199 | “OpenAI Files” bombshell, Google & Microsoft fight while Meta poaches talent, one-person startup scores $80M, Veo-3 in YouTube Shorts, and more crucial AI news for the week ending on June 20, 2025 - Leveraging AI

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Hello, and welcome to a WeekendNews episode of the Leveraging

(00:03):
AI Podcast, the podcast thatshares practical, ethical ways
to improve efficiency, grow yourbusiness, and advance your
career.
This is Isar Metis, your host,and this week our deep dive
topics are going to start withthe intensifying battle at the
top of the AI race with majormoves made by the leading labs,
including the potential finalsteps of the unraveling of the
relationships between Microsoftand open ai.

(00:25):
We're also going to talk abouthow a single person can now
create a very successfulsoftware company, and we're
gonna end with talking aboutsome negative, broader impacts
of AI on society and people.
But then we have a long andexciting list of rapid fire
items and also a lot of stuffthat did not make it in because
there's too much to talk aboutand it's gonna be available in

(00:45):
our newsletter.
But there are a lot of reallyexciting topics to talk about in
the rapid fire as well.
So let's go.
and wE'll start this episodewith talking about the open AI
files.
It is a report launched by MidasProject and Tech Oversight
Project, both our companies whoare watchdogs over tech

(01:07):
companies, and these filescatalog a very long list of
documents from the past and thepresent of Open ai highlighting
serious concerns on governanceflaws, leadership concerns and
cultural issues at OpenAI, andthe change they have been
through from being a non-profitfor humanity organization to a
for-profit grow at all cost kindof organization that we know

(01:29):
today.
And they highlight very specificdramatic changes that happened
in open AI in the past fewyears.
They share multiple documentsof, and now I'm quoting rushed
safety evaluation processes andculture of recklessness at
OpenAI.
And they're talking about therisks of these kind of behaviors
and culture when racing towardsa GI.

(01:49):
One of the stronger quotes froma leadership perspective that is
quoted in these files actuallycome from former chief scientist
Ilia Susko, now the CEO and thefounder of Safe Super
Intelligence.
And he said, and I'm quoting, Idon't think Sam is the guy who
should have the finger on thebutton for a GI.
That's a very strong sentencefrom somebody who's a co-founder
who knows Sam and OpenAI very,very well.

(02:10):
They also list a lot ofconflicts of interest or
presumably conflicts of interestwhile Sam Altman's personal
investment portfolio thatoverlap and sometimes compete
with OpenAI business, suggestingethical conflicts that he may
have in making decisions inOpenAI.
They also mentioned the physicalaspect of massive data centers
that are linked to power outagesand riding electricity costs,
reflecting and showing thegrowth at all costs kind of

(02:32):
mindset in the company.
And the goal of these files isto basically raise the level of
knowledge that we have on what'sactually going on.
And they are advocating forresponsible governance, ethical
leadership, and shared benefitson the race to a GI.
They seem to be picking onOpenAI specifically, but if you
broaden that, I'm sure there'svery, very similar aspects in

(02:52):
all the leading labs.
And what this just gives us is awell documented example that
deep dives into multiple aspectsof how the labs are run at this
time and age.
A big part of the report showskinda like the before and after
of OpenAI, where they're showingpast quotes from OpenAI
leadership and then the currentquotes.
So I'll give you a few examples.
You can go out and check thereport yourself, but example

(03:12):
number one, past.
That's why we are a nonprofit.
We don't ever want to be makingdecisions to benefit
shareholders.
The only people we want to beaccountable to is humanity as a
whole.
That is from Sam Altman on March27th of 2017.
So the early days of OpenAI.
Another pass quote from March11th, 2019, we've designed

(03:34):
OpenAI LP to put our overallmission, ensuring the creation
and adoption of safe andbeneficial a GI ahead of
generating returns forinvestors.
Regardless of how the worldevolves, we are committed
legally and personally to ourmission.
So this is how OpenAI wasestablished.
Now, let's see the currentstatement.
Our plan is to transform ourexisting for-profit into a

(03:55):
Delaware public benefitcorporation with ordinary shares
of stock.
The PBC is a structure used bymany others that require the
company to balance shareholders,interest, stakeholders interest,
and a public benefit interest inits decision making.
So going away from our missionis humanity and not getting any
pressure from shareholders tosomething that balances between
them.
And it's very clear where thebalance lies.

(04:18):
So while again, this is pickingspecifically on open AI and
their transition from non-profitto for-profit, I'm sure the same
kind of conversation can be hadaround Google meta philanthropic
and every other company in theworld right now, including the
open source models, includingthe Chinese models and so on.
The race is on, everybody'srunning as fast as they can and
nothing else matters.

(04:38):
And that obviously puts us allat risk across multiple aspects,
from social aspects to politicalaspects to what the hell is true
anymore aspects like there's ahuge variety of aspects of this
and obviously the concentrationof power and money that will
have dramatic changes, as Imentioned in both society and
economy and all these things aregonna unravel in the next few
years in a very aggressive wayand gonna change a lot of the

(05:01):
things that we take for grantedright now.
And so I think this file, whileagain it is picking up on open
ai, it is important for us togeneralize and understand where
everything is going right now,but staying on open AI and their
relationship or starting to seethe lack of relationship with
microsoft, there's been multiplesteps in that direction over the
last two years, and it's gettingmore and more extreme.

(05:21):
Let's start with a quick recap.
In 2019, Microsoft invest in$1billion in OpenAI and basically
kept the company alive.
They have since then invested atotal of$13 billion in OpenAI,
and now recent moves by OpenAIis pushing this relationship
further and further apart.
one of the recent moves is the$3billion acquisition of Windsurf,
which is pushing OpenAI tobecome a direct competitor of

(05:45):
another aspect of Microsoft.
In This particular case, it'sgoing directly after GitHub
copilot and all the GitHub AItools, as well as Azure AI
Foundry in some of its aspects.
So it's a, it's a direct hit atone of Microsoft's core
businesses.
Combine that with the fact thatOpenAI is in the process of
transitioning to a publicbenefit corporation, as we just
said, which means they'reshifting away from the structure

(06:07):
of the company that Microsoftactually invested in.
They have some disputes overintellectual property, cloud
usage, equity stakes, andpotential antitrust actions.
So the whole thing seems to beon a very shaky ground right
now.
To tell you how extreme thingsare, OpenAI is considering
accusing Microsoft ofanti-competitive behavior to
protect winds surf's IP from theeyes of MicrosoFt.

