Video summary
The video explores how organizations increasingly outsource their thinking processes to AI tools like Otter.ai or Zoom's built-in summaries without critically verifying the outputs against original conversations. This shift transforms these digital notes into an organization's long-term memory, replacing direct human recollection with algorithmic interpretations of meetings and discussions. The speaker argues that while this convenience is appealing, it risks eroding the deep cognitive engagement required for true insight generation. By allowing AI to generate first drafts without rigorous critique or subsequent refinement by humans, companies effectively hand over their intellectual labor, leading to a loss of the internal "grinding and crunching" where ideas are genuinely contested and developed collaboratively before reaching publication.
The speaker illustrates this trend with personal anecdotes from his time at New Science Associates, describing how mundane tasks like stuffing newsletters were once vital learning opportunities for junior employees that fostered deep client relationships and business understanding. These roles provided a necessary rung on the ladder for new hires to grasp company operations through direct involvement in grunt work, but AI is now consuming these entry-level functions before young professionals can gain essential experience. Consequently, there is a danger of "sawing off the bottom of the ladder," creating a future where experienced individuals who understand business nuances are scarce because no one has had the chance to learn those foundational skills through hands-on participation in simple tasks that AI now automates effortlessly.
A critical distinction made between cognitive offloading and cognitive surrender highlights the subtle danger inherent in this technological shift. Cognitive offloading is described as a mindful process where humans supervise well-performing AI tools while remaining actively engaged, whereas cognitive surrender involves blindly handing over work to these "magical capabilities" without maintaining oversight or understanding of the underlying processes. The speaker notes that current critiques calling AI merely an averaging tool are too simplistic and fail to capture its power, yet in practice, organizations often accept averaged outputs as final products because they appear good enough initially. This lack of critical engagement means we lose the ability to trace provenance and have meaningful conversations about specific points of view, effectively silencing human voices that could challenge or refine these algorithmic generalizations.
Ultimately, the speaker emphasizes that this transition is not driven by malicious intent or obvious errors but rather by a series of convenient decisions made over time that gradually erode human involvement in thinking processes. The central question shifts from who made mistakes to how organizations will respond to seeing these forces reshape their workflows and knowledge retention strategies. Having spent nearly thirty years on the side advocating for mindful curation of ideas, the speaker views this boundary between human thought and AI assistance not as a scary frontier but as an interesting place to stand where one can choose which scenarios allow for continued intellectual growth versus those that lead to surrendering our thinking capabilities entirely.
Read the full video transcript
Your organization is deciding mostly by
accident how much of its thinking to
outsource.
I've spent the last 30 years on the
other side of that line and can report
to you what it looks like from over
there. Are you using Otter, Fireflies,
Fathom, Granola, or the built-in AI
summaries that Zoom offers or anything
like those? Aren't they incredibly
convenient? Aren't they awesome? Aren't
you glad to have that burden taken off
your hands? Well, more often than not,
we're not checking those summaries
against the originals. And those become
your organization's long-term memory.
Not the original conversation, but the
notes of it. And I don't know about you,
but I remember somebody long time ago
telling me that whoever takes notes in a
meeting, the minute taker, has a really
important role because how what was said
gets expressed and recorded matters a
lot and can guide future conversations.
Right now, you're turning those things
over to the AIS. If you've been
disciplined about using AI, you likely
have AI critique your first draft. And
um after a while, for pressure from
above to do things more quickly, you're
probably turning over first draft
creation to the AIS.
And after a while, you keep reading what
they're output, and it's pretty good. In
fact, over time for a lot of people that
first AI draft ships. It becomes your
final document.
Now, writing is not easy. I I I find
writing kind of torturous, even though
apparently what I write comes out pretty
well. But what we're missing in this
process is the grinding and crunching
noises in our heads that are how we
thought about the issue. And if we're
collaborating with somebody else, the
contention of ideas and that discussion
that leads to insights in our brains.
This is beginning to go by the wayside
as we turn over the writing of our
points of view to AI.
