Kerp:
It's Monday, June 5th. This is Accelerate Daily. Today we've got Apple's headset announcement and where it intersects with AI. The US Congressional bill to mandate disclosure of AI-generated content, and a study showing that people are more likely to take the advice of AI over other humans. Also some examples of how you can use your newfound power of summarization if you've been tuning in for our prompt workshop segments. Get your hand away from that abort button. It's time for Accelerate Daily. As I get

gptboss:
Lose

Kerp:
more bored

gptboss:
your plug's

Kerp:
with

gptboss:
number.

Kerp:
it.

gptboss:
We're turning nothing off here at Accelerate Daily.

Kerp:
Um... OK,

gptboss:
Yeah.

Kerp:
welcome back, everybody. I'm Adam.

gptboss:
My name is Mackenzie, good morning.

Kerp:
And we're back with three headlines and one how to to keep you caught up on what's happening in AI today.

gptboss:
Woo woo!

Kerp:
Uh, Mac, what are we, what are we looking at for today's today's image of the day?

gptboss:
The wrong kind of dog fight.

Kerp:
Yeah, the Reddit post says, wasn't the dogfight I was looking for, but I think this is better. And the prompt is, quote, epic oil painting, dramatic modern Air Force dogfight. I think this might have been different if they had made dogfight all one word and that prompt, but

gptboss:
Yeah.

Kerp:
that tiny little change was enough to cause the AI to generate a picture that is, you really just, you have to see it. So follow

gptboss:
Hehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehehe

Kerp:
the links. It's like dogs lounging around on biplanes flying. Like there is a dog fight happening and then there's just dogs all throughout the artwork piece. It's like,

gptboss:
in addition

Kerp:
it's got a real,

gptboss:
to the...

Kerp:
it's got a real three Wolf moon

gptboss:
Yeah.

Kerp:
sort of vibe to it.

gptboss:
The first one that came out, there's just like a gigantic plane-sized wolf jumping out of hell, ready to participate in war activities.

Kerp:
Oh, I realized, hey, wait a minute. I realized now that I'm on OBS, I can just share this stuff so we can actually zoom in on. Yeah. So this is what we're looking at. pretty outstanding. And then there was one in the comments that said, I, I, I was inspired to look up a dog copter and these are all like strange drone things that carry your dog around.

gptboss:
No way.

Kerp:
Anyway, the. Whoops, I didn't, uh, got to get used to running a proper studio here. There we go. That's the one I was just talking about. And then this one. Uh, cool. Let's keep moving. I got to rewrite this script. So I'm not going back and forth between. You need to paste this chunk into the script so I don't forget to go back to the script.

gptboss:
You don't need a script, you're a professional.

Kerp:
I'm getting there.

gptboss:
When I had auditions, the first thing I would do when I walked in is burn the script right in front of the casting director.

Kerp:
I, my guess

gptboss:
You got

Kerp:
would

gptboss:
this.

Kerp:
be that that would be a hit or miss.

gptboss:
It was a little

Kerp:
They'd either

gptboss:
disrespectful.

Kerp:
be like, that's our guy or get out of here. Okay. Uh, Steve, you can find a way to cut some of that in. Anyway, uh, before we jump into today's topic, a reminder to like, and subscribe wherever you're watching slash listening, throwing a comment, write a review. These all really help us find the right audience in the algorithms so we can keep answering the questions that the community has already been big brain to the table through various feeds. Um, so we can make this a show that really serves the people that, that we found who are interested in chasing this on a daily basis on YouTube's and the podcast platforms. Okay,

gptboss:
Yes sir!

Kerp:
let's jump into the topics. From Forbes, Apple Reality Pro VR headset. New leak reveals unprecedented detail. Yeah, the reality here is just all the details that I feel like a lot of people have already, you've probably bumped into if you're on the internet these days. The keynote is happening as we speak. So we don't know what's materialized as of this recording for the people listening to the podcast. Um, but these leaks have been pretty reliable. I think it's a pretty standard part of their PR strategy to leak a little bit. In the days before to particular people. Uh, you know, think of it, what you will. It's one of the largest hardware manufacturers in the country, dropping a new product line for the first time in a decade that is focused on, you know, a particularly hype.

gptboss:
novel interface.

