Is OpenAI’s New “o1” Model The Big Step Forward We’ve Been Waiting For?

Channel: Alex Kantrowitz

Published at: 2024-09-13

YouTube video id: l-L3t5CmEYo

Source: https://www.youtube.com/watch?v=l-L3t5CmEYo

open aai has just released its reasoning
model it was called strawberry uh as a
code name but now it's being called 01
as as its public release uh and we're
going to talk all about what it is what
it means and where open Ai and the AI
race goes from here we have the perfect
guest host for us to do it Pary olon is
here she's a technology columnist with
Bloomberg and she's also the author of a
new book out this week it's called
Supremacy AI chat chpt and the race that
will change the world uh and what a week
to release a book because we're just
like here flooded with AI news py great
to see you welcome to the show thank you
it's wonderful to be here I'm a big fan
of the show so just so pleased to be
able to talk to you awesome well thank
you and uh thanks for coming on you know
I was like thinking okay Apple's going
to be the big news of the week so let's
get it out Wednesday and I think at this
point it's amazing Apple intelligence
and the iPhone 16 are kind of like a
distant memory at this point because of
what we've seen over the past couple of
days exactly so let's get to the big
news which is that we've been hearing
rumors that openai was going to release
this strawberry model which are it was
initially called qar and there were
worries about it uh around the Sam Alman
ouster and and it's here and basically
the big thing about this model is that
it does reasoning and so you can ask it
a question and you can actually kind of
see the way that it thinks through the
problem it goes through a bunch of
different steps and that's enabled it to
be much much better more accurate more
competent smarter than any previous
model so I'm like looking at uh some of
the charts that open AI put out with the
release uh I'm looking at accuracy so if
you think about uh competition math uh
GPT 40 was getting a 13.4 accuracy score
on competition math but this thing in
the way that it can think through the
the different problems is getting
83.3 uh score on accuracy uh with
competition math so it jumped from 13.4
to8
83.3 on compet on competition code it
goes from an 11 to an 89 and PhD level
uh Science question actually G GPT 40
wasn't bad at a 56 but 01 preview or o
sorry 01 gets a 78 and 01 preview which
is the model that's out now gets a
78.3 where an expert hum get human gets
uh 69.7
accuracy score I think that's just the
percent of questions they get accurate
that's the accuracy score Pary what do
you think about the release of these
models do they live up to the hype I
mean we basically saw a moment where you
know was Sam Alman pushed out of open AI
because they thought he wasn't being
careful about this type of model like
that was one of the assumptions are are
these a big step forward what do you
think about the release of these models
well first of all I think I would put
this into context of the huge amount of
pressure that open AI has been under for
the last year to get out something
that's going to cause as much awe as
chat GPT did in November 2022 and Dolly
before that um and I I thought it was
really interesting that in the last few
months we've been getting this sense
that hey open AI has been announcing a
lot of things but it hasn't actually
been launching the things that it's been
announcing so for example you know there
was the whole um Sora video thing yeah
where is that yeah where where is where
did that where was that even actually
even
gp40 the the the voice activated the
sort of voice um conversational um
interface I there were I I just met
people in San Francisco last week who
were like yeah I I still haven't been
able to try that yet it hasn't really
fully rolled out um yeah can I tell you
something about that so first of all
like no please I've been uh wondering
where it is cuz I've been signed up to
chat GPT premium for a while and hoped
that I would get that new voice
interface and I haven't gotten it yet
and I over the past so I'll just say
this over the past uh 24 hours I've been
monitoring the uh the chatter around the
strawberry and 01 release and my
favorite exchange is Sam Altman
announcing I don't know if you've seen
this he says roll out complete it's live
up of uh to 100% of chat GPT plus and
team users now this is of the new
reasoning model one of the first replies
is from this guy Matt Paul who goes when
are we getting the new voice features oh
yeah do the thing you said you were
going to do first yeah and Sam responds
hey how about a couple weeks of
gratitude for magic intelligence in the
sky and then you can have more toys soon
I mean oh my goodness it was interesting
but also just like so revealing of the
frustration I think that you're noting
of the fact that like we haven't seen
Sora release broadly we haven't seen the
voice features release broadly and I
think that's kind of the Trap that Sam
and his company has kind of put
themselves in by just releasing a lot of
demo Weare which is very exciting and
but if you start getting people too
excited for too long a time and then you
don't give them anything um then they
start to get a little bit frustrated and
a little bit cynical of what's actually
going on behind the curtain and you know
what it's funny what you just said that
you've been signed up to premium I have
done exactly the same thing because I
signed up to Claude and I'm paying for
two subscriptions to these things but
I'm keeping I I kept chat gbt plus
because I wanted to try out the voice
interface and it still hasn't shown up
on my phone either so here we go um so I
me it's it's funny because I also had I
was waiting for voice forever and it
didn't come so I unsubscribed for from
plus and my subscription ends today this
was not planned so I've been able to
actually try out uh GPT
uh1 uh and it's been pretty interesting
uh but still no voice but I guess 12
hours left before yeah exactly I've been
just you know there's like a monthly cap
there's a very small monthly cap or
something a weekly cap and I have like
not cared about rate limits I've been
like testing uh the crap out of this
thing because I'm like all right uh you
might as well yeah and can I I'll just
share like my initial reaction to it and
I'm curious what you think about it it
it feels like this thing is extremely
math and code based like that's where
the actual breakthroughs are going to be
uh like I said when I was reading these
benchmarks it's competition math
competition code PhD level science
questions that it pauses and by the way
when you're doing the queries you can
actually see it pause and think think
through each step use the word