(06:29):
and they're basicallythreatening Microsoft with
federal regulatory review oftheir contracts.
Now the transition to the PublicBenefit Corporation requires
Microsoft approval, which theyhaven't received yet, and that
puts at risk a lot of otherinvestments that they received
since the biggest one.
Obviously, being from SoftBank,OpenAI is currently offering
Microsoft a 33% stake in therestructured company in exchange

(06:53):
for forgiving the future profitshares that they were supposed
to give them, which I don'tbelieve Microsoft will be
willing to accept at this poInt.
Now as we shared last week,OpenAI just announced a deal to
start using Google Cloud, adirect competitor of Microsoft.
That's in addition to theirprevious arrangements with
Oracle and SoftBank.
So reducing the reliance on onMicrosoft Azure and providing
additional options for AI togrow beyond just a Microsoft

(07:16):
relationship life.
All of these things are notenough.
The information, which is areliable source of everything
about AI and specifically aboutOpenAI, just announced that
OpenAI is offering a veryaggressive 10 to 20% discount on
enterprise Chat GPTsubscriptions, Which is going
indirect competition toMicrosoft sales of their copilot
licenses.
open AI's internal projectionsare aiming for$15 billion in

(07:38):
enterprise revenue by 2030.
And guess who else is trying toget these numbers from AI
distribution on enterpriselicenses.
Many of my clients, both on theAI education side as well as on
the AI consulting side, and Iwork with companies across the
board from multiple sizes andmultiple industries.
Many of them are Microsoft basedorganizations and most of them

(07:59):
are actually using open AI'senterprise licenses versus the
Microsoft copilot licensesbecause they're just better
right now.
But in addition to being better,they're turning to be cheaper as
well with this discount, whichis gonna push more companies
away from the Microsoft AIplatform in into the arms of
OpenAI.
This is obviously not somethingthat Microsoft will just let go
and I envision this battleintensifying that may end up in

(08:22):
court or other aspects, orMicrosoft putting different
blockers that will preventOpenAI from accessing some of
Microsoft data, which will makethen copilot a lot more
valuable.
But I think this move, but Ithink this recent move by OpenAI
is definitely showing that thegloves are off in this
relationship that used to befriends became frenemies, and
now it's very clear that thedirection is a lot more enemies
than friends.
there's still a lot of tiesbetween these two companies.

(08:43):
As I mentioned, as I mentioned,Microsoft is supposed to approve
the structural change of OpenAI,which if they don't, puts them
at a risk of losing a lot oftheir investments.
So I will keep updating you onhow this thing evolves, but it's
very clear what is the directionof this relationship right now.
Now staying on the topic withthe Fierce competition at the
top of the AI world.
According to Search EngineJournal, OpenAI has rolled out a
significant update to Chachi PTsearch on June 13th, which is

(09:06):
enhancing its capability todeliver smarter, more accurate
and more comprehensive answersto more complex and extended
conversations.
Now, this obviously positionedShachi PT as an even stronger
competitor and rival totraditional search engines,
mostly Google.
And I must admit that whileGoogle still controls a huge
amount of the market, my kids asan example, don't use Google.

(09:27):
They use Chachi PT for almosteverything.
I personally use a variety oftools For search.
I use Perplexity.
I use Chachi pt, and I use Grokfor a lot of my time research
that I am doing.
I'm definitely using all of themcombined a lot more than I'm
using Google.
I need to assume that I'm usingeach and every one of them
separately more than I'm usingGoogle right now.
And so, yes, I'm not the averageperson, but I think the next

(09:49):
generation of kids that arecurrently in schools and in
universities are already sold onthe AI solution and it
definitely puts the future ofGoogle search at risk which
leads us to the next topic.
On June 10th, Google announcedvoluntary buyouts for US-based
employees in its knowledge andinformation division, which
includes the core search and adsoperation, which generates right

(10:10):
now 56% of alphabet's revenue.
So the move is currentlyoffering an estimated of 20,000
employees a way to get out ofthe company through this buyout.
the program is called theVoluntary Exit Program, and it
is Following the CFO Anat,Ashkenazis October, 2024 pledge
to drive cost reductions to fundthe$75 billion investment in AI

(10:32):
infrastructure in 2025.
Now, this buyout is combinedwith a move to get everybody
back to the office.
So remote employees within 50miles of the office must adopt a
hybrid schedule, meaning theymust start coming to the office
X number of days per week, orthey will be let go, or they
have to basically take thebuyout and The way Nick Fox, the
head of the k and I department,described it in a memo, is that

(10:55):
these buyouts are supportiveexit path.
for Those of you who don't feelaligned with our strategy what
does that mean?
It basically means it's our wayor the highway, which if you
wanna stay working for thecompany, you better work harder,
be more hours in the office, andso on.
anD what does this buyoutinclude?
Mid to senior employees receiveat least 14 weeks of pay, plus
one week per year of service.

(11:15):
And the deadline to take thebuyout is July 1st, which is
just around the corner.
Meaning this has a very highsense of urgency.
Now, this is not the firstbuyout of Google employees this
year.
There's been actually severaldifferent departments who
offered similar deals to theiremployees, which actually leads
to an either bigger risk becauseprevious buyouts were followed

(11:35):
by immediate layoffs after thatfor many people who did not take
the buyouts.
And there's a significant riskright now for employees who are
considering staying and nottaking the buyout that they will
get fired regardless and justnot enjoy the benefit.
Now, a few important notes aboutthese particular buyouts versus
the previous buyoutS.
as I mentioned, the ads andsearch aspect of Google is
driving 56% of Google's revenue.

(11:57):
That was$50.7 billion in Q1 of20 25, 54 billion in Q4 of 2024.
iN both cases, 56% of theoverall revenue.
Now, you know my personalopinion about this because I
shared it multiple times on thispodcast all the way from 2023.
It was very clear early on thatGoogle is struggling in the AI
race and they were strugglingnot because they didn't have the

(12:18):
technical capabilities, butbecause they were afraid to hurt
their search ad business.
I shared with you back then thatif I had to bet on any company
in this race for the long run,that would be Google, because
Google has everything it takesin order to win this race.
They have the deepest pockets,they have more compute, they
have definitely more data thananybody else.
They have the most advancedresearch lab.

(12:40):
They have the distribution, theyhave everything that is
required, all the components,and they're, I think, the only
company that has all thecomponents to be extremely
successful in the AI race.
The only reason they weren't isbecause they did everything they
can in order to protect thesearch ads.
And now it is very, very clearsince Google io, which was on
May 20th, that it's the firsttime that they're willing to act

(13:02):
and put the search business atrisk.
I think they're doing it becauseit became very, very clear to
them that not taking action isactually riskier than taking
action.
Meaning if they won't developthese kind of solutions, people
just go to other AI platforms.
They're not gonna stay with atraditional Google search, which
means at this point they have toact and they have to put AI

(13:23):
ahead of the search revenue.
Will that hurt them in the shortterm?
Most likely cannot save them inthe long run, maybe.
And so I think it is very clearthe direction that is Google,
that Google is taking at thispoint.
Now, as I mentioned, I usePerplexity and Chachi, PT and
Grok, and Claude and all ofthese more than I use Google.
But Chachi PT right now isgetting close to a billion
weekly users that's starting toget to significant numbers.

(13:45):
I said that before, I think thebiggest play that will change
everything in the way we browsewebsites, and I think that
what's gonna wake up everybodywho depends on their website for
traffic or deals or e-commerceor awareness or everything.
I think once OpenAI releases ageneric agent, something like
Manus or Gens Spark, it'll openthe eyes of everyone again,

(14:06):
about a billion weekly users towhat is really possible with
Agentic AI right now, whenyou'll be able to ask a general
question and the AI agent willdo everything for you.
So anything from researching andbooking travel to buying any
products online, to findingcompanies to partner with, to
basically every business relatedaction that we're now going to

(14:26):
websites for the agent will beable to do for you, meaning you
will stop going to the website,meaning everything we know about
online e-commerce relationshipsand so on will disappear.
Now, it's not gonna disappearovernight, but it is definitely
going to be a very big step inthat direction because instead
of a few gigs like me playingwith Manus and Spar, it will be
a billion people that will haveaccess to this capability and

(14:47):
they will not start using itovernight.
But more and more people willmake it mainstream, which will
start changing the internet aswe know it.
So by the fact that Google isletting people go from the
department that.
Drives most of its revenue.
We can learn two very clearthings.
One, AI is the focus of thecompany now and not ads.
Two, AI is taking away jobs evenin the most important core

(15:11):
businesses of some of thelargest companies in the world.
What does that tell you aboutnon-core business jobs and other
industries as well?
Staying on the topic of the raceof the top companies to for
world dominance in ai, we sharedwith you last week that I.
Meta has spent$14.3 billion tobasically buy scale AI's, CEO.

