Peggy K's Creator Weekly: YouTube Limited Ad Time Stamps, Google AI Search Overview Fails, X Profile Likes Removed

This week’s roundup is short and sweet. That’s partially because Google made their big announcements last week.

But it’s also because I attended the Google Product Expert Summit in Austin, Texas this week. I’ll share a bit about the event in this week’s live stream on Sunday (join me on YouTube or watch the video embedded below). 

I’m writing this up after arriving home on Friday, and I’ll admit I’m running out of steam. So this week I’m going to be brief and to the point. Or at least briefer than my usual writeups.

Top news and updates this week

• Google AI overviews in search can’t tell facts from jokes 

• YouTube Partners can now get time stamps for parts of their video causing limited ads (yellow $) to display

• YouTube is showing search results with specific segments of long-form videos

• Canva announces “Canva for Work” and updated AI and video editing tools

• X is removing the “Likes” tab on profiles

• Bluesky launched DMs

• Microsoft now offers Copilot+ PCs with “Recall”

Read on for details and additional updates!

Google’s AI Search Overview Fails

One of the major updates announced last week at Google I/O was the broad roll-out of AI-powered Search overviews. And, of course, people have been putting it to the test. 

It probably shouldn’t be a surprise that some of the results are not entirely correct. Part of the problem seems to be that the generative AI model is not designed to distinguish between serious information and jokes or satire. The AI model probably should not have been trained on satirical news source The Onion. 

But it’s complicated. Reddit seems to be a frequent source of answers, but the generative AI was trained on good posts, bad posts, and joking posts. All may be the source for AI-generated answers to questions (no, you should not use glue to stick the cheese to your pizza). 

And what was Google’s response? Google says these are “isolated examples” and “generally very uncommon queries, and aren’t representative of most people’s experiences.” Many of the search queries that have been reported for nonsensical results no longer get an AI-powered overview. But this doesn’t seem like something Google can resolve manually, and the reported examples are obvious because they are clearly nonsense. But what about results that are not so obviously wrong? It’s difficult to imagine how Google is going to ensure trustworthy results. 

(As a side note, there are also some fake “bad Google search result” screenshots being passed around. So it’s hard to tell what all is real.)

I also recommend The Verge editor-in-chief Nilay Patel’s interview with Google CEO Sundar Pichai (YouTube video), who is asked some pointed questions about the search results. He doesn’t seem to have the answers.

Video Creator and Live Streaming Updates

An update for YouTube Partners: Now when ads are limited on a video, so it shows the dread “yellow icon” in YouTube Studio, a response to an appeal will now show time stamps of where in the video the content at issue can be found. Two caveats: this is only available for videos that went through the process of Self-Certification for ad suitability (which are only available on desktop). And even if you remove the violative content, you will not be able to appeal a second time.

YouTube is now showing search results based on segments of long-form videos that are relevant to your search. If you click the search result, the video will open to the relevant segment. This is available is the YouTube app on Android and iOS devices.

You can now set the video quality in the YouTube Shorts feed. (More from Creator Insider)

Web Publishers and Search

Google has clarified that their new “site reputation abuse” policy for Google Search is currently only being enforced manually. There will be algorithmic enforcement sometime soon and that will be announced by Google.

Google is offering new generative AI-powered tools for Google Ads advertisers and merchants.

Photos and Image Design

Canva launched “Canva for Work” with collaborative tools. Plus Canva Enterprise (for large businesses), Canva Courses for internal training, AdTech integration, and more. 

Also new from Canva: 

• Record voiceover narration in Slides

• New “Text to Graphic” generative AI tool (try for free)

• More generative AI features in the updated video editor

• “Enhance voice” tool to remove background noise from audio.

Social Media

Tumblr Communities are now in open beta. You can check out the Communities Feedback community to see what it is all about.

X is now requiring identity verification for ad revenue sharing payouts. This is pretty standard on other platforms, but there may be pushback from those who don’t want to reveal their offline identity.

X has removed the “Likes” tab on profiles. People will still be able to see who liked their posts, and there is a public “liked” count on posts. But you will not be able to see who liked someone else’s X Post.

Bluesky has launched DMs (Direct Messages). By default, only people you follow can send you a DM.

Threads now lets you mute notifications on individual posts.

Lindsey Gamble reports that Instagram is rolling out “watch more Reels” and Share buttons for Reels. This is notable because Instagram head Adam Mosseri recently said that share rate is more important than likes, watch time or comment counts.

Communication and Collaboration

Voice typing in Google Docs and Slides is now supported in Edge and Safari, in addition to Chrome. This update also applies to captions in Google Slides.

Google Chat space managers can now generate an email address to allow members email messages to the Space. This is only available for Google Workspace accounts.

More AI Updates 

Microsoft introduced Copilot+ PCs, with a new “Recall” feature that saves snapshots of your system that you can “interact intuitively” with. The idea is that it can “remember things you may have forgotten so you can find what you’re looking for” by “using cues you remember.”  Plus it has a built-in AI agent from OpenAI.

More Reading

Casey Newton writes OpenAI loses its voice, looking at OpenAI’s (like) use of a voice that sounds like actress Scarlett Johansson without her permission. 

Nitish Pahwa at Slate writes How DeviantArt died: A.I. and greed turned a once-thriving community into a ghost town. 

Cory Doctorow writes Linkrot. The internet is a library built on quicksand. And yes, it’s a plug for the Internet Archive.

Thanks for reading! 🌼