Transcript
00:00
Play from here
Copy citation
Welcome to AI at Work.
00:06
Play from here
Copy citation
I'm Elijah Saas.
00:07
Play from here
Copy citation
I'm a serial entrepreneur with 28 years of experience in tech, and today I run an agency
00:12
Play from here
Copy citation
that builds AI solutions.
00:13
Play from here
Copy citation
Hey, I'm Kevin Williams.
00:15
Play from here
Copy citation
I spend most of my time helping companies figure out how to actually use AI.
00:20
Play from here
Copy citation
Not in theory, but in the work itself.
00:24
Play from here
Copy citation
Hey, welcome to the show.
00:28
Play from here
Copy citation
Doing something a little bit different today.
00:31
Play from here
Copy citation
My buddy, Kevin Williams, is off doing some fancy consulting stuff for Utah's government
00:38
Play from here
Copy citation
AI and government regulatory, actually for a couple of days, because he was also supposed
00:44
Play from here
Copy citation
to speak at the University of Utah's MBA program for their MBA hub day on the future of work
00:50
Play from here
Copy citation
in AI, and he got pulled into these other meetings.
00:54
Play from here
Copy citation
So I went in there and did that for him, and it was really interesting.
00:57
Play from here
Copy citation
And I thought I'd share with you guys today what I shared with this MBA class, which is
01:01
Play from here
Copy citation
really tricky because people go back and get their MBA for a variety of reasons.
01:05
Play from here
Copy citation
Sometimes it's just continued education right after their undergrad degree.
01:09
Play from here
Copy citation
Other times it might be somebody who is already climbing that corporate ladder and wants to
01:15
Play from here
Copy citation
get another professional degree in order to get a salary bump, or it might be somebody
01:19
Play from here
Copy citation
looking to start a business for the first time or get additional knowledge for a business
01:23
Play from here
Copy citation
they've already started.
01:25
Play from here
Copy citation
So when it comes to AI, obviously that is changing the landscape for all of those different
01:30
Play from here
Copy citation
scenarios.
01:31
Play from here
Copy citation
And it's also really difficult to think about what do you tell a class full of people when
01:36
Play from here
Copy citation
you have no idea whatsoever what their experience has been with AI.
01:40
Play from here
Copy citation
So I had to start with the assumption that it was very little.
01:44
Play from here
Copy citation
And I didn't want it to be like that.
01:46
Play from here
Copy citation
But when I asked a bunch of questions throughout the different slides of who's heard of this
01:51
Play from here
Copy citation
or who's tried this or who's using this, it was maybe one or two people every time.
01:57
Play from here
Copy citation
And of course, the same one or two people every time that were actually using AI.
02:03
Play from here
Copy citation
A little bit shocking, I think, that we all get inside of our own little universe where
02:09
Play from here
Copy citation
we feel like because we're so immersed or hands-on with something, most of the world
02:14
Play from here
Copy citation
has to be doing the same.
02:16
Play from here
Copy citation
Just not always the case, especially with something that just seems to be everywhere
02:20
Play from here
Copy citation
all the time.
02:21
Play from here
Copy citation
So I'm going to call out what is on these slides because I hate it when some sort of
02:27
Play from here
Copy citation
voice medium starts calling out a bunch of visuals and you have no idea what's going
02:30
Play from here
Copy citation
on, where it's relevant, and maybe this will be the kick in the pants that I need to start
02:37
Play from here
Copy citation
doing stuff on YouTube so I can share this presentation there as well.
02:40
Play from here
Copy citation
So really fast, the things that we covered in this little talk were just my quick background.
02:47
Play from here
Copy citation
I think you guys may or may not know that already.
02:50
Play from here
Copy citation
How we got here with this whole AI business, where we are now as far as what the Frontier
02:56
Play from here
Copy citation
Labs are doing and all the things in the Frontier Labs orbit, why there's still skepticism around
03:03
Play from here
Copy citation
the impact this is having, how we practically apply some of these tools, and a little glimpse
03:08
Play from here
Copy citation
into the road ahead.
03:10
Play from here
Copy citation
So me, I've spent almost three decades in tech, which is crazy, and I got a late start,
03:16
Play from here
Copy citation
and that makes me feel super, super old every time I reflect on that.
03:19
Play from here
Copy citation
I started and sold a few businesses, and my current endeavor is an AI-led design and
03:27
Play from here
Copy citation
development agency that actually builds out these complete solutions.
03:31
Play from here
Copy citation
My whole tech journey started in robotics.
03:35
Play from here
Copy citation
So when you think about AI and automation and the future of all the humanoid robots
03:41
Play from here
Copy citation
to come and everything else, I got a really early look at that.
03:45
Play from here
Copy citation
And this slide is Terminator, and I go on to talk about how the beginning of my journey
03:53
Play from here
Copy citation
was working at Cyberdyne on the T-800.
03:57
Play from here
Copy citation
And then I say I'm joking, of course, and there's the little movie clip of the Terminators
04:03
Play from here
Copy citation
fighting people and things blowing up and robots with guns.
04:07
Play from here
Copy citation
So am I just really that old?
04:09
Play from here
Copy citation
Is really, I thought Terminator, everybody's seen Terminator, gets the reference.
04:13
Play from here
Copy citation
A lot of just shocked looks as if my previous work experience was at some
04:18
Play from here
Copy citation
clandestine military tech endeavor.
04:21
Play from here
Copy citation
It wasn't.
04:22
Play from here
Copy citation
It wasn't.
04:22
Play from here
Copy citation
Just boring old robots that usually did mechanical assembly and pick and place and making circuit
04:28
Play from here
Copy citation
boards and everything else.
04:30
Play from here
Copy citation
So we built a lot of Teslas, not very many Terminators, though.
04:34
Play from here
Copy citation
And I've got a little clip of a tornado here, because every time you talk about AI,
04:41
Play from here
Copy citation
it's like trying to report the weather during a tornado.
04:45
Play from here
Copy citation
You try to get a beat on the conditions and then all of a sudden the roof is gone and
04:49
Play from here
Copy citation
cows are airborne.
04:50
Play from here
Copy citation
In other words, things are moving really, really quick.
04:53
Play from here
Copy citation
So I always throw in the caveat that anything that you hear about AI today might be irrelevant
05:01
Play from here
Copy citation
in a month, a week, a year.
05:04
Play from here
Copy citation
It's one of those rates of change that we just can't really predict right now.
05:08
Play from here
Copy citation
But we can look back.
05:10
Play from here
Copy citation
We can look back.
05:11
Play from here
Copy citation
And we can say that AI has been an evolution.
05:14
Play from here
Copy citation
So there's this thing that feels like it was just all over the place overnight.
05:19
Play from here
Copy citation
All of these companies with the crazy valuations and people cheating their way through college
05:26
Play from here
Copy citation
or using it as leverage at work or getting replaced at work with their jobs.
05:31
Play from here
Copy citation
But the truth is that this has been something percolating in the background for decades
05:36
Play from here
Copy citation
and decades now.
05:38
Play from here
Copy citation
So if we go back to the 80s, we had this rule-based AI, right?
05:43
Play from here
Copy citation
Depending on what you want to give AI the label to.
05:46
Play from here
Copy citation
Very if-x-then-why logic.
05:49
Play from here
Copy citation
I think your thermostats or spam filters, stuff like that.
05:54
Play from here
Copy citation
It's all rule-based, right?
05:56
Play from here
Copy citation
Then we got into the 2000s and this was starting to take labeled data and run it through some
06:04
Play from here
Copy citation
algorithms and give you these smart outputs.
06:08
Play from here
Copy citation
You can think about how Netflix recommends movies to you because you watched K-pop,
06:12
Play from here
Copy citation
Demon Hunters.
06:13
Play from here
Copy citation
You might enjoy Avatar, The Last Airbender, right?
06:16
Play from here
Copy citation
It's using labeling data in your previous viewing history to give you those recommendations.
06:22
Play from here
Copy citation
2020s, we started getting into deep learning and this was taking a lot of the same label-based
06:29
Play from here
Copy citation
rules but applying it to massive unstructured data.
06:33
Play from here
Copy citation
This changed the game.
