Perplexity AI is BS’ing Everyone, Which Mac is Best For Me? visionOS 2.0 is Worth the Beta
HTML-код
- Опубликовано: 6 июл 2024
- We explore WIRED's in-depth critique of Perplexity, from ignoring websites’ request not to scrape content and making stuff up entirely, Apple's new on-device and server foundation models were training on…everything? And a sneak peek into Apple's visionOS 2.0 and iOS 18 with new discoveries!
Try Setapp Today!
Get incredible apps like Downie, CleanShotX, Ulysses, Paste, and more all with one monthly subscription of $10
setapp.sjv.io/c/5517111/34451...
Support the show
Join our member community and get an ad-free version of the show, plus exclusive bonus episodes every week! Subscribe directly in Apple Podcasts or here: primarytech.memberful.com/join
Reach out:
@stephenrobles on Threads
www.threads.net/@stephenrobles
@stephenrobles on X
/ stephenrobles
Stephen on Mastodon
mastodon.social/@stephenrobles
@jasonaten on Threads
www.threads.net/@jasonaten
@JasonAten on X
/ jasonaten
Jason on Mastodon
mastodon.world/@jasonaten
Chapters ➡
00:00 Introduction
04:46 App Shout Out
07:19 Perplexity is BS'ing Everyone
23:00 Apple Intelligence LLM Training Data
36:20 Try Setapp Today!
37:40 Apple Pay Later is Done
40:58 Humane Ai Pin Troubles
42:16 Apple's Help Me Choose a Mac Website
47:50 Apple Vision Pro visionOS 2
56:18 iOS 18 Updates to HomeKit, Passwords App, and Shortcuts
01:04:42 Eddy Cue Lore
01:11:54 Exploring Genmoji
01:14:30 The Great Mouse Debate
Links from the show: 
Наука
What?! I can specify my preferred Apple HomeKit hub!? 🏁
I am enthused and liken it to the iPad FaceTime camera being on the horizontal side for ease of use. 🙌🏻
Yes! Huge win
Hah! I got Mac mini too. I couldn’t get it to show me Studio either.
Haha gotta say work and not hobby I guess 🤷🏼♂️
The audio got better. But now for some reason at 35:17 the audio of Jason sounds weird on my end, not sure if this is for everyone else.
Glad to hear it got better overall! Did Jason’s audio issue persist after that time?
It happened one more time at about 1:06:50, but they were both short, seconds only. It was crackling slightly, but didn’t make the audio unrecognizable. Another great episode, I look forward to them every week.
I can confirm I also heard this audio artifacts at those specified times. Could be something with audio compression on RUclips's video? Anyone who listens audio only podcast want to confirm?
Is it true that Humane fired the person that gave MKBHD the review unit
Hadn’t heard that!
26:00 Jason gonna regret these comments. I guess good on him trying to see the unpopular side, but sometimes trying to do the both side-ism is the incorrect way to go about it. Especially when the AI space is wholly financially motivated. There is no moral high ground these companies will stick to if it will cost em a buck
It's not both side-ism... it's nuance. Also, there's a difference between using information to train an on-device model that enhances Siri, vs indexing the web so you can regurgitate back information someone else has created while abstracting them out of the value chain - which is the point I made a few minutes later.
@jasonraten so I'm hopefully not assuming but just to summarize as a jumping off point - your okay with the scraping of others info to train a model. The other stuff less as fond right? OK so let's start there.
When I analyze things I tend to try simplifying the scenarios down to its base and understand the moral quandaries that arise at that level, get an understanding and then start lumping in complexity because you are right nuance and complexity can change the situation.
So in this case let's say a small mom and pop company wanted to create a siri like app but obviously they don't have the money for a full blown LLM training program so they instead think hey, that Jason guy on that pod is a smart dude, let's just train off of all his vids and hope for the best. I think we would agree that would be plagiarism. They are taking all of your publicly posted content, but it is icky
This is where we might disagree, but doing this same exercise to everyone's data now without their permission I think is just that first sin but done a gazillion times more. I don't think the complexity changes the narrative other than to obfuscate from whom the particular stealing was done. Again you may not agree, but it's a problematic concern that trillion dollar companies just say Oopsy it's the best we can do.