think I do I do and you can you can
correct me on it but um you can see it
like go and and Tackle now I now I'm
getting nervous about thing but you can
see a tackle uh each different step and
there was actually like if you look at
the consumer preferences like the or the
human preferences like where they'd want
to use these new 01 preview models
versus gp40 they actually pre prefer GPT
40 uh for personal writing it's about
even on editing text but where GPT uh
this this 01 model really really thrives
is it's way it's much more preferred for
computer programming data analysis and
mathematical
calculation and that's why I think that
when you when you get into this like
people start to judge is this a step
forward the people who are using it for
more technical activities are really
going to see a jump and the people who
are using it for more stuff like I would
like writing or editing text aren't
going to see it so there might be a bias
among journalists to pan this thing
whereas like people in the technical
fields are going to be seeing that
there's there's real advances so so I'm
curious to get yeah I'm curious to get
your take on like uh on on this what do
you think about this like again going
back to that question yes we know open
AI is feeling the pressure but like is
how should we view this is this the
promised step forward well I think it's
a good strategy what you just described
that if this is a model that is perhaps
more useful for people who um are doing
data analysis they're scientists their
coders um then they've got a specific
models which just has a little bit more
of a clear utility for them whereas if
you're in marketing or you're in
customer service then you're more likely
to use the previous model something like
gbt 40 which is a lot more sort of
language oriented as opposed to
reasoning oriented I think that's no bad
thing in terms of business use case
because I think till now there's been
this kind
of um this effort to try and create
create this well we started off with
this ambition to create artificial
general intelligence right that was the
founding of open AI we want to create um
AI that has the same broad cognitive
capabilities is the human mind um and I
think that actually has a big downside
when you're actually trying to sell
something packaged like that to
businesses which is that you're giving
them this Swiss army knife of a tool
that has all these General capabilities
and the risk is that your end customer
can end up just becoming paralyzed with
indecision on what to do like if you
talk to some businesses who are trying
to figure out how to use generative AI
models I heard from One bank for
instance they asked they put out a
survey to their employees just saying
how should we use generative Ai and they
got more than a thousand responses back
from their employees which is great in
one way but in another way it actually
just makes everything take a bit longer
and in terms of figuring out a strategy
on how best to use it so I actually
think from what you're describing if the
models are starting to kind of fracture
a little bit and there's going to be a
sense going forward that you can use
this type of model for mathematical PA
based processes and data analysis and
then you can use this kind for language
processing maybe that kind of runs
counter to the original ideology of
let's build a broad Godlike General
level intelligent AI maybe it maybe it's
moving away from that but I think like
practically speaking that kind of makes
a bit more sense right so do you think
that Elia was let's say Elia was spooked
by these type of
capabilities I'm sensing the way that
you're talking about it that maybe that
was a little premature oh yeah I think
there's so much that comes out of open
AI that is just completely blown out of
proportion and wrong like for example
when Sam mman was fired there were a lot
of rumors going around that the board
members who fired him did so because
they were big time effective altruists
and they were just driven by this kind
of almost religious fundamentalist
fervor over this idea of being of sort
of being worried about the safety of
humanity under Ai and under under Sam
and and I've heard from people who
really know like who should know that
that is not the case at all um and yet
those rumor well the the the real reason
was that the board was concerned that
Sam was not telling the truth he was uh
he just was not a a reliable source of
information
and there wasn't just the board members
who felt that there were top Executives
who felt that as well right um so that's
such an odd anyway we don't have to
rehash the whole firing again but it's
such an odd reason to I mean yes I guess
you you don't want your chief executive
lying but like there wasn't like any
demonstrable like bad public bad
behavior and the company was crushing it
so I don't know am I giving him too too
easy of a pass maybe a little bit maybe
a little bit because there because the
concern was as I say it wasn't just from
the board members it was top Executives
too and there from people I spoke to but
IIA who led this kind of went back on it
and apologized so was not really such a
strongly held belief or maybe it was
what's your what's your take on it I
mean we don't know why people say what
they say publicly versus what they
actually think and what they actually
say behind closed doors is look at
what's at stake billions of dollars
incredible levels of prestige possible
Nobel prizes you know like yeah that
doesn't that doesn't breed a environment
for honesty yep so let's get back get
back to this um use cases of this stuff
I mean if so so one of the things I've
been thinking about as as 01 has
released or strawberry has released has
been okay so the models are smarter
where's the application I think you sort
of brought that up um yeah it's it's
like are we going to this is one of the
things that has been sort of plaguing
the AI industry even though it's doing
well obviously but it's been like okay
the models are getting smarter and they
do magical things but we still don't
really know how to use them in practice
right this is actually something that I
highlighted there was a grock engineer
who said who said um 01 seems powerful
uh so far but everyone is kind of unsure
what to do with it kind of humorous
place to be after all the
speculation so what do you think happens
as these models get smarter in fields
like science and coding and math it's I
think it's really interesting because
there's this race in Silicon Valley to
make everything more and more capable
which is disconnected I think a little
bit from the sentiment among their
Enterprise like businesses in the rest
of the world who are just like we don't
necessarily need AI That's more capable