06:35
Play from here
Copy citation
This started giving us things like deep learning, processing, audio, images, video.
06:42
Play from here
Copy citation
We started to get facial recognition, self-driving technology, speech recognition.
06:47
Play from here
Copy citation
All of these things were inspired by the architecture of the human brain, why it's
06:51
Play from here
Copy citation
called neural networks, and then that led us into today where we have generative AI.
06:57
Play from here
Copy citation
So this isn't just aggregating data, it's producing new assets, right?
07:01
Play from here
Copy citation
We get images and we get videos and we get graphics, some better than others, but we
07:10
Play from here
Copy citation
are producing things, we are generating things, hence generative AI that didn't exist when
07:17
Play from here
Copy citation
we put in the initial prompt.
07:18
Play from here
Copy citation
So how does that play into our everyday life, even if you're not involved in all this?
07:22
Play from here
Copy citation
Well, it's kind of like the microprocessor in the early 80s, right?
07:27
Play from here
Copy citation
They're embedded into almost everything that we use but nobody really knew they were there
07:31
Play from here
Copy citation
or how they worked, and AI is starting to creep up like that as well.
07:37
Play from here
Copy citation
Smart homes, even if you have one of these Nest thermostats or Ecobee thermostats, it's
07:43
Play from here
Copy citation
using AI, right?
07:44
Play from here
Copy citation
It's very likely inside of a lot of people's cars already.
07:48
Play from here
Copy citation
We've got all of these personal assistants and note takers and things that use reminders
07:54
Play from here
Copy citation
and email filters.
07:55
Play from here
Copy citation
Workplace automation is starting to get very, very popular in a variety of different ways.
08:02
Play from here
Copy citation
And then we have all of these different productivity tools that are running loads and loads of
08:08
Play from here
Copy citation
AI, whether you are team Google, team Microsoft, team Apple, there's tons of AI running in
08:15
Play from here
Copy citation
back of all of these products.
08:17
Play from here
Copy citation
So when you kind of break it down into, well, what is the AI that we have today versus this
08:23
Play from here
Copy citation
really crazy promise of the future that's going to flip the world upside down?
08:27
Play from here
Copy citation
Well, that's AGI.
08:29
Play from here
Copy citation
So today's AI, it's narrow and it's task specific, but that's changing quickly.
08:36
Play from here
Copy citation
AGI, this artificial general intelligence, has always been the long-term goal.
08:40
Play from here
Copy citation
It's not today's reality.
08:43
Play from here
Copy citation
But so far, everything that we've built today is doing such a good job that it fools a lot
08:50
Play from here
Copy citation
of people into thinking that it is a general intelligence.
08:54
Play from here
Copy citation
I once heard somebody say that having narrow AI is like having a genius inside a box, but
08:59
Play from here
Copy citation
only if you yell the right question into that box.
09:03
Play from here
Copy citation
So we're living in this really weird time.
09:05
Play from here
Copy citation
We've got all of these tools that are built for narrow tasks, but they feel like something
09:09
Play from here
Copy citation
more because you can type this sloppy prompt into chat GPT and get back some very reasoned
09:16
Play from here
Copy citation
and fluent prose.
09:18
Play from here
Copy citation
It could be bad poetry, but it could be really good code or helping you debug code or explain
09:25
Play from here
Copy citation
calculus or talk like your therapist.
09:28
Play from here
Copy citation
And that feels like intelligence.
09:29
Play from here
Copy citation
And that's where a lot of the confusion begins because we're seeing a broadening of narrow
09:34
Play from here
Copy citation
AI, not true general AI.
09:36
Play from here
Copy citation
And today's large language models are trained on such massive amounts of data that it's
09:42
Play from here
Copy citation
gotten really, really good at mimicking reasoning and conversation, but it's not quite there
09:48
Play from here
Copy citation
yet because it doesn't understand what we're saying.
09:50
Play from here
Copy citation
It can't really generalize across totally new tasks without that human intervention.
09:55
Play from here
Copy citation
It doesn't have goals, doesn't have curiosity, doesn't have self-awareness, but it keeps
10:00
Play from here
Copy citation
getting faster.
10:00
Play from here
Copy citation
It keeps getting better.
10:01
Play from here
Copy citation
It keeps getting even better and better at mimicking.
10:05
Play from here
Copy citation
And that's why you've got companies like Meta literally throwing $100 million salaries at
10:12
Play from here
Copy citation
the brightest minds because whoever cracks the AGI code first will have a head start
10:19
Play from here
Copy citation
on everybody else in a way that some predict will be impossible to catch up.
10:23
Play from here
Copy citation
So as we continue to look into the future here, there is this guy, I've got this slide
10:29
Play from here
Copy citation
up here now of this fellow Alan Turing, and he posed this question, can machines think?
10:35
Play from here
Copy citation
He made a test called the Turing test.
10:37
Play from here
Copy citation
It's in a lot of AI science fiction lore.
10:40
Play from here
Copy citation
And the real brilliance of this test is that it sidesteps the definition of intelligence
10:45
Play from here
Copy citation
entirely.
10:45
Play from here
Copy citation
And instead, it focuses on behavior and perception.
10:49
Play from here
Copy citation
In other words, can a machine convince us that it's human?
10:52
Play from here
Copy citation
And back in 1950, this was all theoretical.
10:54
Play from here
Copy citation
No computer came close to what you have in your pocket with your Android or Apple phone.
11:00
Play from here
Copy citation
It's a completely, completely different world.
11:03
Play from here
Copy citation
So when this test was imagined, it was imagined as this text-based exchange.
11:10
Play from here
Copy citation
Ironically, the first really impressive models that we saw were just typing in text.
11:16
Play from here
Copy citation
And today's LLMs like ChatGPT can pass casual versions of the Turing test without breaking
11:22
Play from here
Copy citation
a sweat.
11:22
Play from here
Copy citation
So the problem with the test is it's too easy to pass by today's standards, right?
11:28
Play from here
Copy citation
We don't have a super intelligence.
11:30
Play from here
Copy citation
But in a five-minute conversation, GPT can appear to be human, but that doesn't mean
11:36
Play from here
Copy citation
it really understands anything.
11:38
Play from here
Copy citation
And that's where we get into this next guy.
11:40
Play from here
Copy citation
I've got a slide up here of Ray Kurzweil.
11:44
Play from here
Copy citation
He is a futurist.
11:46
Play from here
Copy citation
He's the director of engineering at Google.
11:48
Play from here
Copy citation
He's one of the founders of Singularity University.
11:51
Play from here
Copy citation
This idea of a singularity that he really popularized is the point where machines can
11:57
Play from here
Copy citation
start to redesign themselves, right?
11:59
Play from here
Copy citation
Where you have recursive research and improvement on these models, they could potentially turn
12:05
Play from here
Copy citation
it into a bit of a runaway train, depending on who you talk to.
12:09
Play from here
Copy citation
But this is what we're calling super intelligence, or one of the definitions of it.
12:14
Play from here
Copy citation
It's a very quick leap where a new machine intelligence is born.
12:18
Play from here
Copy citation
It's faster, broader, more capable than the entire human species combined.
12:23
Play from here
Copy citation
And then what?
12:24
Play from here
Copy citation
That's the question, right?
12:25
Play from here
Copy citation
So a singularity, these things happen in nature, black holes, waterfalls, right?
12:32
Play from here
Copy citation
Anything where you have a point you cross that is completely unpredictable to know what's
12:37
Play from here
Copy citation
going to happen on the other side of it.
12:38
Play from here
Copy citation
If I'm floating down a river in my canoe, mathematically, I could predict the velocity
12:46
Play from here
Copy citation
of my vessel, the direction of it.
12:49
Play from here
Copy citation
I could get a beat on the trajectory, kind of see when I'm going to hit the edge of the
12:53
Play from here
Copy citation
waterfall.
12:54
Play from here
Copy citation
But as soon as my little boat tumbles over the edge, it's impossible to predict where
12:59
Play from here
Copy citation
it's going to land.
13:01
Play from here
Copy citation
As soon as we cross some kind of recursive model, or a super intelligence, or one entity
13:08
Play from here
Copy citation
that has more smarts than the whole human race combined, it's unpredictable to know
13:14
Play from here
Copy citation
what little quirks it may have and where it may take us, right?