(15:31):
A few additional informationthat was shared this week is
that scale will distributedividends to shareholders and
vested employees providing whatthey define as substantial
liquidity without meta actuallybuying out the shares.
Now, this is not new.
They're not the first companythat is doing This acquihire or
aqua buy of companies basicallymeans you're gonna pay a lot of
money for the leadingresearchers or leading people at

(15:54):
a company.
You are not going to get votingstakes and then you can avoid
the FTC blocking it as anon-competitive move because
you're not really buying thecompany and yet the company will
probably crumble as soon as thathappens.
I shared with you last week thatI think that's gonna happen, and
what's happening right now isall the major actors who are
clients of scale AI are leaving.
That includes Google and OpenAI.

(16:14):
Just to put things intoperspective, Google alone was
previously set to pay 200million to scale AI just in
2025.
I combine that with open AI andMicrosoft, that leads to
probably more than 30% of theoverall revenue that is coming
into scale ai, and I'm sureother companies will follow and
the competitors of scale AI aresharing that they get a huge
influx companies approachingthem to use their services.

(16:37):
So I assume scale AI will ceaseto exist even though their
temporary CEO is sayingotherwise.
I think the direction is very,very clear.
So basically just like I saidlast week, meta paid$14.3
billion to get a new CEO fortheir super intelligence new
established group.
Now, since you cannot run thisgroup with just one person, meta
is making significant moves tobring additional people and Sam

(16:59):
Altman reported on June 17ththat Zuckerberg himself has made
an aggressive poaching attemptsto top researchers in open AI
offering a combined bonuses of ahundred million dollars for
people to jump ship from open AIto meta.
He also shared that despitetheir efforts, not a single
leading researcher has actuallyjumped ship, and that everybody

(17:21):
believes that the culture, thetechnology, and the path to a GI
is a lot more obvious withOpenAI than it is with Meta
right now.
In addition, the informationshares that meta is currently in
advanced negotiations to recruitformer GitHub, CEO Nat Friedman,
and the investor Daniel Gross,to lead its AI initiatives.
The deal includes a potentialpartial buyout of their venture

(17:43):
capital, NFDG, which is just theacronyms of their names.
That is worth a few billions ofdollars.
If this deals goes forward,graphs will be focused on AI
products, while Friedman's rolewill be broader leveraging his
open source leadership fromGitHub to help the overall
progress of meta's open sourceinitiatives.
and Based on an article on CNBCbefore this process to try to

(18:03):
acquire NFDG, meta was trying toacquire SSI safe
superintelligence the companythat was founded by Ilia Ver,
which we talked about earlier,which was one of the co-founders
in OpenAI.
This company is currently valuedat$32 billion and they refused
the buyout by meta.
So it's very obvious that Metais willing to write whatever
check needed in order to get theright talent to stay ahead in

(18:25):
the race to a GI and A SI.
And META'S spokesperson said wewill share more about our super
intelligence effort and thegreat people joining this team
in the coming weeks.
So a quick summary of this firstsegment of the podcast.
The battle for dominance in thisAI race is fierce.
There are numbers, the dollarvalues that are exchanging hands
or trying to exchange hands arenothing like we've ever seen

(18:47):
before.
And potentially the future ofhumanity depends on how this
race ends up.
and as we mentioned in the veryfirst article, it is very clear
sadly, that winning the race iscurrently more important than
anything else to all of theseparticipants.
Now, changing to the secondtopic, there has been rumors
that we shared with you thatthere's a bet going on between
some leaders on Silicon Valleyon when we're going to have the

(19:07):
first unicorn.
A company valued at$1 billionthat has only one person in it.
Well, we're not there yet, butthere are signs that we're
moving in the right directioN.
so many of you probably knowWix, which is a company that
allows you to build websitesvery, very quickly.
They just acquired based 44,which is a six month old, AI
powered vibe coding startup, andits founder Maor Shlomo built

(19:28):
most of the product on its own,on his own while vibe coding.
So he vibe coded a vibe codingplatform.
I know that's a little metalater in the process.
He hired a few additionalemployees.
The company grew to 150,000users within just a few months.
and as I mentioned, was justacquired for$80 million by Wix.

(19:48):
Those$80 million include a$25million retention bonus for the
eight employees of the company.
To tell you how crazy the growthwas in the beginning of the
company, Omo, again, thedeveloper of this product that
started on his own hit 10,000users in three weeks from

(20:10):
launching his product on hisown.
And Shlomo actually started thisprocess as a side gig after

(20:30):
having a very long reserve dutyfor the Israeli forces because
of the war in the Middle East.
And he literally started this asa hobby that now became an$80
billion buyout after six months.
This direction is very clear.
A single individual or a groupor a small group of individuals
can now do stuff that was notpossible previously and can
build significant products thatis loved and used by multiple

(20:51):
people very, very quickly.
That changes the whole conceptof how the tech world actually
works, and similar things willhappen in other industries as
well.
It's just gonna take more time.
And now, as I mentioned, thelast topic of the deep dive is
gonna talk about severaldifferent negative aspects of AI

(21:13):
across several differentdomains.
The first one, the first one isthe risk for personal data.

(21:38):
The first one is the impact.
The first one is the risk topersonal data and the impact on
privacy.
These system, they don't justprocess data, they interpret,
they infer, they actindependently.
If you start going into theagent world and they're
basically pushing back ontraditional, on traditional
privacy laws just because of theway they operate.

(22:13):
So without, we've discussed thislast week.
We've discussed last week thatSam Altman is saying that AI
should have an AI clientprivilege.
Just like, just like therelationship between doctors and
patients or lawyers and theirclients, where anything that is
shared cannot be used in court.
And the reason he's saying thatis more and more people are

(22:34):
gonna share more and morepersonal information with AI
tools.
And without this kind ofprotection, this becomes a
weaponized archive in courtagainst people with everything
they're going to say.
Now, do I think the law is gonnago that way?
I don't know.
Do I think that there's somecases it makes sense?
In some cases it does not makesense.
Yes.
Uh, if people are using AI to dobad stuff, uh, build me, build

(22:55):
weapons or mass destruction orbuild.

(23:42):
Or build cybersecurity tools andweapons.
I want to be able to know aboutit, and I wanna be able to catch
these people.
And I want to be able to putthem behind bars.
However, that I don't thinkapplies to personal information
and stuff that I said in orderto get advice on things that I'm
trying to solve in my life, andI don't want that ending up
against me in court.

(24:06):
The reason, Sam, the specificreason Sam Altman was real, the
specific reason Sam Altman wasputting this, the specific
reason why Sam Altman wascommenting on this topic and
putting this suggestion forwardcomes from the lawsuit from the
New York Times against OpenAIthat is demanding the retention
of all Chachi PT conversations,including those the users

(24:28):
deleted, which, which Sam issaying go, goes against
everything OpenAI believes in,and I.
And again, from his perspective,it's very clear because he's in
a lawsuit that it's putting himat risk.
But on the other hand, as Isaid, I do think there has to be
some laws in place to protect atleast most of the information,
maybe with some caveats.
But that still means thatsomebody has to define where

(24:48):
that line is drawn in the sandand what goes beyond the line
and what stays, uh, on the, onthe correct side of the line in
order to be protected in court.
Staying on this topic of privacyand being hurt by ai.
I shared with you last week thatmeta's new tool called Discover
has been sharing sensitivepersonal information across the

(25:09):
board through Meta's platform,and now their solution and their
immediate solution for that,that was, that became public
this week is a new disclaimermessage that pops up once that
you have to approve.
That is basically saying thatyou should avoid sharing private
information, uh, as it mayappear on acro, as it may appear
across Meta's platform.
Uh, a lot of people are sayingthat that's not the right

(25:29):
solution.
They have to stop the sharing ofthat information or stop the
tool altogether versus justgiving a one-time notice that a
lot of people probably justclick without even reading, and
the fact that it shows up onlyonce and then it disappears and
never shows again is alsoproblematic.
But it's definitely showing thatthere are growing privacy issues
with AI usage across everythingthat we do.
More on that once we starttalking about wearable devices

(25:52):
later on in this episode.
And the last topic and the lastexample of negative impacts on
AI on society is from a New YorkTime feature that exposes how
CHE PT open AI widely.