we just like we're quite happy with what
you've already put out there like that
is already in pretty impressive and as a
step change up from what was even
available two years ago um so my sense
over the last year has been one thing
that's been lacking from the AI model
vendors like open AI like anthropic like
Google um Microsoft with Azure has been
just to do a little bit more handholding
with their business customers on how to
actually implement the current models
that are available but I think there's
just this kind of um you know rabid
enthusiasm to try and make these models
smarter and smarter we you've talked on
your show before about the arms race um
it's kind of had this it's it's got this
kind of self-perpetuating cycle that the
um you know even like Salesforce last
night just their their announcement got
completely overshadowed by open AI but
they announced agent force which is
basically what they say is the very
first AI platform for businesses that
has autonomous agents there are startups
that have been talking about doing this
but haven't released this and I haven't
seen anything similar I don't know if
you have but from from Google or open AI
or even anthropic people are talking
about it and it looks like Salesforce is
the first to do it so there's this real
um and they're very proud of that um and
so there's I'm just going to call
on that I don't think
Salesforce is going to be the leader in
AI agents or they even the Pioneer I've
heard people talk about things as AI
agents then you take a look at what it
is and it's like there's no agent to it
well and what does an agent even mean
right I mean when I was I was talking to
them about it and they were saying I was
saying well are you using the word bot
and agent interchangeably and they said
oh no no no Bots those are old that's
not agents are new that's that's the new
thing so a bot would be if you're
talking to a customer service chat bot
and it just kind of is based on a large
language model and uses uh predictive
technology to to say something that
sounds about right whereas an agent
can actually resolve your issue by
retrieving data about you the customer
from the Salesforce CRM and the data
Cloud which is one of their big services
that came out two years ago um and so it
means these things can actually take
action but I I'm I'm with you on that
Alex like I don't you know we've seen so
much of this in AI just a lot of a lot
of big announcements a lot of excitement
but then you see businesses actually
just struggling a little bit to to make
these things work for them and work well
yeah and so so I'm just going to read a
couple examples of folks who've been
trying out 01 and why this might be
different and and I think it's your
right to point out the difference
between like chat and Bots um or Bots
and agents because this is like from
somebody who works at this company
called Spellbook legal his name is Scott
Stephenson it's very interesting that he
says he says oan is extremely good at
long document manipulation Docker vision
is one of the hardest problems we've
dealt with dealt with uh like for
example taking a long set of
instructions and modifying a legal
document and 01 is really good at this
when people are overwhelmed by 01 I
think it's because they're thinking of
it as chat still and it's ability to do
work is going to be really good uh once
function calling is fully supported so
you know I think that like with these
Bots it is definitely like you know we
talked about handholding and you know
maybe these companies need to do a
little bit more handholding but I think
often in technology what happens is they
put the capabilities out there and then
people like Scott Stevenson right figure
out um the way to do it because they're
just using it more often and that's why
they're putting this out in preview and
that's the the sort of use cases that
get spread by word of mouth or
productized uh and maybe that's what
happens and I would say that would be
the argument to keep just building these
these oh yeah uh models uh make these
making these models smarter and and
maybe the thing is that they've gotten
so fast so smart so fast that they've
sort of outpaced the utility because we
haven't had enough time or the industry
hasn't had enough time to put their
hands on and try things like this what
do you think and actually digest it yeah
no I think that's that's a very very
valid point um and I think you know
coming for so I'm based in London and I
hear a lot more I'm not not being based
in San Francisco I'm not in the Silicon
Valley bubble so being in London I hear
a lot more of kind of the rhetoric from
people on the east coast of the US from
Wall Street from investors from hedge
funds from healthc care companies um
legal firms and there there's there's a
real sense of skepticism over the last
few months um which I know you you've
you've talked about and but I think it's
not entirely warranted like I think like
this Market correction that's happened
in the last few months was healthy like
it was it was crazy that Nidia became
worth $3 trillion at one point in market
capitalization um and I think that when
people talk about hype I think so much
of the hype is about timelines there's a
sense that um businesses and investors
kind of got this impression that AI was
going to start bringing a return on
investment for companies sooner rather
than later like I was talking to one CEO
of an AI vendor a few months ago and I
was saying so how are you going to
measure how how are we going to measure
success for generative AI in terms of
return on investment for businesses and
you know what he said he said in the
next two or three quarters you're just
going to see more Fortune 500 companies
um report an increase in ebit dah
numbers like they're just going to have
higher profits I I just think that's
ludicrous like really like in the next
two quarters more businesses are just
going to make more money because of AI
and that's how it's GNA take time it's
gonna take way more time than that um
that's been the case with every major
Tech Revolution whether it was search or
um desktop or mobile um you we all know
about the Gartner hype cycle and um
these things just take time to implement
so yes if if the AI vendors in Silicon
Valley is racing ahead on capabilities
then sure maybe the businesses just need
to catch up and so with reasoning uh
reasoning a lot of people have talked
about how it lays the groundwork for
these agents because again like seeing
it it's pretty amazing watching using
these models and then watching it go
through the multi-step process um and
the idea is like all right if you're
going to have an agent that's going to
go and take action for you on your
behalf it's going to need to be able to