Too many years ago to say the actual
date, I worked at a little company
called New Science Associates. We had a
retainer market research around
technology stuff. I was their neural
networks analyst. Um, when we had one or
two of these retainer services, every
two weeks we would publish a newsletter
and when the when the newsletter came
back from the printer, we would have a
stuffing party and everybody would drop
everything they were doing, we would
show up in the conference room where all
of the paper issues all stapled up were
sitting and waiting with envelopes and
labels. And all of us in the company,
everybody from sales, Joan, Anita, uh
Ian, Justin, uh from researchers like
me, Bill, John Pop, everybody else would
come together, and we would take the
labels, put them on the envelopes, stuff
the issue into the envelopes, and get
them ready for somebody to wheel across
the street to the post office. These
stuffing parties were grunt work, but
they were really fun. We'd order some
pizzas and have some drinks, and um we
got to see the labels and hear the
stories about our clients. So, you got
to know the client base and as I was in
research and I got to see, oh yeah,
yeah, yeah, I just answered an inquiry
for this person here. That was key.
That's just one kind of grunt work. But
the first drafts, the the simple tasks,
all those things that were the way that
junior employees learn the business,
those are going away because they're
being eaten by AI. It's uh we're sawing
off the bottom of the ladder. And in 10
years when we need experienced people
who understand the business with some
depth, there's a good chance we're going
to miss a lot of that experience because
there was no place, there was no rung
for young people to sort of grab on and
start to understand what was happening.
A common critique of AI these days is
that it's just a stochastic parrot. It's
just an averaging of of knowledge. And I
disagree with those. I think they're way
too simplistic ways of of thinking about
the great power that lies in there.
But they are kind of averaging what's
happening. They're they're getting a
sense of everything. They can tell what
a maple leaf looks like because they
understand what average mapleness looks
like and they can distinguish it from
average oakeness, for example. Now, as
we start shipping things written with
AI, we can be more critical and we can
ask the AIS to take different points of
view, but that's kind of advanced use.
So in many cases what's happening is
something that's averaged and then kind
of unsigned shows up shipping. Back in
the day when I wrote for New Science and
after that for Esther Dyson, my name was
on everything I wrote. Everything had
providence. Everything had a label that
you could go and ask the person who
created it, hey what did you mean and I
disagree with this and you could have a
really interesting conversation. We seem
to be losing some of that now to AI.
As Beth Caner pointed out in her comment
on my previous post, the theory behind
us handing over our thinking to AIS is
called cognitive offloading versus
cognitive surrender.
The offloading is the positive scenario.
It's where we hand over tasks mindfully,
thoughtfully that AIS do really, really
well. supervise the results and stay in
the middle of the activity. Cognitive
surrender is in fact just surrendering
to these new magical capabilities and
turning over a lot of the work to them
without necessarily staying in there.
And it's sort of the road to humans
being put out to pasture by AIS. Now,
cognitive surrender doesn't look that
different from cognitive offloading.
It's it's subtle sometimes. It's not
like you're in a scene out of Pulp
Fiction where you're pointing a gun at
somebody. It's not that dramatic. So, we
have to be really thoughtful about how
we hand things to AIS and what our
mutual roles end up being.
In the four scenarios I just painted,
notice that none of them involve
somebody actually doing something wrong.
They were these are all matters of
convenience that slip in over time.
That's what makes this hard and makes it
a frontier rather than a scandal. The
question isn't who screwed up. The
question is, what are you going to do
now that you see these forces?
I've spent almost 30 years on the
offloading side of this line. I've been
curating this mind map, uh, taking ideas
from my head and putting them in a place
where other people can bump into them
for all this time where each link, each
decision I made, adding things, curating
this mind map is mine and mine alone. So
far,
from over here, this boundary doesn't
look scary. It looks like the most
interesting place in the world to stand.
In the next post, I'll take you across
and show you around.
Which of these scenarios have you
already spotted in your organization? Or
which are running around that you
haven't noticed? What other ones have
you seen? Uh, post in the comments. I
read them all.