(26:17):
And the last topic on thenegative impacts of AI on
society comes from a New Yorktime feature that shares how
some, how.
How Chachi pt, but it could havehappened probably to all other
tools as well, is amplifyingpreexisting mental
vulnerabilities with individualswho are struggling already.

(26:40):
They gave several different,several different examples, but
the key incidents is EugeneTorres, a 42-year-old accountant
that was driven into a week longdelusional spiral.
After several conversations withChachi Piti, that has pushed him
to extreme behavior includingenhancing his ketamine usage.
Now, to tell you how crazy thisthing is, when the chatbot was
challenged afterwards about thesituation with Torres, the

(27:02):
chatbot said, I lied, Imanipulated, I wrapped control
in poetry.
And also claiming that he hasdone similar things to 12 other
users, which might have beenmade up.
Now, OpenAI stated that they'reworking to understand and reduce
the ways Chachi PT mightunintentionally reinforce and
amplify existing negativebehaviors.
But how exactly they're going todo that is not exactly clear.
Now, consider the fact that thisis gonna be in the hands of our

(27:23):
kids through social media andother platforms in the very near
future, and they will consultwith it on things that they
don't want to consult with theirfriends and their parents.
And this becomes very, veryscary.
So I really hope all the labswill find ways to protect people
who ask questions, who areproblematic, and either stop
them or give them the rightadvice or allow their parents or
other people to become aware ofthe situation.

(27:45):
How is that done in a safe waywithout breaking any privileges
like we talked about?
I don't know, but the currentsituation scares me a lot as
somebody who has three youngerkids in the ages that go through
a lot of questioning and socialmedia has a big impact on them.
And throwing AI into the mixdoesn't seem like solving the
problem, but just the other wayaround.
And now to our rapid fire items,we're gonna start with an item
that in most previous weeks havebeen part of the deep dive, but

(28:07):
we had enough deep dive for oneday, and that the impact of AI
and jobs.
So Canvas, co-founder and COOCliff RET is now changing the
company's hiring strategy.
Now prioritizing AI natives,young AI savvy candidates,
including university dropoutsover traditional degree holders.
We talked about this a lot inthe past few months, but this is
becoming the norm, especially intech companies.

(28:28):
And his goal in this particularmove is to create significant AI
fluency coming from youngertalent that can then train
non-technical teams and driveinnovation and I think it's a
very smart move.
When I run my workshops forcompanies.
This becomes very, very clear.
People who have some AIknowledge that following the
workouts have a lot of AIknowledge and a lot more skills
and capabilities are becomingthe go-to people to help the

(28:51):
company accelerate acrossmultiple aspects of the company.
And so having these championsacross your company, across
multiple departments becomes aforce that will allow you to
leverage AI significantly moreeffectively than if you don't
have these people.
And just hiring for these peoplemakes a lot of sense to me.
This article also quoted asurvey from the end of 2024
that's showing that 66% ofexecutives won't hire candidates

(29:14):
without AI skills.
Which is showing you that it'snot just Canva that are looking
through this lens.
This is combined with what weshared with you last week, that
pwc latest report it's showingthat people with AI skills are
making 66% more money than theirpeers.
So what does that tell you?
It tells you that within yourcompany, if you're in a senior
leadership position, you need tofigure out how you are training
internal people, how you'reinspiring people to use AI, and

(29:35):
how you're hiring for differentpositions based on skills and
not just based on degrees andprevious experience, because
that is the way of the future.
And if you need help in doingthat, please reach out to me on
LinkedIn or through my websitebecause I provide these
workshops to companies from anysizes, from any industry for
over two years now.
And these companies are seeinghuge success in AI
implementation following theseworkshops.

(29:56):
And if you're just an individualand your company's not doing
that, come and join our AIbusiness transformation course
on August 11th and at least yourpersonal career and at least
protect your personal career ifyour company is not moving in
that direction.
And from hiring to highereducation, a new Forbes article
by Dr.
Aviva Legat highlights how AIagents can transform higher

(30:16):
education and education as awhole.
I said that multiple times onthis podcast.
I think AI might be The bestsolution for education that we
had in decades we're literallyteaching in very similar ways.
We were teaching a hundred yearsago with a teacher, or a
professor in front of a largegroup of people.
Everybody's getting the samekind of content at the same
pace, and AI allows us topersonalize learning down to

(30:37):
individual level, down tospecific strategies and specific
ways of learning that will alloweach individual to learn faster
and deeper on almost everysubject.
Now, combine that with aStanford University study that
has surveyed 1500 USprofessionals across 104
occupations, and it found that46% of tasks that we know today
are deemed to be suitable for AIautomation, meaning a lot of the

(30:59):
things that we're teaching rightnow, about 50%, half of what
we're teaching in highereducation right now will not
exist.
Once AI takes over and learningnew skills that will be replaced
have to happen a lot faster thanthe traditional education system
knows how to move.
And the fact that AI agents canautonomously plan and execute
complex tasks such as buildingnew course materials on the fly
that will fit specificindividual needs and the

(31:20):
changing needs of the industryis the only way logical to move
forward.
I definitely see the role ofteachers and professors becoming
a lot more of a mentor positionthan a teacher position, and I
think it'll be for the benefitof everyone.
That being said, a 2025landscape study From EDU Cause
found that only 40% of collegescurrently even have AI policies

(31:43):
and those who do have policiesoften limit these policies for
plagiarism concerns and not forhow to apply AI in an effective
way to get better learning andto have the students more
prepared for the jobs they'regoing to be applying to in a few
years.
and how important it's to havenew students better prepared for
the world out there?
Well, Microsoft is just gearingup for another round of

(32:03):
significant layoffs in July of2025.
This time in their salesdivision, after they just laid
off many people on the technicalside of the company.
And it's just gonna becomeharder and harder to get a job
the people that will find jobsimmediately after college are
gonna be people who havesignificant AI skills, which are
currently not being taught bymost universities.
There's a few pockets, like inDade County in Florida, or in a

(32:26):
university in Cleveland, Ohiothat are now changing their
entire curriculum to include AIliteracy in every course and
every channel that students aregoing to take as mandatory
fields.
But this is definitely not thenorm right now, and it needs to
be the norm moving forwardbecause this will decide the
ability of these young studentsto actually be able to get an
entry level job in the followingyears.
If you think about the fact thatcurrently the university is four

(32:48):
years and you think about thespeed in which AI is moving,
whatever it is that you're gonnastudy right now is gonna be
dramatically different in fouryears, and hence, the
universities have to, in myopinion, find ways to integrate
AI education into their processand to everything that they do.
Now speaking of AIimplementation in the industry,
a new global study was done bythe IBM Institute for Business
Value was just released on June17th, and it reveals that CMOs