like process step by step in order to do
that and this is the way that open AI is
describing it so this is from their head
of research Bob MCU in in uh in um The
Verge we have been spending many months
working on reasoning because we think
this is actually the critical
breakthrough fundamentally this is a new
mod mod ality for models in order to be
able to solve the really hard problems
that it takes in order to progress
toward humanlike levels of intelligence
so I'm curious like do you do you think
this is the step a step toward uh agents
and why why is reasoning this sort of as
he describes it the fundamental Or the
critical breakthrough well I thought I
think that what you were just you
earlier you described the tweet from
someone and what they were describing in
terms of when they played around with
um gbt this new right having to do the
work yes it made it sound like there was
some gray area here where it had almost
agentic properties um why is reasoning
important well I think this just gives
it more broad broader capability and I
think it's really interesting that um
open aai hasn't actually used the word
agent once in its announcement and it
doesn't sound like any of its scientists
have talked about that either um I know
Bloomberg have this like you didn't
Bloomberg publish the um open AI like uh
imagining its future and it has like
five levels and level one is chatbots
level two is reasoning level three is
Agents level four is the AI innovates
and then level five is the AI does work
at the level of an organization so the
the suggestion would be that we're at
level two and that agent level is what's
next yeah yeah um and Al I thought it
was really interesting in in open AI um
see I think right now it's just a matter
of it's like you said they they put the
technology out there and they don't
really know necessarily how it's going
to be used it's sort of the
experimentation is done by everyone else
out in in the world in among businesses
and organizations they did give some
suggestions in their blog post so they
said um I'm just reading from one
section which I thought was the most
interesting was that 01 can be used by
Healthcare researchers to annotate cell
sequencing data by phys is to generate
complicated mathematical formulas needed
for Quantum Optics I'm going to stop
there those are really really specific
things technical very specific and then
they end it with this bit then they say
and by developers in all fields to build
and execute multi-step workflows that's
like anything basically so they go from
they basically that guy who told you
that the um you're going to see this in
IA in the next few cours maybe they're
right I don't know think he was right no
probably not but I can see why he thinks
that right because yeah again this is
going to come to like the very like um
uh labor intensive very value high value
tasks in our economy that's what it's
going to do yeah so but don't you but
don't you think also like already
there's been a lot of concern among
businesses about hallucinations right
from these models now when you give them
the capability of taking action so again
sorry to go back to the Salesforce but
it's just because I have a concrete
example they gave me let which which was
this agent agent in their um in the
customer service chatbot um which could
retrieve data and basically resolve a
ticket in the same way that a customer
service representative would do um
that's putting even more faith in these
models like it's one thing to make a
mistake about information but to
actually screw up the process of you
know a uh resolving someone's complaint
or getting some information for them um
that's a bit more high stakes so I'm I
to be honest I have no idea how this
will play out um and I think businesses
are going to be very cautious just
because there has been so much concern
about hallucinations till now right yeah
and open a even in its blog post says
that it does not solve this does not
solve hallucinations and then you can
um you can did they say that in the
latest did interesting you could also go
to um some of the critics they've
they're having a field day with this
here's Eric sorry Ed zitron uh we have
our big stupid magic trick folks sty
strawberry it's open ai's new model it
will take uh 10 to 20 seconds to give
slightly better answers and it's buggy
open AI is getting desperate and this is
a real deal Pale
Horse yeah I think that's a little bit
over the top what do you think it has to
be I mean it's that's said that's that's
thing but I spoke to someone last night
who had tried it as well I haven't had a
chance to try it myself um and they said
they didn't and this was someone who
tries these models a lot um and they
said they they couldn't really see a big
upgrade or a major step change um or
even a particularly big difference
between what CLA anthropics Claude can
do um but again I I I really need to try
this out myself and and I think it's
quite hard to kind of
put a judgment they literally announced
it last night um so we're we can all
test it we can all play around with it
but I think it will really will really
get a sense of how useful this is when
when people start using it for
work yes okay so um let's end this one
just with a unfair question buying or
selling the hype on on strawberry or 01
I'll go first I'm buying it I'm buying
it I think it's going to be more
difficult for the quote unquote word
Cells versus the shape rotators right
the people who deal in text versus the
people who deal in numbers but I think
this was sort of a necessary step and
you see The Benchmark improvements here
and it's not going to look immediately
just like this you know the way that um
some of the other let let's say code uh
completion you initially was like okay
so it can complete some code and now
it's starting to really look like it can
code things up on its own I think that
this is going to be something that folks
in the technical fields are going to use
um and and it will actually be a
significant step up okay how about you
buying or selling I would say sell only
because I haven't tried it yet and also
because I want to be a bit contrarian um
to you because you said buy so um no I
mean it's that's a hard one to answer I
guess I'll just say sell for now um
because this thing about the benchmarks
I I personally don't fully understand
the benchmarks and what they're weighted
against and what they're sourced against
and you know is are these completely
standardized across the industry um
those are really impressive impressive
numbers but what do they actually mean
and again I I think it's very hard to
really judge this until you try it right
and I'll just give one more piece of
evidence uh on my side I know this is