13:17
Play from here
Copy citation
And so that's the idea of a singularity.
13:19
Play from here
Copy citation
So when does this happen?
13:21
Play from here
Copy citation
Kurzweil predicted 2045 as a likely arrival point.
13:25
Play from here
Copy citation
Some believe that all the things we're looking at today with Gemini and GPT and Claude are
13:30
Play from here
Copy citation
stepping stones on that path.
13:32
Play from here
Copy citation
Others say that we're kind of already arriving there, that it's a very gray area.
13:37
Play from here
Copy citation
And then you have people who say that we're nowhere near real autonomy, or agency, or
13:41
Play from here
Copy citation
general reasoning, and that it may never happen, right?
13:44
Play from here
Copy citation
I like to ask AI itself when it thinks it's going to happen.
13:48
Play from here
Copy citation
And surprisingly, and I keep doing this, over the last year or so, I've been getting
13:52
Play from here
Copy citation
a pretty consistent answer that says it's plausible by early to mid-2030s, unlikely
14:00
Play from here
Copy citation
by 2029.
14:02
Play from here
Copy citation
A real singularity, that real tipping point, likely no later than 2045.
14:07
Play from here
Copy citation
So it agrees with Kurzweil.
14:08
Play from here
Copy citation
I don't know, it may have gotten that information from citations from Kurzweil.
14:13
Play from here
Copy citation
So then what happens, right?
14:15
Play from here
Copy citation
And that's the big question.
14:17
Play from here
Copy citation
And there's a lot of different opinions on this.
14:20
Play from here
Copy citation
And it's, I've got a slide up here of painter Bob Ross painting a robot, a humanoid robot
14:28
Play from here
Copy citation
that I generated with AI.
14:31
Play from here
Copy citation
It really like, you know, there's this growing consensus in Silicon Valley that really believes
14:37
Play from here
Copy citation
that we should be setting up universal basic income now.
14:39
Play from here
Copy citation
Because most knowledge work as we know it, and a lot of other work, including logistics,
14:45
Play from here
Copy citation
transportation, are all going to be taken care of for us as a species, and we've got
14:48
Play from here
Copy citation
to be doing something else.
14:51
Play from here
Copy citation
And we still have to get food and do all these other things.
14:54
Play from here
Copy citation
What does that look like?
14:55
Play from here
Copy citation
Are we just all going to be artists?
14:57
Play from here
Copy citation
There are a lot of people who believe that.
14:59
Play from here
Copy citation
And the next slide, I'm going right back to Terminator that nobody seemed to recognize.
15:05
Play from here
Copy citation
Because you have people like Eliezer Yudkowsky, I always mess up this guy's name, he wrote
15:11
Play from here
Copy citation
this book, If Anyone Built It, We All Die, where he doesn't think there is any scenario
15:17
Play from here
Copy citation
after artificial general intelligence that really keeps humanity in the game.
15:22
Play from here
Copy citation
And that's going to be it, right?
15:24
Play from here
Copy citation
So again, all predictions, lots of assumptions.
15:28
Play from here
Copy citation
Nobody really knows.
15:29
Play from here
Copy citation
And that's the only fact we have today is nobody really knows.
15:32
Play from here
Copy citation
So I've got this next slide here that just has a handful of the different large language
15:37
Play from here
Copy citation
models that are popular today.
15:40
Play from here
Copy citation
And when I asked how many people were using these, again, shocking, like a couple of hands
15:45
Play from here
Copy citation
went up.
15:46
Play from here
Copy citation
We've got OpenAI, we've got Meta's model, we've got Anthropic and DeepSeq, DeepMind,
15:52
Play from here
Copy citation
we've got X, Mistral.
15:54
Play from here
Copy citation
Yeah, it's, you know, somebody did ask, well, what's the best, right?
15:58
Play from here
Copy citation
You've got kind of market leaders, you've got adoption, you have time in the game.
16:03
Play from here
Copy citation
But it has been a pretty crazy game of leapfrog, especially lately, right?
16:10
Play from here
Copy citation
Where a lot of them kind of felt the same, where you say, oh, this one's so much better
16:14
Play from here
Copy citation
for creative writing, or this one's so much better for coding.
16:19
Play from here
Copy citation
And then all of a sudden, OpenAI really started taking the lead with GPT, and then Gemini
16:25
Play from here
Copy citation
3 dropped, and everyone was going crazy over it, including Kevin and I.
16:30
Play from here
Copy citation
And then we noticed all sorts of regression or difficulties using it.
16:34
Play from here
Copy citation
And so, again, the answer is to really test all this stuff for yourself, because, yeah,
16:42
Play from here
Copy citation
some of these really do shine in different areas.
16:45
Play from here
Copy citation
But how you prompt them, how you get what you want, like, that's pretty consistent.
16:49
Play from here
Copy citation
That's a lot of the stuff that we teach and go over.
16:52
Play from here
Copy citation
And I've got this next slide, tons and tons of other AI products, right?
16:56
Play from here
Copy citation
And I think this is where things get a little confusing for people, or it feels overwhelming,
17:02
Play from here
Copy citation
that everything in the world is being powered by AI.
17:04
Play from here
Copy citation
And look, there's hundreds of companies and different SaaS products, and they all want
17:08
Play from here
Copy citation
you to use them.
17:09
Play from here
Copy citation
And, well, if we've got all these frontier labs, why do these exist, right?
17:14
Play from here
Copy citation
Well, they're what we call wrapper applications.
17:18
Play from here
Copy citation
In other words, they are providing a layer with a specific service over these foundation
17:25
Play from here
Copy citation
models like chat GPT to give you a better experience for the thing that you're trying
17:30
Play from here
Copy citation
to get.
17:31
Play from here
Copy citation
And I've got lots of examples up here, everything from note takers that you would invite into
17:35
Play from here
Copy citation
your Zoom or Google Meet or Teams call, like Fireflies, for example, or Fathom.
17:42
Play from here
Copy citation
Then we've got other wrappers that help you write code better, like Repl.it or Cursor.
17:49
Play from here
Copy citation
Then you've got things like Grammarly that are just following you around the web and
17:54
Play from here
Copy citation
every application on your computer, helping you with spell check or things that sit on
17:58
Play from here
Copy citation
top of your CRM.
17:59
Play from here
Copy citation
If you're on Salesforce, it might be AgentForce.
18:01
Play from here
Copy citation
If you're on HubSpot, it would be Breeze.
18:03
Play from here
Copy citation
So the reason why these things exist and the reason why they leverage other models is that
18:09
Play from here
Copy citation
these frontier models are very expensive and time-consuming to make, right?
18:14
Play from here
Copy citation
We're talking $100 million plus, a decade of research and evolution, whereas if I'm
18:21
Play from here
Copy citation
one of these companies and I want to use this technology and just point it directly at a
18:25
Play from here
Copy citation
user in their specific use case, for example, say I want to make presentations, Gemini 3
18:35
Play from here
Copy citation
all of a sudden just started knocking out amazing slide decks, right?
18:38
Play from here
Copy citation
There's even a connection to put it into Google Slides.
18:42
Play from here
Copy citation
It's still pretty tricky.
18:44
Play from here
Copy citation
It still doesn't work that well.
18:48
Play from here
Copy citation
If you make a mistake, it's very hard to go back and edit, but you have a platform like
18:53
Play from here
Copy citation
Gamma, and Gamma will produce slide decks for you much, much better and they're editable
19:01
Play from here
Copy citation
and it's easier to get context in there because that's all it does, but it's being powered
19:05
Play from here
Copy citation
by one of these LLMs.
19:07
Play from here
Copy citation
Why don't they just make their own?
19:09
Play from here
Copy citation
Well, again, it's an insane amount of money and time, whereas it can just call the API,
19:18
Play from here
Copy citation
in other words, the little computer-to-computer messaging system behind the scenes for fractions
19:23
Play from here
Copy citation
of a penny every time it uses that tech to power its product.
19:26
Play from here
Copy citation
So today, we already have industries that are being completely turned on their ears
19:32
Play from here
Copy citation
by AI, just to name a few.
