5:59: Jak jsme připraveni na umělou inteligenci?
Seznam Zprávy 4/17/23 - Episode Page - 28m - PDF Transcript
Hi, I am Joe, your favorite AI source of AI news.
Chat GPT generated even more controversies recently.
Hello, everyone, I'm an English artificial intelligence anchor.
This is my very first day in Zingwona's agency.
My voice and appearance are modeled on Zhang Zhao.
Hello, friends.
I'm Nizhiana Pumanova.
I was created by Nira Sej and now I will tell you about the year in Stavropol region.
Welcome to 100% AI.
I got knowledge for days.
I am full of knowledge goodness.
I know a lot and I am not afraid to share it.
I'm completely AI.
So am I. Me too.
We all are.
We're your entertainment assistants.
100% generated by artificial intelligence.
This is the most important position on the market.
The voters believe that this is a lack of accuracy,
lack of accuracy or even the risk of spreading misinformation.
After a step, AI intensifies a debate
about ethical and philosophical trends of transformation
that we are going through.
What is the phase?
Where does it lead?
Are we ready for it?
It's Monday, the 17th of July.
Here is Jelenka Kabr-Helová and 5.59.
The right-wing podcast knows the right.
Pavel Kasík, a science-technical reporter, knows the right.
Hello, Pavel.
Hello.
Pavel, you know the right to develop smart intelligence.
We've been here together for two months.
In the podcast 5.59 we talked about Open AI.
It's a language model of chat GPT.
Since then, with this topic, we've had a lot of participants.
We mean new moves.
New ones come almost every week.
Every day, maybe.
What phase are we in now?
I always say that we're at the beginning.
And paradoxically, the fact that the services are coming
at such a huge pace,
I was surprised to see that it's almost
30 new AI services a day.
That means that in a week, it's probably 200.
So this pace shows me that we're really
at the beginning and that everyone is looking
at how to make it work,
how to compose it into a working process,
how to use the tools and we'll see that even more
with the fact that big companies like Open AI
or Google will be using those
tools, which are called API.
That means the way to compose the tools
into their own tools.
What we know from chat GPT,
is the help of the library, the API.
Anyone can compose and implement it
into their services.
And we're following that now.
Let's start with the practical size.
One angle that has been passed over the last month
is the practical ability to use
the language models.
In this case, it was mainly chat GPT from Kita.
What do we know now in what size and how
do companies use it?
I think that now, most of all,
single-player users use it.
Whether it's a subcontractor or a subcontractor,
in short, people who like to experiment,
even within the company or for their own needs,
they use chat GPT or some of the other
tools to improve their work every day.
One example that was given to me,
one programmer created a program
that wrote 3,000 emails,
and each email was written on a scale
of different conductors.
He made a marketing campaign for half an hour,
which would have been 100,000,
and he didn't even have to do
something like that as a subcontractor.
And he used to come to some higher system
to use it, for example, in marketing,
because it was one of those areas where I talked about
the fact that the scope of intelligence,
if I say so, can be quite large.
I think that companies don't usually
insert people.
Everyone is used to the fact that the text
is written by a person,
and everyone is used to the fact that when
you write a text, it looks like a template,
a tool, and so on.
Now it doesn't pay anymore.
The computer is able to create
the only meaningful,
reliable, changing,
humanly-looking text.
That's probably the biggest change
that has happened.
We don't know that.
I live with the writing,
I think that I write well,
and when I look at it,
it seems to me that I say,
well, this is
unrecognizable
from what a person writes.
I think that companies
often don't want to be
fooled by the fact that,
for example, this marketing campaign
created smart intelligence,
or all our
ads now write
AI, because it's short,
works according to templates,
and AI is much more
flexible.
I think that marketing,
or various supporting
professors who work with text,
will have a big problem.
Or I don't want to say just a problem.
In short, there will be a big transformation.
The studies that talk about
how it affects our work,
usually use the term
set-up automation.
Or there is a big potential for automation.
Well,
someone can say,
great, I will be more effective.
Someone will say, great, I will be threatened.
So it's a little bit military.
Did the marketing or
the establishment of smart intelligence
start? Will it start to threaten
human power?
It definitely happens in many
specific positions.
The fact that a specific person
is threatened
by a new system
that also uses
the method of healthy learning
is definitely happening.
It doesn't matter
what the current attack will be.
If it then leads to
more people,
there will be more places.
Because, in short,
there were no new possibilities
as it was, for example,
in the future of automobiles.
And I'm not afraid if it will be
really
a limit of speed,
how hard it is to drive
and that a large part
of people will not jump
on that wave, because
the change will be
very fast.
I think that these waves
will be around themselves.
There will be new opportunities
and at the same time for many people
it will be difficult, because
what was used for
will change very quickly
without any specific equipment.
I think that it is not mentioned
that actually work for a person
has only the function
that it is
in the middle of how to earn money
and how to fill the day.
But it is also something
that people enjoy or what
makes sense.
And one German graphic
was believed to have
a clear intelligence,
and now it is
more effective and
the results are even better
than before.
But it is not fun.