unfair but um I I just saw a thread with
a stream of folks who've like tested uh
01 on their personal benchmarks like a
lot of people have their like own
personal tests M uh to of AI and it's
it's just like person after person
talking about how much better it is uh
than really their previous tests okay
okay but the the other thing here is is
really like maybe yeah I don't know like
this is preview of course and yeah
that's right you you you mentioned it
and and you're totally right that we've
seen Sora uh uh which has been a lot of
hype and not released yet and we've seen
the voice uh GPT 40 voice application
and the the status of that is is still
you know up in the air and you know
should we uh give open AI a couple weeks
of gratitude for the magic intelligence
in the sky and have our more of our toys
soon I don't think so I'll just have
have in in spite of everything I've said
there is an argument that actually maybe
behind the scenes Sam mman and open AI
have just realized like they need to
focus on one thing that really matters
and if I was going to look at those
three things like the voice interface
Sora and reasoning capabilities hands
down the bigger deal is reasoning
capabilities so maybe what's happening
behind the scenes is they've just put a
pause on these two other things which
aren't that important and the reason I
say that is because last week I talked
to some people in the valley and a
couple of times it came up that there
are the there's growing content about
Sam maltman's General lack of focus and
people are saying he's disorganized and
they're trying too many things and I
spoke to one VC who even said that he
was going to meet Sam mman that weekend
and I said if you could give him one
piece of advice what would you tell him
and he said I would tell him that he
needs to look at everything he's doing
and he needs to pick two things and just
stick with those two things and he said
even if one of those things is this big
chip infrastructure project that he's
that's fine do that but then just do one
other thing and nothing else because
there I think there is a concern that
open AI has just got too many pans on
the fire and um so so yeah so I would
say yeah I wouldn't discount strawberry
and sl01 just in case actually maybe
what's going on here is that they've
realized they need to just focus on the
higher Stakes uh bet that counts for
more
yeah Pary you're convincing me here
because uh there there have been sort of
Rumblings and rumors going around that
open AI is lost it and too many people
left look El is gone he just raised a
billion and John Schulman's gone and and
he's at anthropic now and you know the
the rest of the leadership team has
thinned out and there's been staff
Exodus and open AI ruined itself and I
think we can say definitively well maybe
not defin but we can say with good
confidence now that the the company is
still pushing the envelope and I was
just on H the chatbot um uh chatbot
Arena where they battle different
chatbots against each other and you vote
which gives the best output and GPT 40
is still the winner there so oh really
more so than Claud uh oh yeah it's it's
definitely above it's actually Gemini is
second and this is where people vote on
what's best so now human manipulation
could be responsible for some of it but
I do think that the of open AI has been
uh prematurely reported that being said
there's a lot of competition you know we
can we can agree on that but they're not
dead yet yeah
yeah and they also want some more money
so it brings us to our next story this
is actually coming from your outlet uh
Bloomberg has a report out that open AI
fundraising is going to put the
valuation of the company at 150 billion
and a couple weeks ago we just did a
show talking about how it was going to
be 100 billion and that's crazy uh but
uh they are according to Bloomberg in
talks to raise 6.5 billion from
investors and another five billion from
debt uh in debt from Banks so they're
going to basically bring in 11 billion
uh dollars and it's the round like we
talked about in this on the show in the
past uh Thrive Capital Microsoft Nvidia
and apple of are all talking about
getting involved so let me ask you this
you you obviously just wrote the book
about these companies um what's the deal
with taking all this money and getting
this massive valuation and still being a
nonprofit how's that even possible but
they're not a nonprofit right they're
capped profit and one of the reports
that came out a few weeks ago there was
a report saying that there's some
interest among um the management of open
AI to restructure the
company in such a way that it could
become even more investor friendly so go
even further in the direction of being a
for profit
organization um as so the only way that
it's really a nonprofit is that it's it
has this nonprofit board which has a
significant amount of power that is very
unusual for a company for example it had
the power to fire the CEO or did it it
didn't really right because he came back
in the end yeah um so I mean that's just
what I kind of tried to make the point I
wanted to say in my book was that I was
just so fascinated when chat gbt came
out that two people Sam Alman and Demis
hbis had become like these Rivals to
create artificial general intelligence
and they both started out with really
humanitarian goals like really
altruistic goals like curing cancer
solving climate change Sam mman wanted
to you know improve the wealth of people
um across the PO across the population
um but it was just so hard to keep those
goals in check when money became such a
focal point because building AI was so
expensive and um and they just ended up
getting sucked into the gravitational
pull of big tech companies I mean that's
just there's no two ways about it h but
what I really wanted to highlight was
that both of them knew from the start
that they needed to protect this this
technology and how it was used and that
it wasn't misused so they tried to
that's why s that's why Sam tried to
start open AI as a nonprofit and that's
why Demis tried to break away from
Google and also become something like a
nonprofit he spent years trying to do
that and spent a lot of money and time
trying to do that and if they both
failed to create these structures of
responsible governance like if I'd had
that's really what my book is about is
govern governance I would never use that
title because no one would buy the book
but that's
just gave away the game here
so but yeah that's basically what this
is about these companies are now you
know they the tech is now controlled by
a handful of very large big Tech organiz
uh companies um in a regulatory vacuum
there's no proper oversight and there
were you know people tried to put that