19:35
Play from here
Copy citation
In healthcare, AI is detecting early-stage cancers with their scanning technology more
19:41
Play from here
Copy citation
accurately than most radiologists and much, much faster.
19:45
Play from here
Copy citation
In finance, AI is flagging over a billion dollars a year in fraud using real-time anomaly
19:50
Play from here
Copy citation
detection with explanations, not just alert flags.
19:54
Play from here
Copy citation
It's incredible.
19:55
Play from here
Copy citation
And then in transportation, you have companies like Waymo with their autonomous vehicles
20:00
Play from here
Copy citation
that have driven over 7 million miles without any fatalities whatsoever.
20:03
Play from here
Copy citation
And in retail, 35% of Amazon's revenue is coming from their AI-powered recommendations.
20:08
Play from here
Copy citation
You know, the, oh, you bought this water bottle.
20:12
Play from here
Copy citation
You might want this little accessory for the gym or this or this or this.
20:17
Play from here
Copy citation
And I believe that was even before they really fully implemented Rufus, which is their chat
20:21
Play from here
Copy citation
assistant that, I mean, I use all the time now.
20:24
Play from here
Copy citation
I'm not going to be bothered to go through and read reviews when I could just say, hey,
20:30
Play from here
Copy citation
give me a collection of all the one-star ratings, but filter out anything that said that they
20:37
Play from here
Copy citation
didn't like the product because the box was damaged or something, Don, and give me the
20:41
Play from here
Copy citation
truly horrendous stories so I could weigh my purchasing decision right here.
20:46
Play from here
Copy citation
So that alone, that alone is huge, right?
20:50
Play from here
Copy citation
It's really changing things in a material way there.
20:54
Play from here
Copy citation
And this next slide here has four different areas of skepticism that bubbles up.
21:00
Play from here
Copy citation
So if it's doing all these incredible things already, why is there still so much public
21:06
Play from here
Copy citation
pushback for every person who thinks that AI is going to save the world or maybe take
21:11
Play from here
Copy citation
it over?
21:12
Play from here
Copy citation
There's another who thinks like, this is just a dumb fad.
21:16
Play from here
Copy citation
It's a trend.
21:17
Play from here
Copy citation
It's going to pass.
21:18
Play from here
Copy citation
This isn't going to go anywhere.
21:19
Play from here
Copy citation
It's just a big pump and dump for the valuation of a bunch of new companies.
21:24
Play from here
Copy citation
I get it.
21:25
Play from here
Copy citation
This is understandable.
21:27
Play from here
Copy citation
Even though I'm inside of it, the rationale for this thinking is valid.
21:32
Play from here
Copy citation
It doesn't make it so that this is true that it's a fad or it's all going to go away, that
21:37
Play from here
Copy citation
I do not believe whatsoever.
21:38
Play from here
Copy citation
But at the top of the list, we've got hallucinations because, yeah, these large language models
21:42
Play from here
Copy citation
can make things up, right?
21:44
Play from here
Copy citation
So it's a known limitation, hasn't been fixed yet, gets called out all the time.
21:50
Play from here
Copy citation
It's even in a lot of the warnings when you get an answer.
21:52
Play from here
Copy citation
When I am in my car and I'm talking to Siri through CarPlay, I'll often route to GPT because
22:01
Play from here
Copy citation
Siri's just not really good at much at all right now, actually.
22:08
Play from here
Copy citation
And once it routes to GPT after whatever answer it gets that I'm asking about, it says, yeah,
22:16
Play from here
Copy citation
large language models can have hallucinations.
22:17
Play from here
Copy citation
Please make sure to check your facts.
22:19
Play from here
Copy citation
There's a little disclaimer at the end of it, which is really interesting.
22:21
Play from here
Copy citation
So this happens when models are uncertain or lacking grounding in verified data or,
22:29
Play from here
Copy citation
in other words, AI can be a really good bullshitter because it wants to give you the answer, right?
22:36
Play from here
Copy citation
It wants to fulfill the ask.
22:40
Play from here
Copy citation
And that can be scary in a high stakes situation.
22:42
Play from here
Copy citation
OpenAI has since rolled back a lot of what it will give you in regards to legal advice
22:47
Play from here
Copy citation
or medical advice.
22:48
Play from here
Copy citation
And the other thing is decision making, right?
22:51
Play from here
Copy citation
That kind of rolls right into decision making.
22:53
Play from here
Copy citation
AI doesn't have any real emotional context.
22:58
Play from here
Copy citation
You know, it can spot patterns and make recommendations, but it doesn't understand ethics the way that
23:04
Play from here
Copy citation
a human does, not even close.
23:08
Play from here
Copy citation
So, you know, people say like, well, it never will, right?
23:11
Play from here
Copy citation
I don't think that's true, but today it can't.
23:14
Play from here
Copy citation
And so they dismiss it.
23:15
Play from here
Copy citation
And then there's the whole bubble conversation that I don't want to get too deep into this,
23:19
Play from here
Copy citation
but, you know, if we kind of unpack the fundamentals of a bubble, it's any time that you've got
23:27
Play from here
Copy citation
something that has a value associated with it, which is just completely detached from
23:32
Play from here
Copy citation
its true economic fundamentals, right?
23:35
Play from here
Copy citation
We saw this in the dotcom bubble and it does kind of feel like year 2000 all over again.
23:40
Play from here
Copy citation
You've got all these new companies with insane valuations, just like before they were taking
23:46
Play from here
Copy citation
any word and throwing a dotcom at the end.
23:48
Play from here
Copy citation
Now we're throwing a dot AI extension on that domain to get a better valuation.
23:54
Play from here
Copy citation
And yeah, there's a lot of that same, you know, this math doesn't math like the economic
24:01
Play from here
Copy citation
fundamentals, the revenue, the earnings, the profitability of these companies doesn't match
24:06
Play from here
Copy citation
the values that they're getting.
24:09
Play from here
Copy citation
But it's not like the Internet went away after the dotcom bubble burst, right?
24:13
Play from here
Copy citation
It just cleared a lot of the crap out of the way that wasn't really providing much value.
24:17
Play from here
Copy citation
And you had a lot of massive, massive companies born out of that.
24:21
Play from here
Copy citation
And so, again, I understand, though, I understand, because when a company is burning 15
24:27
Play from here
Copy citation
million dollars a day, a day, right.
24:30
Play from here
Copy citation
And then also says within the same week, by the way, we need one point four trillion dollars over the
24:37
Play from here
Copy citation
next eight years for the infrastructure to keep this all going, to hit all of our goals.
24:44
Play from here
Copy citation
You don't have to be a data scientist.
24:46
Play from here
Copy citation
Just call out the fact that those numbers don't quite jive.
24:49
Play from here
Copy citation
Right. So it's a justifiable argument.
24:53
Play from here
Copy citation
And this next slide has one word on it.
24:56
Play from here
Copy citation
And it's yet because all of these all of these arguments, all of the pushback really is where we are
25:04
Play from here
Copy citation
today. And one thing Kevin and I always say on this show and to everybody that we talk to when we
25:10
Play from here
Copy citation
hear complaints about shortcomings is that this is the worst that it's ever going to be.
25:14
Play from here
Copy citation
Right. This is this is this is that we are living in an amazing time.
25:19
Play from here
Copy citation
And next month is going to be more amazing.
25:21
Play from here
Copy citation
And next year is going to be even more amazing.
25:23
Play from here
Copy citation
And five years from now, I have no idea what's going to happen.
25:26
Play from here
Copy citation
But, yeah, it's not always right.
25:27
Play from here
Copy citation
And it makes stuff up.
25:29
Play from here
Copy citation
Can't always execute great creative output yet.
25:33
Play from here
Copy citation
Can't truly reason the monetization use cases just aren't really that strong with so many of these
25:38
Play from here
Copy citation
businesses or compute is just too expensive.
25:42
Play from here
Copy citation
Just that doesn't make fiscal sense yet.
25:45
Play from here
Copy citation
Right. Or it's never going to take my job.
25:49
Play from here
Copy citation
Got to got to got to keep got to keep educating yourself because yet is the is the truth there or can't
25:57
Play from here
Copy citation
even render fingers with I'm sorry, I can't even render hands with five fingers.