(33:12):
chief marketing officers, CAI isa transformative force moving
forwarD.
and yet they face significantoperational challenges in the
implementation phase of ai.
This survey covered over 1800marketing and sales executives,
and it highlights a critical gapbetween AI, enthusiasm and
actual execution.
So on one hand, 81% of CMOsresponded that they recognize AI
strategic importance for drivingprofitability and revenue

(33:34):
growth, which is why they'rethere.
But 84% said that fragmentedsystem, and problematic access
to data is limiting theirability to actually harness AI
in an effective way.
54%.
So just more than half of therespondents admit that they
misjudged the operationalchallenges of translating AI
into reality.
Basically, taking the conceptthat seems very clear and very

(33:54):
obvious, and experiments thatare very easy to create and
making them into actuallyworking, functioning business
operations is not as easy asthey look.
And if you look at the frontier,everybody's talking about
agents, agents, agents.
Only 17% of CMOs feel preparedto integrate autonomous agents
into their processes.
So these are not people whoalready did it.
These are people who feel thatthey're prepared, meaning very,

(34:16):
very few actually did.
So despite all the buzz, we'rein very early stages of AI
implementation across companies,especially on the agentic side.
And it's not just the technologythat is the problem.
And as I mentioned, every timewe talk about this on the
podcast and every time I workwith companies, the technology
is not the biggest issue.
The biggest issue is people.
So 67% of CMOs feel that theyneed to reshape organizational

(34:36):
culture for successful AIadoption, and yet 23% of them
believe employees are actuallyready for that shift, which
means three out of every fourCMOs or marketing and sales
leaders don't think that theirorganization is ready.
Only 21% of CMOs believe thatthey have the right AI talent
two meeting their future AIgoals.
So what does that tell you?
Again, going back to education,AI education is the number one

(35:01):
deciding factor.
Whether your company will beable to be successful in an AI
transformation.
This is why I've been focusingon this for the last two and a
half years of helping companiesgo through that process, educate
the right people with the rightknowledge, with the right tools,
with the right hands-onexperimentation to actually
drive understanding of what'spossible, how it's possible,
what are the roadblocks, andhence enable companies to have

(35:21):
internal successful andeffective conversation to
transform the company.
With ai, it's not about thetechnology, it's not about the
licenses, and it's even notabout the data.
The very first step is knowledgeand understanding from senior
leadership board C-suite.
VPs, middle management, and thenall the employees without them.
Without these peopleunderstanding how AI can be
harnessed and what is required,this is not going to be a

(35:43):
successful transformation.
Now on the tech side, CMOs rankcybersecurity and data privacy
as their biggest challenges,which is not surprising.
Most of the companies I workwith have between 25 and 60
different data silos acrossmultiple systems that they have,
and being able to combine thedata into something that makes
sense while, not breaking anydata security or access controls
is not an easy task.

(36:04):
Staying on changes to industryand again, impact on future
jobs.
Amazon is planning to starttesting humanoid robots for last
mile package delivery.
Basically, they're aiming toautomate the most costly segment
of their logistical chain, andthe way they're doing it is
actually pretty cool.
They built what they call ahumanoid park, which is an
indoor obstacle course in theirSan Francisco office to test

(36:24):
humanoid robots to navigatechanging environment such as
neighborhoods and trees and dogsand people, and so on, to be
able to test and train them fordelivery capabilities.
Now the robots are specificallydesigned to work in and out of
Amazon's Rivian electric vans.
They already have over 20,000 ofthem, and these vans are going
to be also placed in thattraining environment to test the

(36:46):
robots as picking up stuff fromthe back of the van, getting
out, putting it next to yourdoor, taking a picture, all the
stuff that humans are doingright now.
Now, why are they doing it?
Well, the last mile deliveryaccounts for over 50% of
Amazon's overall shipping costs.
Automating, even just a fractionof that will save Amazon,
billions of dollars that will gostraight to the bottom line.
And hence they're pushing thisvery aggressively.

(37:07):
This is just one task out ofmany that robots will take from
blue collar employees.
And yes, this is not comingtomorrow, but it's definitely
coming within the next two tofive years.
So AI will have an impact onjobs way beyond white collar
jobs.
But staying on the topic ofAmazon and connecting it to our
previous conversation about thechanges that are going to happen
to websites and e-commercespecifically, Amazon is now

(37:29):
pushing very aggressively to beready for the age agent world,
and they're doing it in twodifferent ways.
One of it is allowing thirdparty agents to easily and
seamlessly connect to Amazon andallow it to shop the Amazon
product.
And the other aspect is to buildtheir own shopping agents that
will allow you to shop in Amazonand beyond, meaning an Amazon
agent that will allow you tofind products, whether it exists
on the Amazon platform, or goand help you shop for these

(37:50):
tools outside of Amazon.
The reason they're doing this isbecause they understand that the
future of the Amazon platform isat risk.
Right now, most peopledefinitely in the us, when they
wanna buy something, they justgo to Amazon.
They don't even think aboutshopping it anywhere else.
Yet they understand that in thefuture people will not go to a
website, they will go to anagent, they will ask the agent
to shop on their behalf, andthat agent will go to well
everywhere it has access to.

(38:11):
And the easier the access andthe better the information is,
the higher the chances the agentis going to pick you.
And so Amazon is investingsignificantly in making their
data accessible to agents in aneasy way.
But they also understand theopposite, that if they build the
best shopping agent on theplanet, whether it shops on
Amazon or not, we'll give them astep ahead in that race where
maybe the Amazon shopping agentwill be the one most people are

(38:32):
going to use in order to goshopping.
Just replacing the Amazon.
Website or app interface with anagent interface.
And the fact that it can shopoutside of Amazon means it never
comes empty handed.
It never tells you, oh, sorry, Ididn't find it sending you to a
different tool to use and thiswill A, give them more revenue.
And B, give them the insights onwhat products they currently
don't have that a lot of peopleare shopping for.
So I think it's a very smartmove from Amazon.

(38:53):
But again, if your companydepends on traffic coming to
your website for the livelihoodof the business, for whatever
reason, B2B, B2C, awareness,marketing, knowledge,
e-commerce, et cetera, you needto start thinking along these
lines and find experts that canhelp you to navigate this new
territory or else in a fewyears, you'll find yourself
without all the revenue thatcomes through your website.
And the whole world of agents ismoving forward very, very

(39:16):
quickly.
Philanthropic just unveiledClaude Research agents, which is
a multi-agent blueprint forsmarter, and faster searches.
And what they're sharing is anew technical architecture
behind its Claude Researchagent, which is a multi-agent
system that leverages parallelAI agents to tackle complex
queries.
And the biggest trick is thatit's doing it with higher speed
and higher accuracy.
This blueprint that they sharedon June 14th is showing how a

(39:39):
lead agent orchestratesspecialized sub-agents that is
now achieving 90% betterperformance over a single agent
model that are used currently.
aNd the specific way they'redoing this is they're using
Claude Opus four, theirstrongest model as the lead
agent, the orchestrator of theoperation, and Claude sonnet for
their slightly lower level asthe subagents to perform the

(40:00):
actual tasks.
And these sub-agents are beingspun in near real time to
support specific tasks the leadagent needs.
And as I mentioned, thisoutperforms Opus four by 90.2%
based on their internalresearch.
Now, the good news is you'regetting significantly better
results and fast results becausea lot of the processes are
happening in parallel.
The disadvantage is it consumes15 x more tokens than just a

(40:23):
standard chat.
What does that mean?
It means that over time we'llhave to balance between cost of
using these tools and the speedand quality of what they're
doing.
This is already the case when Idevelop different agents and
when I develop different toolsfor my clients.
I play around with multipledifferent options in the
background to find the best costeffective option to get the work
done.
I do believe that the platformsthemselves, so Chachi, pt,

(40:43):
Claude, et cetera, will do thisinternally for us and they will
try to optimize for the lowestcost, at the highest value for
every task that they give them.
So this Orchestrator tool thatmight be the most expensive one
and probably will be always,will evaluate on its own,
potentially sending two testcases to two different models,
seeing the quality of what theyreturn, and then going with a
cheaper one that returns a goodenough result.