in place we as a society have not
figured out how to balance Innovation
with um with proper oversight so what do
you think all this money is going to go
toward then I mean are they going to
even be able to justify that valuation
they're losing money a lot of money now
I mean what's the business case for
investing in
them well I mean they get a revenue
share with Microsoft right from um
through Azure so if Azure does really
well then micros then sorry then open
aai can do quite well they've got the
subscription business which is actually
pretty steady they've got something like
200 million active users chat gbt so you
know I mean it's it's not a bad business
long you could say longterm that can
grow to something quite impressive but
these things take time right I mean with
Amazon it took many years for it to
become a profitable business and maybe
people are looking at open AI in the
same way the problem is that as you well
know uh building and improving
Foundation models is incredibly
expensive and it's not just the compute
right it's the salaries like these
senior AI scientists and Engineers how
much are theying well okay so I spoke to
someone about this who uh from an
organization that was trying to
hire um you that that just was asking
questions about salaries and so one kind
of mid-tier AI startup was asked so how
much would you pay for like a senior AI
kind of executive and they said it's
basically around um over the course of
four years $6
million um I've heard elsewhere you know
two to three million doar a year
uh salary options over time yeah yeah
and so one way they're trying to make
the money back is potentially raising
the subscription fees for these Bots I
mean they they don't make any money on
the Bots right now um but one report
that came out of I think the information
this week is that um let's see it says
in internal discussions at open aai
subscription prices ranging up to $2,000
per month were on the table uh so
basically the thought is maybe that
these new reasoning models again the
ones that were released this week
preview and mini so there's like an
actual model that open AI has that you
know I think will eventually make its
way into the product and if the models
do live up to this sort of um improve
benchmarks and become super valuable for
some cases maybe they can charge much
more is that one way that they end up
justifying the
valuation maybe but I can't see it being
that high pry High totally can totally
see big numbers being tossed around
though and I'll give you why the reason
why I think that is I think it was last
year I was talking to somebody who was
from an an AI company that worked very
closely with open Ai and they they
talked to people at open AI all the time
and they were like oh yeah so Sam Sam
and open AI they're doing another
funding round and he wants to raise a
hundred billion dollar he was like
they're saying that's what they're
saying in the company he's trying and I
was I remember responding you mean a
hundred billion valuation right and he
said no no no they actually want to
raise a 100 billion um I don't I could
never corroborate that with anyone but
my sense is that big numbers get tossed
around a lot right like trilon among
yeah yeah exactly exactly that's a great
example um the whole s trillion dollar
chip thing was it 7 trillion yeah it was
a 7 trillion yeah yeah raising um so a
lot of huge ridiculously galactic Iz
numbers get tossed around when people
talk about open Ai and probably inside
open AI too um but you know
fundamentally to your point do they need
to raise the price yes they probably do
I mean look at us we've both been paying
for both Claude and chat gbt um because
it's sort of affordable um but maybe
they they can afford to raise the price
if they put in these capabilities where
these systems become integral to your
work process then yeah maybe you'll just
have to pay more and maybe they can get
away with it is uh is one reason why
companies like Nvidia and apple and uh
Microsoft might be involved in this
round is just because like even if they
don't make their money back if openai is
able to keep pushing the envelope then
their valuation will just go up like
it's a smart investment like a almost
like there was a moment where opening
eyes said if you're going to invest in
US you guys should think of it as a
donation and maybe in in these companies
cases they that's fine because it is
it's going to work out for them if it
works out and also they can afford they
can afford to do it right like it's not
like apple has just been told like they
had to put 13 billion doar in an escrow
account to pay this Irish tax fine from
the European Union and that's even
that's like it sounds like a huge number
but really it's just pocket change for
them so an invest like an investment in
open AI I think the company that's
completely on the Forefront of AI right
now and it literally Sparks the gender
of AI boom almost two years ago I think
is worth it for them yep okay one more
question for you about funding so every
now and again you get these Rumblings
that open AI might be trying to raise
from uh like basically Gulf Oil States
I'm going to get it wrong Qatar Saudi
Arabia but you know of the of that
nature um it just kind of seems
interesting to me that it's going to
these states and like uh companies that
initially started with this mission of
AI being beneficial for the world being
like you know really need to do like
give an oil you know an oil oligarchy or
uh I don't know authoritarian type
country a stake in what we're doing what
do you think about that well let's uh
align our or do we align ourselves with
the oil oligarchy in the Middle East or
the tech corporate oligarchy in the
west what to choose I would go with the
tech ones but
anyway yeah it depends which side of the
uh which part of the world you're in and
what on what your worldview would be on
that I mean there's so much money out
there right they've got Sovereign wealth
funds which means they can make these
huge bets on risky Investments and um
Dubai and the UAE there's been a lot of
there's been a big push to try and um
diversify away from their very very big
dependency on oil and try to move
towards more more of a tech-based
economy um so it absolutely makes sense
for them to be putting money into
something like this right and you've
been following these companies pretty
closely obviously writing the book about
them um is there going to be a point
where they just kind of run out of
Runway like is this is this
sustainable is what
sustainable this all the money being
spent on the training of the models in
AI I mean it just it's so Capital
intensive um like eventually and and
you're like we were talking earlier
about how it's going to take some time