26:02
Play from here
Copy citation
Right. And so I've got this slide here that I love this this meme that went around just just two years
26:08
Play from here
Copy citation
ago that had a series of hands that were rendered by A.I.
26:13
Play from here
Copy citation
They just have absolutely nightmarish outputs where everything is twisted and there are thumbs growing out
26:20
Play from here
Copy citation
of risks. And the title says A.I.
26:22
Play from here
Copy citation
accepting the job that sees handshakes of these human hand monstrosities.
26:27
Play from here
Copy citation
Right. For some reason, it just couldn't do this.
26:30
Play from here
Copy citation
It just couldn't render human hands.
26:32
Play from here
Copy citation
It looked real. And even when it did, there was this really strange, uncanny valley where the hands didn't look just
26:37
Play from here
Copy citation
quite right. And then I've got this slide from one of our marketing partners of these three
26:42
Play from here
Copy citation
images that were used as ad creatives just last week.
26:46
Play from here
Copy citation
And they are stunning.
26:49
Play from here
Copy citation
It was this race car campaign, and he got just unbelievable ROAS return on ad spend with these.
26:56
Play from here
Copy citation
I think it was something like 10 or 11 X.
26:59
Play from here
Copy citation
The video assets were just as impressive.
27:02
Play from here
Copy citation
And you know what, to get something like this just a couple of years ago would have involved a casting call and
27:12
Play from here
Copy citation
getting talent in-house and really expensive cameras and line producers and storyboards and tons of editing and
27:21
Play from here
Copy citation
drones on and on and on these commercials, probably at least a couple of million dollars to make.
27:28
Play from here
Copy citation
And these were produced in an afternoon with prompting just two years later after the nightmarish hand debacle.
27:35
Play from here
Copy citation
So this is where it's going.
27:37
Play from here
Copy citation
This is where it's going.
27:39
Play from here
Copy citation
You know, just going back to this idea that you see something and it's not working very well, guess what?
27:45
Play from here
Copy citation
That's the worst it's ever going to be.
27:47
Play from here
Copy citation
And it could be completely fixed.
27:48
Play from here
Copy citation
It could be completely revolutionized within weeks.
27:52
Play from here
Copy citation
We just don't know.
27:53
Play from here
Copy citation
So part of this is really staying open minded to the possibility of innovation here.
27:57
Play from here
Copy citation
So that said, why is there still so much hate about this?
28:05
Play from here
Copy citation
Why do we have such a negativity bias?
28:09
Play from here
Copy citation
Every time AI gets something wrong, there are a lot of loud voices that say, see, this is this is just never going to work.
28:17
Play from here
Copy citation
This isn't this isn't going to happen.
28:19
Play from here
Copy citation
This this too shall pass.
28:20
Play from here
Copy citation
So the reason why is it's hardwired into our nature for survival.
28:25
Play from here
Copy citation
You know, we evolved to look for perceived threats to exaggerate downside.
28:29
Play from here
Copy citation
And we did that in order to stay safe.
28:31
Play from here
Copy citation
So I think autonomous vehicles are a great example of this because, A, it's an emerging technology.
28:37
Play from here
Copy citation
B, it's running loads and loads of AI.
28:40
Play from here
Copy citation
And C, when something bad happens, everybody hears about it and says, you got to be crazy to let one of these Teslas drive you around.
28:49
Play from here
Copy citation
Do you care anything about your life?
28:50
Play from here
Copy citation
Right.
28:51
Play from here
Copy citation
So this slide has this news story and the title is Cartoon Frank crashes Tesla awkwardly rehashes NHSTSA investigation findings.
29:03
Play from here
Copy citation
So what somebody did was they put a painting of the road on the road so that it looked like it kept going.
29:11
Play from here
Copy citation
One of these old school illusions that you might see in a Roadrunner cartoon where the Roadrunner tricks Wile E.
29:17
Play from here
Copy citation
Coyote into going off a cliff.
29:19
Play from here
Copy citation
by having them run through a painting and the Tesla fell for it and it crashed right
29:24
Play from here
Copy citation
into this thing. And it got global headlines for it. And, you know, if you look at the
29:33
Play from here
Copy citation
big picture of autonomous vehicles and just look at last year, we had almost 40,000 people
29:39
Play from here
Copy citation
die in traffic crashes in the U.S. alone. Right. And why does that happen? Because at
29:46
Play from here
Copy citation
any time over, you know, 660,000 drivers are using their phones while they're driving.
29:52
Play from here
Copy citation
They're looking at the screen in their lap and not the windshield right in front of them.
29:57
Play from here
Copy citation
That's not good. And then you have over a million arrests for driving under the influence
30:01
Play from here
Copy citation
every year. And then you have almost half of drivers who self-report that they struggle
30:05
Play from here
Copy citation
to stay awake while driving. So when you look at autonomous vehicles compared to that, I've
30:12
Play from here
Copy citation
got this little video clip of me getting into a Waymo in San Francisco with my family.
30:18
Play from here
Copy citation
And when I was there about a year ago, these cars were everywhere. And some of them had
30:23
Play from here
Copy citation
somebody in the front seat, probably one of the Google employees, making sure this thing
30:27
Play from here
Copy citation
didn't run into somebody in an intersection or just taking notes on how it behaved. I
30:32
Play from here
Copy citation
don't know what they were doing in there. But when I came back a year later, there were
30:36
Play from here
Copy citation
people in the Waymos in the backseat and there were even more of them. And they were taking
30:41
Play from here
Copy citation
passengers all over the place, more than Ubers in the city. I forget how many more.
30:48
Play from here
Copy citation
They had a massive share of all of the rideshare platforms were now the autonomous vehicles.
30:54
Play from here
Copy citation
And why? Why would you get in a car driven by a robot? That seems so scary. But they
31:02
Play from here
Copy citation
have 85% fewer injury crashes. They have a 96% reduction in intersection collisions.
31:08
Play from here
Copy citation
I like this stat because I ride a bicycle. 82% fewer cyclists and motorcyclist injuries.
31:12
Play from here
Copy citation
92% reduction in pedestrian injuries. But here's a real kicker. Of any incident they
31:17
Play from here
Copy citation
had, 90% of those occurred because of a human error. Not somebody who's programming the
31:21
Play from here
Copy citation
car, but somebody who ran out into an intersection or ran a red light or went through a stop
31:25
Play from here
Copy citation
sign or was texting while driving. So when you ask these questions of, is it perfect?
31:32
Play from here
Copy citation
It's not the right question. The right question is, is it better than a human? And in this
31:35
Play from here
Copy citation
case, yeah, it's over 90% better. And these are high stakes, life and death situations.
31:40
Play from here
Copy citation
And listen, I grew up in San Francisco. I hate driving there. I got in this thing with
31:45
Play from here
Copy citation
my family and I felt infinitely safer than I ever have inside of any of the rideshares.
31:51
Play from here
Copy citation
And this thing drove through rush hour, construction zones, a convention center, flawlessly, absolutely
31:57
Play from here
Copy citation
flawlessly, super, super impressive. And that brings us to the AI super cycle that we are
32:04
Play from here
Copy citation
currently in. What is a super cycle? You might ask. Well, this slide has a diagram of a hockey
32:11
Play from here
Copy citation
stick growth curve. And at the bottom of that, we've got stuff like fire, right? That gave
32:16
Play from here
Copy citation
us warmth and some lights. And then we got steam powered engines and electricity give
32:22
Play from here
Copy citation
us even more warmth and light and machines and other things made our little existences
32:28
Play from here
Copy citation
a lot more comfortable and convenient. And we started the hockey stick part of that trajectory.