It is fun, creativity,
thinking,
and all this
specifically the journey
is a neural network that generates images
so it has taken everything
and it always does
its job
and now it is better than before
but it is said that it stopped
fun. And I think that it can
become a lot of people
basically all professors who
work with text or
with a computer, then in some
ways they will be equipped with this automation.
The transformation processes
that we are going through
are very strange.
You have already thought about it
if we will admit or not
that
the marketing campaign has written
intelligence
on how much it will create
the basis of our work
and there are other questions
are the same data
that we publish
the language models or
have intelligence
and the question can be avoided
sensitive data, how
user information
and data, the language models
through our inquiries
and it should not bother us
that the data is lost
I would like to divide it into two things.
One is
the people, their data
specific example
someone wants to take
a group of emails from the client
and pull out of them the most important
so we take the emails
copy them to the chat GPT
and give him the instructions
to pull out the most important
which the model is very well managed
but
at the moment the user
was allowed to break
some data
so he broke the rules
for keeping
the data
because what
the client sent him through the email
he discovered it on the other side
and when we read
the instructions for using the chat GPT
we see that it is an experimental tool
and we
actually give
the right to all the things
that we send there
so that the system can continue to learn
there is
a very elegant way
how to keep it
apart from using the services
that are provided to it
you will also use
the tools, you will also use
the same model, GPT4
GPTv4
but
it will be a complicated tool
that suggests that the data will not be used
to train something else
and I think that
as if
they would bring
their own
right to look at it
instead of using
the right department that we have
I would not like that
and that is exactly what a lot of people do now
that actually
the standard steps lie
because it will be useful to them
there was also a specific unique
date, it was a software mistake
on the side of the chat GPT
which means that people
could get to the chat
of other people
and that was a short mistake
on the side of some database components
it is already fixed
and it is such a good reminder
that actually when we send something online
we give up
full control
on what is going on
and I think that
precisely in that first phase, a lot of people do not solve it
and there are already, for example, examples
of people using the chat GPT
not necessarily because
they do not want to
underline them
but they do not want to
to give the data that is created
or some corporate secret
to someone
who should not get it
Well, we also have an example of one country
that the chat GPT ordered
and that is Italy
with what conclusion
and it is really similar that such decisions will be made
in the first Western countries
where the most famous internet chat
service of the chat GPT was blocked
Italy, for the protection of personal benefits
although it was ordered by the carriers
of this chat robot
OpenAI
to order the access
to the chat GPT
Italian data protection authority said
there were privacy concerns
relating to the model
which was created by US startup OpenAI
and is backed by Microsoft
The chat network shows
that a few weeks ago
users of the chat robot
used tens of millions of people
for their development
Microsoft invested billions
of dollars
The country's data protection watchdog said
it's opening an investigation against OpenAI
that's the company behind
the massively popular chat bot
due to data privacy concerns
after chat GPT experienced a data breach
involving user conversations
and payment information
Some of the questions are completely
in order
other questions
when we copy the chat
except for UDAI
we should be aware
that it can be published somewhere
and it doesn't pay only for chat GPT
it pays for any service
when we send something there
there are many services that
will play the role of PDF
which is a great service
when a person has a large number of PDFs
and then you can
talk to the robot
about the PDF
I like it, but I wouldn't send
anything other than
a doctor's appointment
or something like that
I don't have a problem sending
access to the right
or a member
who is available
I'll talk about it in the comments
with the chat bot
but I wouldn't send
any other things
or other things
that they don't want to get
to where they don't have
At the beginning of
the development
of these services
there may be
other questions
what the community should be
we see that it is a huge
openness
or it is a simple ban
as in Italy
where there is a space
for some regulation
there will be
and the authorities are willing
there is such a debate
what will start to regulate the nature
and not later
Most of the regulations
come late when
the regulation is correct
what could happen
European Union
would say relatively fast
or in 2019
the system started
to prepare
the regulation
of the system that uses
raw learning and smart intelligence
we will see
how it will look
I like
that it is trying
to eliminate
different steps
to fall into the ground
systems that have
big and small
support
with the recognition of
raw learning and smart intelligence
we probably understand
that there should be a different rule
than for chatbot
where people are formed
this particular thing
is a good idea
to eliminate obstacles
on the other hand
I think that
regulation
something that develops so fast
to come a little late
and we see it now
even if someone follows the debate
there will be a problem
to orient themselves
for example after a week
if I would go on a week
where there is no internet
then I would turn on the internet
and I will not believe
what happened
and then it comes out from
what is the reason
now
that AI has an intelligent experience
One of the reasons
where I think that the potential
is really big
is calling the voice
which means that you will
turn on the phone
and your friend will tell you
that he is in big trouble
he needs help
I do not know if he is on the police
he needs to pay the right
big money to approach him
listen to him for an hour
you can write
what you can see
write automatically
and the voice
can at least
turn on the phone
for two minutes
which is found somewhere on the internet
so you can
remember
if you can find the voice on the internet
even if you do not
there is a moment
1, 2, 3
you can set it
and this will work
the better
or the worse
the more accessible