to figure out the use cases even though
there probably are use cases there but
like how much time does this have before
it has to show results that return an in
on the investment maybe a few I'd say a
few years and I think there isn't as
much pressure on the big tech companies
like I I I wouldn't it's very tempting
to just say it's not sustainable because
it's so expensive and it's moving so
fast and there's no obvious use cases
but you have to remember these these
companies have billions of dollars on
their balance sheets they they can
totally afford this and they can afford
to keep spending like this for the next
few years at least um I mean look how
much money Mark Zuckerberg spent on the
metaverse and that really went nowhere
um and he's still like meta's still
doing fine in fact meta as a stock has
outperformed all the other big tech
companies in the last I think since the
start of the year thanks to its
advertising and still called meta by the
way so maybe they haven't fully given up
but yes he's in it for the long ter for
the Long Haul um yeah I I I think think
they they I think they will continue to
spend quite big on this um for the next
few years and because they you know
although they want to they do need to
see their Enterprise like again I'm
thinking from the hyperscalers
perspective but you look you talk to
someone like Google and they have I like
I literally asked someone from their
Cloud team last um last week um you know
what what their inbound interest was for
um from Enterprise customers for Gemini
and for their for their API um and they
were just couldn't they couldn't give me
a number but they were like it's so it's
just crazy the amount of um usage that
we have and I was like has it has it
doubled in the last year doubled would
be an understatement like the rate of of
of usage increase it has really gone up
so they're doing fine but the the
question is just like um whether
businesses will continue spending on
these models um we can go into this if
you want but this is probably for a
whole other episode like how do you how
do you measure return on investment
right it's not simple and it's
absolutely not and I think people are
starting to accept that they're starting
to see buying AI Services as a bit like
having email like it's hard to measure
as a business how much having email
actually makes you money but if you took
email away from all your employees your
business would just fall apart it's
absolutely integral to operating um and
so I think there's a sense that AI will
become like that it will become a really
important component of making a business
more productive but it will be really
hard to measure exactly how it does that
in numbers yeah but the thing is most
email programs are free and this is very
expensive to produce and run but I think
that's a that's a qu we'll we'll cross
that bridge when we come to it but I do
want to hear a little bit more about um
sort of your perspective on Sam and
Demis and if there's stuff that the
public uh don't don't know about them
that they should know and then I also
want to talk about this story about how
uh some local news sites are using
people uh who are who've turned
themselves into AI avatars uh uh to read
the news and like do local news programs
so why don't we do that right after this
and we're back here yeah okay great
sorry um everything's good we have a
couple more
minutes oh you're asking me right yeah
yeah
yeah yeah yeah in terms of my time yes
I've got I've definitely got time okay
great okay and we're back here on big
technology podcast with Pary Olsen she's
the author of Supremacy AI chat GPT and
the race that will change the world it's
out this week okay so Pary I'm curious
like you I think a lot of us have a
perception of Sam Alman um maybe many
but somewhat fewer have a perception of
damis aabis who's the head of Deep Mind
at Google what do you think uh what did
you learn about the two of them that is
not fully represented in the public
Consciousness that is important to
know well I what I learned from both of
them is that they both were incredibly
Mission oriented people um they had very
different personalities Sam was a very
outspoken um charismatic individual he
was an entrepreneurs Guru um he even
after his company his first start up
looped essentially failed by Silicon
Valley standards he somehow managed to
create this incredible respectability
around himself among um startups and
venture capitalist in Silicon Valley as
this almost Yoda of um of of
entrepreneurial advice he would um post
blog posts with just like 99 piece of
pieces of advice for for startups and
he's just an incredibly good
communicator and someone who would also
at the drop of a hat help other
entrepreneurs so i' I've spoken to
entrepreneurs who said yeah I just you
know sent him an email I didn't think
he'd respond and then he responded right
away and introduced me to someone who
helped me raise money and so he has
really engendered a ton of Goodwill
among Stardust because he's very
responsive like that um and it's
actually given him an incredibly
powerful position in the valley just as
a as someone who's incredibly well
connected and that was even before he
started opening in AI when he was the
head of white combinator Demus is you me
deep
mind oh no Sam Sam you're talking about
yes yeah I'm talking about Sam I'm
talking about Sam um Demus on the other
side of the ocean very different kind of
personality um he started off as a he
was a chess Prodigy growing up um one of
the best chess players in the world when
he was only 13 and he was obsessed with
games he I don't know if you ever played
the G the video game theme park
but I know that's demises he did as a
teenager or something he did it when he
was 17 he C co-designed it with Peter
molu at um bullfrog Productions and then
he went to Cambridge really young and
and started his own video game um
company um which in the same way that
Sam's first startup failed deus's first
startup also failed and both
entrepreneurs really learned a lot from
those failures and both kind of came out
of that with this goal of wanting to to
build artificial general intelligence
Demis did it first he was actually the
first to kind of put his neck on the
line as a um among the scientific
Community to do something that was
really a fringe Theory um back in the in
2010 and start a company that actually
wanted to build human level AI it was
seen as very crazy and nobody would put
money I actually spoke to an investor
yesterday who was one of the VCS who
didn't back Deep Mind and now they
absolutely are kicking themselves for it
now um but it was just they had no
business model and the only way they
could make money was um raise money was