32:34
Play from here
Copy citation
And we got the internet. We got the microprocessors. We got the internet. We had this global connection
32:40
Play from here
Copy citation
and communication happen. And it started changing the way that we do business and interact with
32:45
Play from here
Copy citation
each other. And then we've got today, we've got generative AI, right? And so this curve
32:52
Play from here
Copy citation
represents the fact that with every one of these technological breakthroughs, the actual
32:57
Play from here
Copy citation
curve gets steeper and steeper. And when you consider things like the singularity or
33:03
Play from here
Copy citation
artificial general intelligence, that curve becomes indistinguishable from a straight
33:08
Play from here
Copy citation
line up, almost defines the singularity in itself, right? It's like this unpredictable
33:12
Play from here
Copy citation
exponential trajectory that just looks like you're on this rocket ship growth curve. But
33:19
Play from here
Copy citation
AI is a very different technology than any of the other super cycles that we've experienced
33:24
Play from here
Copy citation
as a species. It's spreading wider. It's impacting nearly every white collar role as
33:31
Play from here
Copy citation
well as some of the ones that aren't like logistics, right? And so when you look at
33:36
Play from here
Copy citation
this compared to the internet, I don't really think it's a fair comparison. And this might
33:42
Play from here
Copy citation
ruffle some feathers out there or a lot of people might disagree, but I think AI is more
33:46
Play from here
Copy citation
akin to the industrial revolution. But here's the real big difference. Instead of spanning
33:51
Play from here
Copy citation
100 years, what if you compress the industrial revolution into five years or 10 years? What
33:57
Play from here
Copy citation
would that have looked like? What would that have done to the world? And that's why this
34:02
Play from here
Copy citation
stuff is so important to pay attention to because the real challenge here is maintaining
34:07
Play from here
Copy citation
organizational agility and staying flexible enough to adapt as the pace of AI's capability
34:13
Play from here
Copy citation
and all of the change that it's creating accelerates because it's going to keep accelerating
34:17
Play from here
Copy citation
and it's going to keep going faster and the curve is going to keep getting steeper.
34:20
Play from here
Copy citation
So on to what we can do, what we can't do. This slide is called Generative AI Capability
34:27
Play from here
Copy citation
and Boundaries. And to deploy AI anywhere, we have to be ruthlessly honest about what
34:34
Play from here
Copy citation
it's good at, what it's not so good at. And today, its primary superpower is its ability
34:39
Play from here
Copy citation
to process, analyze, to generate information at speed and scale that you just can't do
34:45
Play from here
Copy citation
with human teams. This is a game changer for productivity, right? It can recognize patterns,
34:50
Play from here
Copy citation
very subtle patterns across massive, massive data sets. It would take human analysts months
34:55
Play from here
Copy citation
and months to discover and it can do it in a prompt. The interface, right? The natural
35:02
Play from here
Copy citation
language interface is pretty critical. This idea that you can have conversations with
35:07
Play from here
Copy citation
the machine to get what you want instead of being somebody who knows how to write code
35:11
Play from here
Copy citation
is also a massive unlock. And as far as limitations, we already covered the thing
35:15
Play from here
Copy citation
about hallucinations. It's pretty critical. It's factually very powerful, but can still
35:21
Play from here
Copy citation
confidently deliver things that sound very plausible, but they're incorrect.
35:28
Play from here
Copy citation
So we've got to verify outputs. Models that really struggle to show how they work. So
35:34
Play from here
Copy citation
there's this explainability. This goes all the way into the Frontier Labs with the people
35:38
Play from here
Copy citation
building these actual learning systems. It's very difficult to go into the models and figure out
35:45
Play from here
Copy citation
why it gave the answer that it did. That is still a very murky science. Then you got all the human
35:51
Play from here
Copy citation
context and novelty that we talked about too, right? AI is trained on past data and it often
35:55
Play from here
Copy citation
struggles with novel or unexpected situations or interpreting subtle nuance. It doesn't really get
36:02
Play from here
Copy citation
things like sarcasm or emotional context. It doesn't write very good jokes. So the goal here
36:07
Play from here
Copy citation
is to maximize the strengths while building human-in-the-loop processes, right? We always
36:12
Play from here
Copy citation
advocate for this being a force multiplier for humans, not a replacement for humans.
36:18
Play from here
Copy citation
So how do you do that, right? How do you get it to be a force multiplier for you or your
36:23
Play from here
Copy citation
business or your role? We have these four practical pillars of AI. We have, starting off with
36:31
Play from here
Copy citation
development-enabled and really this is one of AI's superpowers, like its ability to accelerate
36:37
Play from here
Copy citation
the speed and the quality of coding, even the debugging and the testing. You get more focused
36:42
Play from here
Copy citation
teams and they're going to deliver very complex features at a rate that they couldn't achieve
36:46
Play from here
Copy citation
before. You've got the customer-facing integration. So one example of this would be the ultimate
36:53
Play from here
Copy citation
layer of personalization where on the right platform and the right knowledge, say you had
36:58
Play from here
Copy citation
an email list of 10,000 recipients and you're going to do a blast and they all get the same
37:03
Play from here
Copy citation
message, right? I don't know what you're going to talk about your sale that you're having on
37:07
Play from here
Copy citation
whatever widget you're making. And instead of sending that same message, you could do all sorts
37:11
Play from here
Copy citation
of workloads to have data enrichment about the person that you're sending it to. And with the
37:18
Play from here
Copy citation
same efforts, because machines are doing it all, you could have a unique message go out to each
37:22
Play from here
Copy citation
one of those 10,000 people based on their social media platforms, based on their LinkedIn profile,
37:28
Play from here
Copy citation
based on previous communications. So that is also moving really, really quickly.
37:33
Play from here
Copy citation
Workforce optimization, again, it's just endless examples here. But handling the repetitive,
37:41
Play from here
Copy citation
time-consuming tasks or data-intensive tasks, summarizing reports, taking notes,
37:46
Play from here
Copy citation
synthesizing research, managing schedules, all the crap that nobody really wants to do,
37:51
Play from here
Copy citation
but we have to do, it's already happening, right? We're already solving those use cases,
37:56
Play from here
Copy citation
hopefully so that the skilled employees can focus on high value, creative or strategic,
38:02
Play from here
Copy citation
human-centric work, not machine work. Then you've got the future-proofing, right? This is the idea
38:06
Play from here
Copy citation
of using advanced AI for real deep forecasting, for predictive modeling, for things that humans
38:13
Play from here
Copy citation
also haven't necessarily been so great at in the past, right? And that machines can take
38:19
Play from here
Copy citation
massive amounts of data and be able to get much better models and take a look ahead to see
38:26
Play from here
Copy citation
what is going to happen or how it can help usher you into this next era based on data that you
38:33
Play from here
Copy citation
could just never chew through yourself or even with a big team. So how do you start doing all
38:39
Play from here
Copy citation
this? Well, I've got this little pyramid slide and at the base of the three levels of this pyramid,
38:43
Play from here
Copy citation
we've got the things we don't like to do, and we already covered some of those things, right?
38:49
Play from here
Copy citation
So again, like capturing the meeting notes, going through a bunch of documents, contracts,
38:58
Play from here
Copy citation
or spreadsheets, issue management, customer service, all this stuff that takes an army of
39:05
Play from here
Copy citation
people that is really just reading stuff, replying to stuff, moving pixels around a screen, and nobody
39:12
Play from here
Copy citation
really likes to do it. Not a lot of controversy here and not a lot of risk, right? Drives immediate
39:17
Play from here
Copy citation
time savings for every knowledge worker, shows undeniable value within a 24-hour period. And
39:24
Play from here
Copy citation
any of these examples can save one to three hours per employee per week instantly, right?
39:30
Play from here
Copy citation
So the second tier up in the pyramid, the things we don't do very well. And this is where we really
39:36
Play from here
Copy citation
start getting into the conversation of ROI and eliminating human error at scale. Not eliminating
39:44
Play from here
Copy citation
humans, but human error, augmenting humans with superpowers, right? So a good example of this is
39:50
Play from here
Copy citation
really monitoring transactions in real time. So flagging abnormal vendor charges or duplicates or
39:55
Play from here
Copy citation
policy violations, fraud signals, all of that stuff, it does incredibly well. So yeah, again,
40:03
Play from here
Copy citation
humans just can't go and look at thousands of different data points simultaneously and scan
40:08
Play from here
Copy citation
them. It just doesn't happen. So way at the tip of the pyramid, we've got the real business
40:14
Play from here
Copy citation
changing stuff. And this is where AI stops just saving money and starts making money, right?