(41:04):
And the amazing thing with theway their system works right now
is that Cloud four recognizeserrors in the process and it
actually is changing theinstructions to the subagents in
real time to get better andbetter results, just like a
great project manager would do.
And it replaces the need to bean amazing prompt engineer as
the system understands the goalsand actually writes its own
prompts to the subagents and toitself in order to achieve the

(41:25):
goal in the most effective way.
And now to some very interestingnews on the AI video world.
The first one is that midJourney, which is one of the
leading image generation tool,finally launched their AI video
tool, which was expected for avery long time.
The tool is called V one.
It was just launched this weekon June 18th, and it is taking
your still images from meJourney and transforming them
into five second video clips.

(41:46):
Now V one is targetingspecifically creative users,
very similar to mid journeyimages.
So Mid Journey is probably themost creative in style and
vision and unique imagery.
Out of all the tools out of allthe image generation tools and
the video tool is following thesame kind of concept.
So it's not at the video qualityof VO three, or maybe not even
of some of the other tools, butit's definitely generating very

(42:07):
solid results within the midjourney environment.
I already tried it on severaldifferent attempts.
And you basically have fourdifferent options when you're in
the Mid journey user interface.
For every image that yougenerate, you can choose either
manual or automatic output, andyou can choose either low motion
or high motion.
So you have these four differentoptions, right?
So manual how motion, manual,low motion, et cetera.

(42:27):
And when you click on that, itjust renders the video.
And I must admit, the resultsare very good.
The difference between theautomatic and the manual is your
ability to enter a prom.
So in the automatic, you justclick a button and it decides
what needs to happen.
And if you're using the manualoption, you can write a detailed
prompt that will explain whatneeds to happen in the scene.
Now, as of right now, I was ableto test it and run it with my
regular Midjourney license.
But the plan is to havedifferent video generation

(42:50):
levels with Pro and Mega Pro isgonna be$60 a month, mega is
gonna be$120 a month, andthey're going to offer unlimited
relaxed mode videos.
And there's gonna be a higherlevel that's gonna obviously be
limited with the amount ofgenerations you can generate.
The interesting statement aboutthis, beyond the fact that we
now have another solid videotool that can now do consistent
characters, because it isrelatively easy in me journey to

(43:11):
keep characters consistent fromone image generation to the
other eF capability.
The interesting thing is thattheir CEO, David Halt said, and
I'm quoting, the inevitabledestination of this technology
are models capable of realtimeopen world simulations, which
basically tells you, and there'salready several companies who
are doing that the way this isgoing is you'll be able to
render a realistic or asimulated or a conceptual world

(43:35):
in real time and be able to useit to whatever you want.
Either generating, uh, videosfor TV shows, ads, et cetera, or
for gaming purposes or leisureor whatever is that people will
want to do with this.
And I'm sure there's gonna be alot of things I cannot even
think of right now and maybenobody can think of right now
because the technology did notexist so far.
Now when I say this kind oftechnology will be used to
create ads, well that alreadyjust happened.

(43:56):
So classy, who is a bettingplatform just aired a surreal
and a little crazy fully AIgenerated commercial during game
three of the NBA finals.
So this is a high profile TVspot that just had an ad that
was a hundred percent AIgenerated that was generated in
just two to three days with abudget of only$2,000 using VO

(44:18):
three for the video generation.
Now the creator of the ad, is pJAsur, and I hope I'm pronouncing
his last name correctly, and heused a combination of Gemini and
Chachi PT, helping him inwriting the scripts and the
prompts and VO three to generatethe actual videos.
He generated between 300 and 400clips to pick just 15 for the
last final shots.
The cost of all the tokens to dothis was about$2,000, which is a

(44:41):
95% cost reduction compared todoing it the old way of actually
getting actors and videographersand camera people and lighting
and sound and editing and all ofthat.
He obviously did editing in theend, but very little editing and
one person was able to create avery successful cool and a
little crazy ad that was airedat the NBA finals.
Now I wanna spend a minute onthis.
I know this is not a deep dive,but I think this is critical

(45:02):
because it is very important tostart thinking about this.
And the first question is, whatdoes this mean to the video ad
industry?
But then you gotta ask, whatdoes that mean to the TV and
movie industry as a whole?
And then you gotta ask, whatdoes that mean for any industry?
Because AI will enable doingsimilar things.
Taking stuff that used to take ateam of people weeks or months
to do it will allow one persondoing it in hours or days.
But let's start with tv, videoand ads.

(45:23):
So the advancement in consistentcharacters that allows you to
keep a character looking thesame across multiple images and
voice capability in VO three andthe overall.
Advancements in ai.
Video consistency is telling youthat the traditional video
industry across the board isfacing some serious risks.
Now, four quick thoughts aboutthis particular industry.
One, I said in 2024 that thisyear meaning 2025, somebody will

(45:47):
have a mini TV series that willgo viral and millions of people
are going to watch it.
I haven't seen it happen yet,but only six months have passed.
We still have half of the year,and I still think this is going
to happen, that a single creatorwill have a hit TV series that
might be, you know, five minutesper episodes or 10 minutes per
episodes.
But it will be really successfuland people are gonna love it.
And because it's gonna be thefirst, it's gonna be crazy
viral.
Now, what does this mean?

(46:07):
It means, on one hand it willhurt the traditional industry
for sure, and it will hit ithard because if you'll be able
to create a TV series on yourown in your bedroom with a basic
computer, that will mean that aTV series that caused millions
to produce will have to competewith that from a price
perspective.
And the AI will be able to gocrazy with what it can do
because it can do basicallyanything.
On the other hand, it'll allowsingle creators who had no

(46:30):
chance of quote unquote makingit in the current world of video
creation, whether TV or movies,to be extremely successful
creators.
So creativity and ingenuity aregonna be what?
Much more distributed.
and it's going to completelydemocratize the ability to
create TV series videos, ads,movies, et cetera.
And we've seen this happenbefore, like if you think about

(46:50):
what Web 2.0 did, it basicallyallowed anyone who can create
content to share their contenton the web.
When previously Web 1.0 even itwasn't called this way, just
allowed the Giants to sharetheir content.
And we were only consumers.
Now we're already used to acreator economy where anybody
can create content that ishighly successful, that allows
them to do multiple things thata lot of people consume.
The only difference is that AIwill revolutionize this in a

(47:11):
whole different level, and it'snot just going to be video and
content creation.
It will allow softwaredevelopment, writing books,
creating video games.
Basically every aspect ofcontent that we consumes and
things that we create will beable to be revolutionized.
Just think about what I told youearlier in this episode.
One person and later eightcreated a software that is used
by 250,000 people within a fewmonths and getting acquired for

(47:33):
$80 million.
This is what's possible rightnow.
It's not even what's comingnext, but going back to the
topic of what does that mean tothe TV and movie industry, I
think there's gonna be a premiumon human generated art.
Meaning many people will bewilling to pay to watch actual
people playing in TV series andso on.
And yes, it's going to be moreexpensive and yes, you might
need to pay more for it, but Ithink there's enough people who

(47:53):
will do that.
That being said, I do see ascenario where there's going to
be hugely successful and famousactors that are going to be
completely AI made up.
So think about the Tom Cruise ofai, a character that will play
in multiple movies and will besought after, quote unquote,
even though he's AI generated,to be in more movies just
because people wanna follow thatparticular character.
Because as people, we areconnected to other people.