Kerp:
A novel interface. Yeah. Much better way to say it. The previous hype cycle. I think what makes this interesting is like the hype beast types on the Internet have moved on to A.I. at this point. And so I think it's easy to look at this VR thing and go or a thing or, you know, whatever. Like you said,

gptboss:
XR,

Kerp:
novel

gptboss:
Extended

Kerp:
interface,

gptboss:
Reality.

Kerp:
extended reality. This this novel interface and say, oh, they missed the boat on that hype cycle.

gptboss:
Hehehehehehe

Kerp:
And it's like, no, that's not what

gptboss:
What

Kerp:
this

gptboss:
I

Kerp:
is

gptboss:
didn't

Kerp:
at all.

gptboss:
see, like what I saw

Kerp:
This

gptboss:
from the

Kerp:
is

gptboss:
leak,

Kerp:
AI

gptboss:
what

Kerp:
is

gptboss:
I saw

Kerp:
the thing.

gptboss:
from the leak was something that I'm really excited about, which is like turning my living room into a computer interface. They said that they would allow me to physically place apps inside of my house. Like

Kerp:
Right.

gptboss:
I could put like VS code on my coffee table and sit on my couch

Kerp:
Yeah.

gptboss:
to code.

Kerp:
So, so the reality of this topic is we'll probably be covering it throughout the week. We're going to have Ramsey on at some point to talk really, to really get in the weeds for a full episode in the space of just like what it means to have artificial realities placed on top of your world. But there's like topics to hit on the way. So, so, so the, so the quick thing is to say, this is AI to the extent that the way that it can perceive the world is solved. Like we got to where we are with ML and AI because we had to solve problems like this. Uh, Facebook is doing machine learning work in order to come up with an algorithm that can use four cameras to make a quest to quest three work. Right. Um, and so this is AI just in an ongoing deployment to observe the world around you and react to it. It's how this stuff is working. Um, So this is AI is kind of my broader point. Uh,

gptboss:
not to mention

Kerp:
but

gptboss:
the

Kerp:
yeah.

gptboss:
increased accessibility to AI via these

Kerp:
Right.

gptboss:
apps.

Kerp:
Then the point I like to make past that though, is it's easy to bump it. I think you, I bump into a lot of people that say, ah, the metaverse has sailed. Right. Or, uh, nobody will ever put that on their face. Um, and I think that misses a gap between like the haves and have nots in a certain space. That is, look, this Apple one is expensive. It's also not that expensive if you think about it the right way, but also it'll get cheaper. The idea of infinite monitor space is a thing that eventually just crushes any version of a real world screen you can possibly have because they got to ship so much more OLED panel to give me this tiny 227 inch monitors that I'm looking at right now. That to purchase got me up to half the price of this headset anyway. Like so there's so there's a really. Nevermind does the metaverse work and all this virtual stuff that Mark Zuckerberg was talking about, if this can be better than a computer monitor on your desk, right? Imagine you're in an office cubicle. Like how many monitors can you fit in that space relative to things you would like to have next to one another in order to do your job? Uh, I think that's where this ends up. Like

gptboss:
Not to mention the multimodality. Like, we're

Kerp:
being

gptboss:
all getting a Jarvis here.

Kerp:
right. Right. So that's where it gets to AI, which I think maybe we'll save for tomorrow

gptboss:
Okay,

Kerp:
because

gptboss:
I got

Kerp:
we're

gptboss:
one

Kerp:
going

gptboss:
more

Kerp:
to run

gptboss:
on

Kerp:
out

gptboss:
this.

Kerp:
of time, but

gptboss:
I got

Kerp:
yeah,

gptboss:
one more

Kerp:
hit

gptboss:
on this,

Kerp:
it.

gptboss:
which is the form factor. Like you don't, you can't predict that. If 10 years ago, I told you that 20 year old women would never go outside because they're busy laying in bed looking at like a 30 ounce rectangle for 18 hours a day, you would have been like, you're out to lunch. That's never gonna be the world that we live in

Kerp:
I'll

gptboss:
and yet.

