvements on the glut at attempted revolutionary software from last year.
Then, I'm reminded that "revolutionary" and "incremental" aren't mutually exclusive. A watched pot never boils. These minor changes can make an enormous difference over time.
If you buried your head in the sand in the middle of 2023 and came up for air just now to look around, I think you'd see revolutionary change since mid-2023, especially with image and video generation (and the LLM understanding how to do images).