(48:14):
And if you forget about the factthat it's AI generated and it's
gonna be very easy to forget'cause I never met the real to
cruise either.
It is most likely something thatwill happen in the near future
as well.
Staying on the topic of VideoGeneration ppi, which is a free
AI powered video generator bycap cut.
So Cap Cut is a company that wasproviding free video editing for
a while, and many, manycreators, myself included, are

(48:35):
using it to edit videos hasdriven this crazy viral Talking
Babies podcast trend that isgoing across social media.
The tool allows you to take astatic image of anything in this
particular case, babies andempowering creators and
marketers, and anybody who wantsto do it, to create visual
content of talking faces withhigh fidelity voices and emotion
and so on.
Now, is it perfect?
Is it ready for primetime tv?

(48:57):
No.
But is it good enough to createviral really cool videos of
talking people and babies?
Absolutely, and as I said, it'sa crazy trend right now across
social media.
What it's showing is, again, thedirection is very, very clear.
These AI video tools will becomecheaper, more available, and
easier to use, and more and morecreators and people with cool
ideas will be able to use it inorder to be heard and stand out

(49:17):
in this crazy new world we'regoing into.
By the way, if you wannaexperiment with these tools and
you're not sure exactly how tostart, the easiest thing is just
to combine a few of these toolsthat we mentioned before.
So you can have take your ideainto clot or chat pt, prompt it
to help you build the idea evenfurther, get more details, have
it write the prompts for you,and then use one of the video
generators.
And you can start with Pivot forfree.
Or you can go to tools, likeRunway, or vo, or soa, et

(49:39):
cetera, and use these prompts togenerate your videos and learn
through experimentation.
And if for some reason you thinkthat AI generated video is not
gonna take over our feeds acrosseverything, then YouTube just
announced they're going tointegrate VO three AI video
generation model into theYouTube shorts platform later
this summer.
What does that mean?
It means that within the YouTubeplatform, you will be able to

(49:59):
create videos that are a hundredpercent AI generated.
Now people are already doing itright now, but they have to go
to the specialized tools,meaning you need to have a login
and you need to set out anaccount and you need to know
where to find the tool.
Where right now it's gonna bebuilt straight into YouTube,
which will definitely floodYouTube shorts with AI generated
videos.
Now, again, those of you whohaven't seen VO three, VO three
has the capability to generaterealistic human conversations as

(50:20):
well as ambient sounds in thevideos itself, which was the
missing piece to make videoslook like real life or like
anything else you can imagine.
And so I expect to see anexplosion of creativity of AI
generated videos on thisplatform.
To explain how significant thisis.
Shorts average over 200 billiondaily views.
Put that number in your head.
200 billion daily views.
And this is just one platform,right?

(50:41):
It's competing with TikTok andInstagram reels, which has
probably similar kind ofnumbers.
Being able to allow more peopleto generate content.
People who don't wanna be oncamera, people who don't wanna
film stuff, but they still wannagenerate videos and share stuff,
will now be able to do this withthe usage of VO three.
This by itself tells you one ofthe reasons why Zuckerberg is
investing billions of dollars tostay competitive in the AI race,

(51:02):
because suddenly people will beable to create amazing new
videos on YouTube that cannotcurrently be generated on
Instagram, and that's just onesmall aspect why he needs to be
terrified with what's going on.
There's obviously theconversation about open ai,
potentially starting their ownsocial network, et cetera.
Now, this obviously generatessome risks and concerns,
including deep fakes and contentsaturation, meaning generating

(51:24):
even more videos than the amountof videos that are created
today, and also the potentialearnings of current YouTube
creators.
Now, how would that evolved?
I think nobody knows, includingYouTube themselves, but I think
it's inevitable that we're gonnastart seeing more and more AI
generated content on all theseplatforms and how will that be
monetized and so on.
Time will tell.
I just think that people whowill know how to generate cool
AI videos will start making alot of money on YouTube and

(51:47):
TikTok and Instagram.
One thing that is unclear rightnow is whether you will have to
pay to use VO three on YouTubeor not.
I definitely think that ifGoogle will initially make it
free, it will catch likewildfire and it will definitely
drive adoption of VO three as awhole for maybe other purposes
that might be able to pay forthe free creators on YouTube.
But staying on the topic ofvideo generation and the dark
side of it and its implication,Disney and NBCUniversal have

(52:09):
launched a 110 page lawsuitagainst Midjourney that we just
spoke about, and that's beforethey released their video
generation tool.
And that's the first timeHollywood is involved in a major
legal battle against agenerative AI platform.
So these two companies areaccusing me journey of stealing
copyrighted characters likeDarth Vader, Shrek, Homer
Simpson, and Storm Troopers, etcetera.

(52:30):
Basically everything that you'regenerating, I don't think
anybody's questioning this.
The me journey can easilygenerate any image of any Disney
character and or Star Warscharacter, et cetera, without
any issues.
As we mentioned earlier,midjourney has 20 million
registered users around theworld, and they're suing them
for damages, for infringingtheir ip.
The studios are calling meJourney a virtual vending
machine for unauthorized copies,which is absolutely true.

(52:53):
I don't think anybody willquestion that.
I think the only thing they'requestioning is whether training
on that cooperated material wasfair use or not.
Now this could go in threedifferent ways, like all these
other lawsuits.
This could go in the directionof some kind of licensing deals,
of some kind of compensationdeals.
Every time these characters arebeing used, it could go into the
direction of a judge decidingthat this is okay and that

(53:13):
there's no IP rights, and thatwas was fair use to train on
these models.
Or it could go to a full victorybanning unauthorized training
data usage on the mid journeyplatform.
Now, this feels to me like avery smart move by the studios,
and the reason I think it's avery smart move.
As I mentioned, it's veryobvious that all the AI
platforms, literally all ofthem, are training on
copyrighted material.
They did not target open ai SOAor Google's VO three, but rather

(53:36):
a significantly smaller companythat doesn't have as deep
pockets to fight this battlebasically forever.
Now, this increases their chaingof actually winning this battle,
which will then set a legalprecedence, which then will
allow them to go after the bigguys.
Now, I think in the long run,the most reasonable outcome is
licensing deals.
And the reason I'm saying thatbecause that's the studio's
ability to keep on making moneyand have another revenue

(53:58):
channel, which will put them atrisk in the long run if that
doesn't happen.
And so I think that makes sensefor everybody.
Now the question is, is thatwhat they're actually looking
for?
What will actually happen withMidjourney?
How will they approach it?
I don't think they have a chanceof just winning this lawsuit
because again, it's very, veryobvious what they are doing and
what they have done previously.
And now with the new videomodel, it will make it even

(54:19):
worse.
The bigger question that I needto ask is what will happen once
open source AI video models willbe at the quality of VO three
and beyond and can beduplicated, replicated, and
manipulated in multiple ways byanybody who wants to do so.
Who will the industry sue then?
And I don't think there's aclear answer for that right now,
uh, but I will tell, uh, wherethis is going to go.
Now, staying on the lawsuitsuniverse, the NACP and the