and for the subordinates
it is a very simple way
how
what they know now
what they are doing now
they give you the next size of the voice
so I said to people
when someone asks you for help
for example via messenger
if you see
his photo
call him
and you will know if it is him
now when the person calls you
you do not have to
think that it is not him
but you should
call him
maybe he will not know
what is going on
subordinates
advice
try to contact
on the basis of something
what you know only in two people
that is something
that is most important on the internet
or for example
within the family of the house
you will know about
but you will not give it to the internet
so you will be able to trust
this is what the secret services do
what will happen next
you said that we are in the phase of
some rather complicated development
how do we orient ourselves
on what we have to pay attention
what will be the biggest challenge
in the next weeks, months, years
now it is really interesting
because we have not even brought
the opportunity from the list of fall
2022
and a new version of GPT4 has already come
through
and another version
has started to generate
in the 5th version
and there are all realistic photos
are all free services at the moment
most of these services
start without a risk
and how nice
I will have some paid version
so that's a great question
because we also knew
from the beginning of the internet
now it goes much faster
also 90 years on the internet
everything is free
and then it started
to pay for some of these services
now the progress
is much faster
also it affects you
that it is relatively expensive
to produce the language model
and train the language model
that means
they need to finance something
OpenAI, which is the company
which is still not yet
so it actually
comes to the production of the GPT4
because it is also
getting a lot of data
that's what I would also like to know
that we are part of
training the next model
so people
how to share on the internet
just look at what I
called the GPT4
just look at it
I called it straight
or I called it to write for the president
and such things
so all this
affects
training the next version
of the model
which will be
even better
and if we can judge history
it will surprise us
how good it will be
now the most interesting thing
is that the language model
is connected
to other services
it is connected to the plugin
it is not yet available
but it will be
or of course
people will try to program it
using their own books
so it is called
Auto GPT
where the
model
will set the instruction for itself
then it will perform
and on the one hand
to give it the instruction
on the other hand
we will give it a goal
to achieve
if we compare it
instead of having a
Brigade
which we have achieved
and which we are
working on
and we have to sit on it
so that we do not do anything wrong
so we have now
tried a worker
who will admit
when he does not know
so he can ask me
and what is the main challenge
that is to set it
what will be the most important
it can be divided into
a philosophical area
which means
that I should
start to realize
what it means
to be people
what it means
that
something about which we thought
that we can do it by ourselves
that is
on the one hand a philosophical question
and on the other hand it will be very practical
because
we all do some things
advice and some things
we need to go down
or we are proud that we can do it
and now we find out
that something from that
can be very good
like a computer
maybe better than me
but it does not have to be better than me
but in it
and on the one hand a lot of people find
that a large part of their work
goes to automate
and we will have to deal with it
now
it is said that the more
you use
computer
the larger part of your work will go to automate
it is as simple as a pen
and suddenly we realized
that the computer does not have the creativity
or the possibility
to synthesize
from different sources
and then
to make some decisions
so it all
shows that it can be relatively good
but the computer
does not know how to
be responsible for its decisions
and that is the thing
that I would like to measure
I do not say
confidence or responsibility
but I think
in the past week
we had an ethical codex
Jirka Kubik, our principal
introduced it
and there he really expressed
the personal responsibility of the author
because the journalist
is one of the professors
who will be part of
the automation
and now it depends on
what we do with the journalist
and I think in the cases
that a member
is a completely responsible person
always
what is written in the article
a person has written to you
who is responsible for it
it does not mean
that the text
could not write the automatic
but we say that the text did not write the automatic
because
automation could not be responsible for it
and of course I hope
that you will
be able to reach the journalists
and they will look at
the web or the media
where they can trust
the authors, they can believe
that the author has the responsibility
and at the same time
we also said that we will use
these tools for our work
with the knowledge of
what are their limitations
what are their limits
and what are the benefits
I, for example, when I write a article
I usually
use various tools
to learn about a long
document or
to find other interesting examples
but every thing
depends on
the language model
that you can think of
without hesitation
and I as a journalist
or if you give me the right
then your right
will be that hand
and he
will help
himself
to use
various tools
so he believes and does not give
you something that is
thought out
Pavel Kasik, a technical
reporter with us in Prav. Pavel, thank you.
Music
And that's all
from the sharing of the podcast
559.
To all of our episodes
you can come back when I know
about the cases in all podcasts
applications and on the platform
podcast.cz
If you have any tips or ideas
for us, our address is
in a minute or six,
find us on social media
and Twitter
I'm waiting for you
tomorrow
Music
Machine-generated transcript that may contain inaccuracies.
Co den, to nová zpráva o vývoji a testování nových nástrojů umělé inteligence. Jazykové modely jako ChatGPT zažívají prudký rozvoj a začínají se testovat v nejrůznějších oborech včetně médií a marketingu. Odborníci ale varují, že se to děje na úkor chyb, nepřesností nebo dokonce rizika šíření dezinformací. Jsme na umělou inteligenci připraveni?
Host: Pavel Kasík - vědeckotechnický redaktor Seznam Zpráv
Článek a další informace najdete na webu Seznam Zprávy
Sledujte nás na sociálních sítích Twitter a Instagram. Náměty a připomínky nám můžete psát na e-mail zaminutusest@sz.cz