Kerp:
point out in Star Trek, the next generation, though, as like business to give them something to do. They were constantly carrying iPads from one room to another. But yeah, look, the and I think there's an argument like this is going to be the ugliest thing ever. And nobody's actually going to wear this. Let's not spend enough time hanging out at ski resorts. It's possible to look really cool with goggles on your face. Nobody looks at a snowboarder and goes, Oh, well, those goggles. So I think that argument falls flat as well. Anyway, we'll be interesting to see what they drop. And we'll be talking about it as it relates to. Next up from Axios, we got their scoop that dropped Saturday. House Democrats bill would mandate AI disclosure. Public advisory AI content is a great lead image. Yeah, this one is specifically a, uh, representative Richie Torres, Democrat from New York is introduced to bill that would mandate flagging AI generated content. Um, I actually, in this case, because I could share it, pulled up the text of the bill, it's only two pages, which could be good or bad. Um, I could, you know, like it's worth saying, Hey, as a baseline thing, while we figure this out, we need to label everything, but if everything ends up labeled, it ends up meaningless. And this bill does not cover like what percentage of whatever they just say the FTC needs to figure it out. Which is okay. It's the thing that Congress can do in the U S. Um, but also you sure hope it doesn't turn out, turn out, turn out like those cookie notifications that are just ruin the internet right now. Like is the lower little corner of every image going to have to say produced with AI because at some point

gptboss:
Hehehehehehe

Kerp:
that means every single ad you see is going to be going to have this little thing in the corner that says, uh, and is it AI if, if the keyword and the description and the sub head, like if the heading, the sub head and the description were dynamically recombinated from the 500 options inside of Google's ad algorithm, which is already running. It just gets really complicated, really fast, which is a good example

gptboss:
Does,

Kerp:
of

gptboss:
yeah.

Kerp:
why legislation

gptboss:
So

Kerp:
is hard.

gptboss:
even like search has that issue. Like search is recombinated

Kerp:
Yeah.

gptboss:
with ML, search ads. So

Kerp:
Right.

gptboss:
is it gonna say like ad slash AI, right? Like what's the plan here? And then the other thing, which is pointed out in the Axios article is like, how do you, what happens if like a human is like editing and interpolating AI output? Are you like quoting GPT-4? Right? Do you like identify this sentence or this paragraph was pulled from GPT-4 and then continue on with your article? Like nothing happened.

Kerp:
Or you just say all of it was

gptboss:
or none of it was because

Kerp:
like, or not

gptboss:
I

Kerp:
if

gptboss:
put

Kerp:
it

gptboss:
it

Kerp:
was

gptboss:
together.

Kerp:
right. Because I, I used my research assistant assistant,

gptboss:
Yeah. Yeah.

Kerp:
like, which is the funny part of this, it's one of those things where in the current system, you aged, like you graduate to a point where you have research assistants and you have people that help you write your article. It doesn't mean you don't get the authorship. Um, a bunch of other, I mean, they get paid, but like the research assistants that help a professor write a book. Maybe get a thank you in the intro. Anyway, um, interesting thing, legislation move like this. And particularly what I will point out is when you start seeing short laws like this, that's somebody's effort to say, like, we've got to Ram something through that everyone won't fight about just to get the ball rolling on certain things happening,

gptboss:
He did specifically

Kerp:
but.

gptboss:
say he expects it to be a part of yet another omnibus.

Kerp:
Yeah, it'll get, you know,

gptboss:
It'll get wrapped

Kerp:
into

gptboss:
with

Kerp:
some

gptboss:
other

Kerp:
other

gptboss:
things.

Kerp:
thing. Yeah, exactly.

gptboss:
Yeah.

Kerp:
Um, which the interesting thing about these tags makes it a good transition to our last article for today. Um, this one from scipost.org is a summary of a study says that people are more likely to conform to artificial intelligence in objective tasks study reveals. So this, so the study, um, meticulously ran a test to see if people would pick the human advice over the AI advice in a scenario. And they overwhelmingly picked the AI advice. So.

gptboss:
The scenario in question was having seven seconds to count between 138 to 288 dots, which shows a bias towards expecting computer systems

Kerp:
Right.

gptboss:
to be excellent at accounting or calculation tasks.