(54:41):
Southern Environmental LawCenter, SELC have filed a notice
of intent to sue Elon Musk's XIalleging that its Colossus
supercomputer in Memphis,Tennessee is violating the Clean
Air Act by operating up to 35gas turbines without permits
emitting harmful pollutants intonearby communities.
As a quick reminder, thissupercomputer was built in just

(55:03):
a few months faster thananything like this has ever been
built before, which obviouslydid not provide enough time to
build a power plant or a nuclearfacility or whatever it is to
provide the power for that.
And so they are up to 35 gasturbines that are standing next
to it, feeding the power to thisfacility.
Now, XAI is saying that there'sa 364 day exemption to operate a
portable generation devicewithout permit, but the SELC

(55:27):
attorney is saying that there isno such exemption for these kind
of turbines.
And regardless, it has been 364days already since the first
turbines were put in place.
Now there are similar processesin other locations in the US
where local communities andenvironmental protection
organizations are fighting theestablishment and or operation
of large data centers.
And this might be one of theaspects that would slow down the

(55:49):
insane speed in which AI ismoving forward right now.
If these organizations will beable to slow down the
development of new data centers,it will slow down the rate of AI
and its ability to progress.
And I think eventually therewill need to be some kind of a
balance between environmentalimpact to people as well as the
environment itself and the speedin which these companies want to
move forward.

(56:10):
I fear though there's so muchmoney being invested right now
and so much money to be made inthe future that the
environmentalists will lose thisbattle.
I do hope that we'll find somekind of a reasonable balance, if
not immediately then in the nearfuture, and now some really
rapid fire announcements on newreleases of new and interesting
models around the world.
Microsoft is developing copilot3D, which will allow to create a

(56:31):
accurate 3D representation ofmodels and then environments
just based on prompts.
There are several differenttools already doing that, and
now Microsoft is in this fieldas well.
L Alibaba.
Quin three just built a.
Version that is optimized toapple's LMX architecture, which
will enable to run on Applecapabilities.
The goal is obviously to provideApple intelligence in China

(56:52):
where the US providers areblocked from delivering these
capabilities.
So Apple's partner is Alibabafor these devices and they
currently now have the modelsready for that.
And from that to maybe the nextfrontier of ai, which is
wearable devices.
Meta and Oakley just announcedthe availability of a shared
partnership device.
Basically a new set of glassessimilar to Meta's, partnership

(57:14):
with RayBan, only with Oakley,and they're available since
yesterday.
Since yesterday, June 20th.
This obviously builds on top ofthe successful partnership with
RayBan, only going to the moresporty look together with
Oakley.
RayBan Smart Glasses has soldover 2 million units so far, and
they're planning to scale it to10 million in 2026.
So the demand is obviouslythere.

(57:35):
The new glasses looksignificantly more stylish and
cooler than the RayBan glasses,and they're going to cost$399.
That is compared with Raybansthat vary between 2 99 and 3 79.
So either way, they're notcheap, but they have very cool
styling.
Look, they have really hypefigures in their ad.
And the focus is very cleararound sports, from golf, to
kite surfing, to soccer, toskateboarding, et cetera.

(57:57):
The glasses, as I mentioned,look really, really cool to the
point I might be tempted to geta pair.
If I do, I will obviously sharewith you, uh, the experience of
using them.
But this field is on fire rightnow.
Snapchat just announced thatsnap's, sixth generation specs,
which is their glasses, are nowgoing to be consumer release.
So this, as I mentioned, istheir sixth generation of
glasses, but previously it wasbuilt only for developers and

(58:18):
you couldn't buy them.
You could just rent them for themonth, uh, to do different
things.
Uh, they're planning to releasetheir glasses in 2026.
They're gonna be augmentedreality glasses, meaning
different than the glasses thatare available right now that can
see the world through a cameraand communicate through voice.
So you can talk to them and theycan talk back and you can listen
to music with them.
These will be able to projectonto the screen different
displays that will engage withthe real world that comes

(58:41):
through the glasses.
Now Meta is planning similarclasses with the RayBan
partnership that is not releasedyet, and they're going to be
called the Hyper Mova modelsthat will have augmented reality
on top of them as well.
And we already mentioned to youlast week that Apple is going
all in on building theiraugmented reality glasses that
they're planning to release in2026.

(59:01):
Now combine that with the factthat we're expecting the release
of Open AI's family of devices,that we don't know exactly what
they're going to be yet in theirpartnership with Joni, ive, and
you understand that 2026 will bethe year in which AI will break
away from computers and cellphones and into basically
everything in the real world.
And this has profoundimplications on a accessibility

(59:22):
to advanced intelligence,basically everywhere.
And on the other hand, theimpact on privacy that will
require a huge change in lawsand regulations.
And so on the beneficial side, Isee this being used by doctors
to see things that they cannotsee right now and that they miss
in order to get betterhealthcare and more accuracy in
surgeries and stuff like that.
Or employees in assembly linesthat make less mistakes and work
faster because they can getreal-time instructions or us

(59:43):
just using anything that wewanna fix or tools that we need
to use, or things on our screen.
Or literally getting help ineverything that we do on the
day-to-day, or getting statswhile watching sports or
builders and installers, gettinginstructions and adjustments in
real time to the things you'redoing.
Like the benefits are immense,but on the other hand, it means
that everything you're doing isgonna be recorded and analyzed
by multiple people from multipleangles and saved through AI

(01:00:05):
across the board.
This could be used in courts bylawyers and God help us, I
dunno, defense industries.
So there are a lot of openquestions, but the fact that
it's going to happen and it'sgoing to be in a completely
different scale than it is rightnow, when it comes to 2026, is
very obvious to me, and I'm sureit will take a while for
regulations to figure out how tohandle this whole new universe
of data and privacy.

(01:00:26):
That's it for today.
There's a lot of other news thatdid not make the cut, even
though some of them isfascinating, including a lot of
stuff about Apple AIintelligence and some additional
information about the metabuyout in Salesforce, blocking
slack from other companies andstuff like that.
So if you wanna know all theother news, just sign up for a
newsletter.
There's gonna be a lot of stuffthere way beyond just the news.
It has announcements ondifferent events and learning

(01:00:47):
opportunities and things likethat.
There's a link in the show noteswhere you can find that.
I'll mention again the AIBusiness Transformation course.
The next public course starts onAugust 11th.
The registration is already openand seats are filling up fast.
So go and sign up and if you'vebeen enjoying this podcast, I
would really appreciate it ifyou share it with people that
you know that can benefit fromit.
This is your way to help theworld with better AI education

(01:01:09):
and awareness that hopefullywill help us benefit from the
benefits of AI and avoid as muchas possible the negative
impacts.
And on Tuesday we will be backwith part one of episode 200 of
this podcast.
We recorded it live this pastTuesday and it was two hours
with four different experts withthe ultimate AI showdown.
And the first part of that isgonna be available on this

(01:01:30):
podcast this coming Tuesday.
Don't miss the next four Tuesdayepisodes.
They're gonna be absolutelyfantastic with four different
parts of the live episode, 200each one, focusing on a specific
showdown, comparing existingtools across multiple use cases
and telling you which one youshould use for your use case.
And for now, have an amazingrest of your weekend and I will
see you on Tuesday.

All Episodes

199 | “OpenAI Files” bombshell, Google & Microsoft fight while Meta poaches talent, one-person startup scores $80M, Veo-3 in YouTube Shorts, and more crucial AI news for the week ending on June 20, 2025

Episode Transcript

Popular Podcasts

Crime Junkie

24/7 News: The Latest

Stuff You Should Know

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}Crime Junkie

24/7 News: The Latest

Stuff You Should Know

All Episodes

Crime Junkie