to come to Silicon Valley where people
are selling the future and they raised
money from Elon Musk and Peter teal and
then of course they they they buil deep
mind I mean Demus I think is like a
really fascinating character he's
absolutely obsessed with games and he's
also really really good at winning games
if you ever play him at foosball he will
beat you he's just like very very good
um or or or soccer um and I think maybe
that's also a trait he shares with Sam
that they they both um you know are very
competitive and want to win and Demis
also has an incredible um almost
obsession with um winning a Nobel Prize
this is what I've heard from people who
worked with him that that was kind of
the goals that he put for the company um
but I think both both men you know
played in in my mind I wanted to kind of
hang the story on these two guys there's
so many other there's a huge cast of
characters in the field of AI but I
don't like to tell a story about AI
governance sometimes you just need to
boil it down to a couple people to make
it just a little bit clearer and more
simple um and so that's why I focused on
just those two okay One More Story I
want to get to before we head out which
is um I was looking up your interviews
on YouTube and saw you like stand before
the camera and deliver this monologue
and as I'm watching I realized oh that's
not Pary it's AI Pary and Jesus you're
the perfect person to speak about this
because this week I read a wild story uh
in the in wired about a Hawaiian
journalist who worked at um this paper
called The Garden Island and basically
it is a print type not print it's a
words news site and it's covering this
island of you know a couple thousand
people maybe 10 no maybe like 70 th
people or so and the news site all of a
sudden started to do news video and who
are the anchors of this these News
videos they're actually AI avatars that
will basically take news from the site
and then read them to you and you can
watch the video as if it's like a
regular uh bit of news so just thinking
about the direction of AI we were like
talking earlier about like the ways that
it might be applied what do you think
about the fact that um you know we're
starting to see effectively news anchors
uh being I don't wouldn't call it
replaced yet but being mimicked by AI
avatars like the one that you uh became
in your video and actually reading real
news to real people on news sites like
the Garden Island I think in a nutshell
I think our future is just going to be
so noisy we're going to have so so much
information in so many different forms
it's going to be not just text we're
going to have so much more video we're
going to have so many more images
because AI is just going to be
generating so much of it um and that's
come like whenever I talk to businesses
about how they're using generative AI
they always always say we're not
replacing jobs we're we're we're not
replacing humans we're augmenting them
which I think like it sounds a bit just
disingenuous and that's just their way
of protecting themselves reputationally
but it's kind of true like I think we
are in the case of the newscasters like
it's not replacing newscasters right
it's just creating more videos for
people to watch potentially of news so
that they don't necessarily read it on
text it's just another way of consuming
news and that's what I see potentially
happening with generative AI is that
well I I I genuinely cannot see major TV
news networks replacing their anchors
with generative AI anchors anytime soon
you know if ever but I certainly could
see you know local radio stations local
webs not even local TV stations but like
really small Outlets that just don't do
video now here's a chance for them to do
video and so we're just going to have
lots lots more video content available
to us through all sorts of different
platforms yeah it's amazing I was like
thinking about it for a big technology
like what would have happened and
obviously the technolog is not there yet
but like what would have happened if we
would have fed some AI the news stories
about strawberry and and the open AI
evaluation and then written like one
sentence sort of perspectives from each
of our side and then fed the AI your
book and then say create a podcast
episode I mean that stuff is crazy now I
would never do it I think that our stuff
is better than what you would get from
AI generated slop but I yesterday I I
spoke with an editor who had his body
scanned and um uses this video uh as an
added added feature on his website and
it was amazing he was like first of all
he was like working with the software
company and was like my accent is uh is
and my English is um not as good as your
technology is making it out to be so I
wanted it to sound like his quote was
dirtier you know I think he meant
rougher not like I don't think he wanted
it with swear words and stuff um and and
they did it and it sounds a lot and
looks a lot like him and they're getting
thousands of views on each of these
videos they're getting thousands of
views really see this is what I don't
get is do people actually like this and
I think what is going to happen I mean
you said thousands of views right so the
more people see AI generated content the
more they get used to it and the more
they get comfortable with it so even
though avat video avatars of news
anchors and this gentleman you mentioned
probably look a little bit on County
Valley actually if you watch that kind
of content for long enough you start to
just be okay with it and then maybe like
when it comes up in your Tik Tok feed or
Instagram or you find it on a website
you just you click on it because you're
okay with it you're used to it exactly
oh god well everybody this has been a
real discussion between real me and I
hope that this is real you Pary is this
real give a sign of life I don't know I
couldn't tell th there's a thumbs up all
right right your avatar couldn't do that
cuz it's just yes see there we go okay
so um but I I uh am so glad we got a
chance to speak and I do encourage
people to pick up the book it's called
Supremacy AI chat GPT and the race that
will change the world in man Supremacy
it's the right the right title for it
right oh thank you yeah uh well
congratulations on the release par me
great speaking with you thanks again for
coming on oh likewise this was this was
really fun thank you and it was for me
as well all right everybody thanks for
listening on Wednesday I'll have an
interview with the CEO of LinkedIn Ryan
roslansky so stay tuned for that yes
we're talking about AI of course we are
there Microsoft subsidiary uh and uh and
then I'll be back on Friday with Ronan
breaking down next week's news thanks
again for listening and we'll see you
next time on big technology podcast