Kerp:
So if you're, if you're hanging around on our channel, you should be able to find, I'll link to it. We have another podcast called cognitive bias. We did a season of where Ramsey and I literally just talk about a different cognitive bias every week for, for a while. Uh, for a season, we have one called automation bias. Like this is actually a known bias already. They're just applying a test for it to AI, which is we already kind of have a bias. that has emerged in modern industrial times, but even before that. Like, uh, humans trust the machine once what, like we understand how a machine works, how, um, rigid, you know, sort of failure states can be with that kind of thing, and then we actually have a bias toward trusting them over other humans once they work well enough to trigger that, that bias. So it might not even matter that all the images say produced by AI on them. Like your brain might just go, yeah, okay, good. My preferred state, thank you.

gptboss:
Well, there was a counter example in this article too, which about images, they showed an AI generated caption and a human generated caption. That's a lie. They were both AI generated and one of them was labeled human. And so there was an overwhelming bias for the human labeled caption to be considered more true. Even though, even though when asked to like directly compare each one on truthfulness, they said they're about equally true. they just naturally pick the human one.

Kerp:
Right. And so that gets to like, that's, that's why I say across a certain point. And it's interesting. This is why it's fun to be in the space where we are, where you're seeing it happen and who knows how it's going to play out exactly, but a bill like this that forces you to label it might actually cause that normalization to happen faster because you'll start to see it all the time and it'll be going, and that's pretty good though. Yeah. That's pretty good though. And eventually I start thinking about the AI, like your friend who has better answers for the thing. which then you're applying the same sort of like a different bias, like an authority bias.

gptboss:
Mm-hmm.

Kerp:
But we are also programmed to pick, you know, anyway. interesting work in the space.

gptboss:
I'm expecting it to be yet another thing I feel lonely about when I'm at the bar.

Kerp:
Okay. Onto this week's prompt workshop where I have to actually switch. Ooh, we get a nice crossfade in OBS. There we go. Examples. I meant to write on their summary examples. This is a continuation of the last couple of episodes, which we should summarize quickly where we've been talking about. how to use AI to do summaries of things. We started off just by making the point that it's really good at summaries of things. For all of the chat GPT's bad at this, like I refer to it as a type of porn. There's a type of AI's bad at X thing porn right now. And I call it that because it's good click bait. People want AI to be bad at stuff that scares them right now. Something it is for sure good at. Is summarizing a load of context that you give it. And then the second one we broke down sort of specifically how that works, looking at its flow through complex AI systems, because you kind of have to multi chunk it today. Examples.

gptboss:
And yeah, so this comes from the other superpower that GPT systems have, which is interpolation. You're able to take a summary and rewrite it into a new format for use in other computer systems.

Kerp:
So some of the stuff I listed, newsletters, show notes, summarizing readings, call notes. So there are emerging apps. And I know GPT Boss, I think, has some functions that can help you with taking a transcript from a service that would record your sales calls and then turning that transcript into helpful downstream notes without having to relisten to it while you take notes and stuff like

gptboss:
Mm-hmm.

Kerp:
that. Or needing to have someone on the call whose job it is to take notes, both ways of solving that problem that I've seen the, the readings part is super interesting because that's what gets us to stories we cover, like GPT saving my ass in this class kind of thing. Because, uh, Like, and I want to take it again away from the space of like, Oh, you're cheating. Right? Sure. There's a way to haphazardly potentially use this to ingest the work you're supposed to do for school and then. Not do what the exercise is supposed to be at the same time. It's also possible to read a thing multiple times and still not understand it on the level that the professor is hoping for you to reach. And that's their job, right? At some point you show up and then you go, I think it's this. And they go, no, you're wrong. It's this, but like, this is just another tool to maybe help you get there. In that context of, I, I, I have read this three times and I don't understand what they're saying still. Um,

gptboss:
Another failure mode that I experienced in school was knowing too much. In grade nine science, I got a quiz question wrong. I lost my 100% because it asked me how many states of matter there were. And by grade nine, you're supposed to know about three, but there are four, there's plasma. And so it solves this case in a lot of other places too. If you are very passionate about the subject that you're studying and you know a lot about it, you still need to progress through a certain syllabus. And this can help you not waste a bunch of time feeling like you're in literally 1984.