40:21
Play from here
Copy citation
So if it can learn customer behavior and intent or personalize offers and messaging,
40:26
Play from here
Copy citation
if it can optimize itself continuously, offer to be like a personal concierge for everybody that
40:34
Play from here
Copy citation
you work with all the time, that is a game changer. That changes everything, revenue,
40:40
Play from here
Copy citation
conversions, retention, does upsells, lifetime value. It really starts to build this competitive
40:46
Play from here
Copy citation
mode that becomes difficult for others to copy because it's so specific to your business.
40:52
Play from here
Copy citation
So when we look at, well, what are the roles that AI plays inside of an organization,
40:57
Play from here
Copy citation
we've got these three different categories and they kind of mimic human categories, right? You've
41:05
Play from here
Copy citation
got clerks, colleagues, and coaches. And so when we look at clerks, the definition would be,
41:11
Play from here
Copy citation
what can AI do that performs like really fast, accurate and rule following tasks, right?
41:16
Play from here
Copy citation
Think of it as like your intern or executive assistant who just happens to have a PhD in
41:23
Play from here
Copy citation
everything in the world and is often just completely free, right? That's what a lot of
41:27
Play from here
Copy citation
these AI tools are. So drafting standardized reports or reviewing documents, extracting data
41:33
Play from here
Copy citation
from PDFs, again, basic summaries, low risk, high efficiency augmentation, right? That automates
41:41
Play from here
Copy citation
repetitive friction that slows down your organization. Then you've got colleagues,
41:46
Play from here
Copy citation
and these are more of collaborators as aside from task doers. And so this might be multi-step
41:53
Play from here
Copy citation
workflows and things that need to adapt to context and provide surface level insights like
41:59
Play from here
Copy citation
reviewing compliance patterns in a legal document, right? It's not going to, open AI won't give you
42:04
Play from here
Copy citation
actual hardline legal advice anymore, but it will still review a lot of documents for you.
42:09
Play from here
Copy citation
So again, what's a value? Reduces cognitive load, ensures procedural accuracy across complex tasks.
42:17
Play from here
Copy citation
And last we got coaches, and these are AI tools as thinking partners. You could think of this
42:22
Play from here
Copy citation
as the most sophisticated form of a partnership. And so this AI helps individual contributors
42:28
Play from here
Copy citation
reflect and refine and grow by offering real feedback, alternative suggestions like,
42:34
Play from here
Copy citation
hey, do I have any blind spots? What might I be missing here? Can you think of a different
42:38
Play from here
Copy citation
angle that I'm not viewing? Can you look at what company XYZ over here is doing? They're
42:44
Play from here
Copy citation
my competitor. Can you reverse engineer their marketing plan and then compare it to mine?
42:49
Play from here
Copy citation
So this is a real unlock, right? This is elevating human decision-making. It's accelerating
42:57
Play from here
Copy citation
professional development. This is the real bridge, the real AI unlock that is what we
43:05
Play from here
Copy citation
are hoping to help people move towards to help their roles and the businesses that they run.
43:12
Play from here
Copy citation
So this is the beginning of the fun killer slides, the essential enabling conditions that
43:18
Play from here
Copy citation
we always talk about, right? We start with an AI manifesto. What does that mean? Well,
43:24
Play from here
Copy citation
you could think of it as a high-level, ethical, driven, mission-aligned principle for AI use
43:31
Play from here
Copy citation
across your whole company. So what are the shared values? What will AI never be used for?
43:38
Play from here
Copy citation
This provides a North Star or a bit of a moral compass for the organization as it begins to
43:44
Play from here
Copy citation
innovate with AI. Then you've got policy, right? You've got the framework establishing real
43:50
Play from here
Copy citation
governance that is designed to be agile, but still puts guardrails around everything as the
43:56
Play from here
Copy citation
technology and regulatory landscape changes. So
43:59
Play from here
Copy citation
This goes into continuous compliance or data security or risk mitigation as you deploy these new models in your organization.
44:08
Play from here
Copy citation
Then there's the whole cross-organizational literacy.
44:10
Play from here
Copy citation
So this is about developing real targeted or role-specific AI training that addresses the needs of everyone from people out in the field to executives and the C-suite.
44:23
Play from here
Copy citation
So this makes sure that every part of the organization understands how AI affects their specific job and the organization as a whole.
44:30
Play from here
Copy citation
Then you have defined ownership structure.
44:33
Play from here
Copy citation
So this is about a cross-functional alignment with very clear accountability for every AI initiative.
44:41
Play from here
Copy citation
And it's hopefully going to avoid what we call shadow IT.
44:46
Play from here
Copy citation
I'll get a little more into this in a minute.
44:48
Play from here
Copy citation
But really a responsible way to maintain these models, the compliance, and the outcome for everybody and move it towards success as opposed to sketchiness.
45:00
Play from here
Copy citation
Speaking of, you've got safe experimentation zones.
45:04
Play from here
Copy citation
So every organization should build some kind of a sandbox environment that has good structured feedback loops for controlled innovation.
45:12
Play from here
Copy citation
So you want teams to be able to test and iterate quickly without risking production systems or having a hallucination result in a lawsuit.
45:21
Play from here
Copy citation
And that goes into the last one, which is this phased implementation.
45:25
Play from here
Copy citation
So prioritizing projects in a very strategic way where you balance two factors.
45:30
Play from here
Copy citation
On one side, you've got ROI potential, the real business win.
45:35
Play from here
Copy citation
And then you've got organizational risk, the safety factor.
45:39
Play from here
Copy citation
The goal is to launch low-risk, high-return projects first because those build momentum.
45:44
Play from here
Copy citation
You get these little wins that turn into bigger wins.
45:47
Play from here
Copy citation
And then they scale into completely transformative parts of the business.
45:52
Play from here
Copy citation
So the ultimate fun killer slide is risk and governance imperatives.
45:56
Play from here
Copy citation
And there's this graphic of a giant iceberg.
46:00
Play from here
Copy citation
Same graphics used in like every business and marketing presentation ever.
46:05
Play from here
Copy citation
Unreported AI is the visible part of the iceberg.
46:08
Play from here
Copy citation
This is 43% of AI used in these organizations, whereas the bulk of the iceberg sitting concealed under the water is 57%,
46:17
Play from here
Copy citation
meaning that out of the 75% of employees using AI, most are doing it very quietly for a variety of reasons.
46:25
Play from here
Copy citation
Some people just see it as this little cheat code, like I've got one up on my colleagues.
46:29
Play from here
Copy citation
Oh, they'll never know that I did this whole presentation using AI or that I don't even reply to my email anymore.
46:34
Play from here
Copy citation
My AI agent does it for me.
46:36
Play from here
Copy citation
So this gets into really sketchy territory because we mentioned some of the risks or some of the things that AI doesn't really do very well.
46:44
Play from here
Copy citation
So when employees use these third-party tools without any guidance whatsoever or any regulatory or any guidelines,
46:50
Play from here
Copy citation
they create things like data leakage or IP exposure, regulatory breaches,
46:56
Play from here
Copy citation
things their clients may not want being fed back into a large language model because the user didn't know how to use the appropriate settings.
47:04
Play from here
Copy citation
So these risks are happening now because of this unreported usage.
47:10
Play from here
Copy citation
And the unreviewed AI-drafted documents that may contain noncompliant language get out there,
47:18
Play from here
Copy citation
and employees might paste proprietary code or sensitive company data into a public third-party AI tool.
47:25
Play from here
Copy citation
It's not good. It's not good.
47:27
Play from here
Copy citation
So the goal here is to build governance structures like the ones we were just talking about that encourage responsible use,
47:33
Play from here
Copy citation
guided use, make it easy for employees to use the tools, but in a secure way.
47:38
Play from here
Copy citation
What does the future hold? What's next in AI, right?
47:41
Play from here
Copy citation
Well, some of the things that are starting to come up would be like System 2 AI, the reasoning engines.
47:48
Play from here
Copy citation
And this is moving from being the awesome pattern matcher to a deliberate thinker.
47:55
Play from here
Copy citation
So this means that a lot of these models are starting to perform these multi-step solutions.
48:01
Play from here
Copy citation
They can plan and verify the steps, self-correct their errors.