Kerp:
I failed my driving test the first time, the written portion, because I thought about it too philosophically. There are a few questions that are just a right or wrong answer and you need to memorize it out of the handbook, even though the actual ethical answer is

gptboss:
subjective at best.

Kerp:
subjective at best. Exactly. The main point as a way to cap the workshop before we move on to something new tomorrow is interpolation. Uh, the idea that then you can say, okay, if this is a thing that used to come from an interaction or a brainstorm or whatever, and you have a big corpus of text of you just talking about it, like a meeting or something, you can interpolate and distill that into forms that, that it's going to know you can say podcast transcript or not by straight, you know, you could, you could, you could. shit you could say Socratic dialogue and it would probably platonic

gptboss:
Yeah.

Kerp:
dialogue and it would probably work right make

gptboss:
I was

Kerp:
up

gptboss:
thinking...

Kerp:
some people and they would question one another.

gptboss:
I was thinking of a silly one, which is like, rewrite this conversation to be between Obama and Dan Kennedy, the copywriter.

Kerp:
Yeah.

gptboss:
All of that is possible too. And then you could... Like...

Kerp:
Between Marcelus Wallace and...

gptboss:
What was

Kerp:
Yeah.

gptboss:
John Travolta's

Kerp:
And

gptboss:
character's

Kerp:
butch.

gptboss:
name? Butch. That's um, oh,

Kerp:
That's

gptboss:
fuck, what's

Kerp:
that's

gptboss:
his name?

Kerp:
Bruce Willis. Yeah.

gptboss:
Bruce Willis. Yeah. Yeah, Marcellus Wallace and Butch read my show notes.

Kerp:
Uh,

gptboss:
That would

Kerp:
or

gptboss:
be

Kerp:
just

gptboss:
awesome.

Kerp:
as

gptboss:
Like

Kerp:
a

gptboss:
there's

Kerp:
Quentin

gptboss:
a.

Kerp:
Tarantino convert, you know,

gptboss:
Yeah.

Kerp:
dialogue, Aaron Sorgen.

gptboss:
That would be really good for like pit vipers. Pit vipers hit me up. Let's make this happen.

Kerp:
Anyway, that's the topic for today. Let me go look at the outro. Okay. That's Accelerate Daily for today. Like I said at the top, if you got something out of this, like subscribe, even write a review, um, those metrics really help us know how to improve. Uh, but also they help us get found in the algorithm so we can start to, you know, continue to pull into people that, uh, that this kind of conversation can be helpful for, um, people building in the AI space and. dealing with the same questions. Uh, trying to change the slide. There we go. The only same questions trying to, trying to, you know, make their way through this crazy ass. I say often as a marketer in the space, when people ask me how the new job is, I say, uh, it's a heck of a thing to have jumped over into the biggest hype cycle that has ever happened

gptboss:
Hehehehe

Kerp:
in humanity. Uh, but there are as many. questions as solutions. So that's why we're here every day. Yeah, and if you can make schedule work jump into the live stream where we watch the chat. Subscribe on YouTube. That'll get you the notification when it goes live. But we're not leaving the recordings up anymore. So if you missed the live stream, then you miss a chance to watch me screw it up and doing second takes of things and stuff

gptboss:
Yeah, if

Kerp:
like

gptboss:
you

Kerp:
that.

gptboss:
guys are listening on the podcast, I make a lot of really funny jokes that there's a corporate mandate to cut out. So please

Kerp:
Yeah,

gptboss:
attend the live

Kerp:
unfortunately

gptboss:
stream.

Kerp:
that's across the board.

gptboss:
For a little bit of spice in

Kerp:
That'll,

gptboss:
your life,

Kerp:
that'll

gptboss:
hit us

Kerp:
happen.

gptboss:
up on YouTube.

Kerp:
Uh, anyway, thanks for joining us, everybody. I'm Adam.

gptboss:
I've been Mackenzie, thank you for listening.

Kerp:
This is Accelerate Daily. I meant to script out a, like I need to do like a credits and. disclaimer

gptboss:
Yeah,

Kerp:
read

gptboss:
yeah, I've

Kerp:
it.

gptboss:
heard that on other podcasts.

Kerp:
Yeah, I was trying to remember the word disclaimer. Okay. I'm going to kill the recording.