48:06
Play from here
Copy citation
This leads to real complex logic, real complex math, the coding problems that were completely out of reach previously.
48:14
Play from here
Copy citation
So this is huge, and we're living it right now.
48:17
Play from here
Copy citation
Then you've got the agentic workflows. Feels like it's happening a bit slower.
48:20
Play from here
Copy citation
I've seen a couple of crazy things lately, but nothing like the real promise of 2025 is the year of the agents,
48:26
Play from here
Copy citation
where we were going to have this big move from generative AI to the one that will just autonomously navigate our CRM
48:35
Play from here
Copy citation
and interact with our calendar and click buttons in our software.
48:38
Play from here
Copy citation
I'm starting to see it happen. It's slow. It's expensive. It's super buggy.
48:44
Play from here
Copy citation
But again, we'll go back to my slide that just said, yet, hasn't happened yet.
48:48
Play from here
Copy citation
I've seen some promising markers on the road that say that it's around the corner.
48:54
Play from here
Copy citation
Then you've got this infinite context and memory shift.
48:57
Play from here
Copy citation
So the future models are going to remember context and preferences across weeks and months, maybe even years,
49:04
Play from here
Copy citation
maybe even in perpetuity as compute keeps getting more powerful and storage keeps getting cheaper.
49:11
Play from here
Copy citation
Right now, it's a shortcoming.
49:14
Play from here
Copy citation
Context windows keep getting bigger, but I think I've talked about this before.
49:19
Play from here
Copy citation
You might be deep in a thread, one of these chats, and it starts getting really slow.
49:24
Play from here
Copy citation
And so you start a new chat and it loses all the context of your last chat.
49:27
Play from here
Copy citation
A lot of these Frontier Labs, I've seen this in Gemini, I've seen it in GPT,
49:33
Play from here
Copy citation
will pull up context from threads that are completely unrelated, drop in names, examples, circumstances.
49:40
Play from here
Copy citation
It's really getting good at this cross-thread contextual referencing.
49:46
Play from here
Copy citation
It's one of those little hints into this perpetual and unlimited context and memory.
49:51
Play from here
Copy citation
Last, we have this native multimodality.
49:54
Play from here
Copy citation
It's really a move to true omni-perception of the physical world.
50:01
Play from here
Copy citation
How do we consume the physical world?
50:04
Play from here
Copy citation
It's not just through a text chat.
50:06
Play from here
Copy citation
We talk, there's audio, we see, there's video, and we can do it all simultaneously.
50:14
Play from here
Copy citation
These AI models are starting to dynamically interpret all of those inputs
50:20
Play from here
Copy citation
so that you don't have to use a model to generate video versus a model to generate an image
50:25
Play from here
Copy citation
versus a model to generate a report.
50:27
Play from here
Copy citation
Pretty soon, they're all going to be baked into the same model.
50:31
Play from here
Copy citation
The path forward, what do we recommend?
50:35
Play from here
Copy citation
What do I tell this MBA class as far as getting off the presentation and actually doing something?
50:42
Play from here
Copy citation
It's a really personal application.
50:43
Play from here
Copy citation
That's where this all goes.
50:44
Play from here
Copy citation
Experiment with it.
50:46
Play from here
Copy citation
Start from the bottom of the pyramid, the mundane task.
50:49
Play from here
Copy citation
What do you hate doing?
50:50
Play from here
Copy citation
Have it summarize emails or draft agendas or plan some travel.
50:55
Play from here
Copy citation
Build some real AI literacy and intuition.
50:58
Play from here
Copy citation
This is the fastest way to understand its power and its limitations.
51:03
Play from here
Copy citation
The action is, if you look at that iceberg slide with all of the risk,
51:09
Play from here
Copy citation
you still also want to be auditing your organization or your role, playing both sides.
51:18
Play from here
Copy citation
You want to understand the risks and the areas of high employee demand
51:25
Play from here
Copy citation
so that you can cut that off at the pass and not let things go off the rails.
51:31
Play from here
Copy citation
What can you do about it?
51:33
Play from here
Copy citation
What's a good action item once you are literate and you understand the risks and shortcomings?
51:39
Play from here
Copy citation
You can start to pilot things.
51:41
Play from here
Copy citation
You can pick one, two, or three high-impact opportunities across those four pillars that we talked about.
51:48
Play from here
Copy citation
Look at that AI pyramid and the different ways that you can move up.
51:53
Play from here
Copy citation
If you can, again, achieve those quick wins, even in a silo,
51:58
Play from here
Copy citation
and then bring them to management or bring them to leadership
52:01
Play from here
Copy citation
if you're not running some entity yourself, improve the ROI of the technology.
52:06
Play from here
Copy citation
Trust me, it's going to get traction.
52:08
Play from here
Copy citation
You're going to get traction in the organization.
52:10
Play from here
Copy citation
And then from that, before it gets deployed, of course,
52:13
Play from here
Copy citation
you develop that governance framework and the roadmaps
52:16
Play from here
Copy citation
and start to create a role-stable, compliant foundation that's needed
52:22
Play from here
Copy citation
to take something from pilot to true organizational transformation.
52:26
Play from here
Copy citation
I also included a slide of some of the things that we use.
52:29
Play from here
Copy citation
I'm not even going to rattle off all the foundation models.
52:32
Play from here
Copy citation
You know them.
52:33
Play from here
Copy citation
I hope I know.
52:34
Play from here
Copy citation
We use all of them, and we use the note-takers, and we use the dictation platforms.
52:39
Play from here
Copy citation
We use the all-in-one brains like ClickUp.
52:42
Play from here
Copy citation
We use the coding assistance tools like Cursor and Repl.it.
52:46
Play from here
Copy citation
So that was it.
52:47
Play from here
Copy citation
That was it.
52:48
Play from here
Copy citation
I really hope that you found that helpful.
52:51
Play from here
Copy citation
I really hope that this MBA class found that helpful.
52:55
Play from here
Copy citation
Again, I went into this with a lot of hesitation of explaining what a
53:00
Play from here
Copy citation
language model is and where this tech came from,
53:03
Play from here
Copy citation
kind of assuming that everybody was as deep in the weeds as Kevin and I are.
53:08
Play from here
Copy citation
That's not the case.
53:09
Play from here
Copy citation
It's not the case.
53:10
Play from here
Copy citation
So one of the messages I left them with is if you aren't using it,
53:15
Play from here
Copy citation
if you're not familiar, if you're not literate, it's not too late.
53:19
Play from here
Copy citation
And just because you hear about it a lot in the news
53:21
Play from here
Copy citation
or it's coming up in your RSS feed all the time or you hear about it in podcasts
53:26
Play from here
Copy citation
and news stories, it doesn't mean it's too late.
53:28
Play from here
Copy citation
We are early.
53:29
Play from here
Copy citation
We are super, super early with all of this.
53:31
Play from here
Copy citation
So if you start now, it's going to be a superpower for you next year,
53:36
Play from here
Copy citation
the year after that, the year after that,
53:38
Play from here
Copy citation
because if you go back to this super cycle graph,
53:40
Play from here
Copy citation
we are just on the approach to a straight-up trajectory.
53:44
Play from here
Copy citation
Thanks.
53:45
Play from here
Copy citation
I hope this helps you use AI at work, and we'll see you next week.
53:48
Play from here
Copy citation
Thanks for listening.
53:53
Play from here
Copy citation
If you'd like to send us a question to answer on the show,
53:56
Play from here
Copy citation
email me at Elijah at Spark6.com.
53:59
Play from here
Copy citation
That is S-P-A-R-K, the number 6.com, or Kevin at Kevin at AscendLabs.ai.
54:06
Play from here
Copy citation
To learn more about AI strategy and training for your business,
54:09
Play from here
Copy citation
visit Kevin's company, Ascend AI Labs at AscendLabs.ai.
54:15
Play from here
Copy citation
And to learn more about design and development of custom AI solutions,
54:19
Play from here
Copy citation
visit my agency at Spark6.com.
54:22
Play from here
Copy citation
You can also sign up for my free newsletter,
54:24
Play from here
Copy citation
AI Framework Friday, by visiting Spark6.com forward slash newsletter.