While thousands of kids ask the
GPT chat
to write their practical school work and
fashionista Pope Francis
memes created with apps like
Midjourney
emerge , a handful of top
AI
experts and industry executives, including
Elon Musk,
raised a cry in the sky today and called for
an "immediate" six-month pause
on the advancement of those systems.
Because?
They now say
they are risky
for society and humanity, according to an open letter they signed.
This person does not exist: we asked Midjourney to compose it by adding a series of commands about gender, age, origin and other characteristics.
We have lived surrounded and assisted by her for a long time, but since last December her presence has become more powerful and sophisticated.
She already wrote, drew, animated and "reanimated" objects and people, but now she
seems capable of creating them
.
Artificial Intelligence
(
AI) has finally given birth to a new generation -called "generative" for its ability to conceive original content from the remix of pre-existing materials-, propelling in an amazing way the possibilities that we had already glimpsed: at
concerts holographic images
of Maria Callas and Elvis Presley, the increasingly perfect memes on
smartphones
and
applications
that
age
or
they childish
our voice and
physiognomy
.
Ayr Disney presented to society a
new breed of robots and virtual toys
that in a very short time will be able to share our spaces, so real.
The Tinkerbell fairy will accompany us in the shower.
Diego Maradona and Pelé greeted Messi on Monday
with credible voices;
Maradona can interact with attendees in the Tango D10S, and Pope Francis wears a white down jacket: these are some examples of what devices like the MidJourney and Dall-e can create based on a series of instructions.
But beware, its performance varies enormously depending on the skill and
know-how
of its operator.
Thus, there is already a whole
army of "image engineers" or "prompt engineers"
, experts in uploading the most precise instructions and techniques for the new methods of assisted creation.
collaborations
Creative human tasks already seem headed for a type of
collaboration with AI programs
that -apart from ethical and legal controversies of all kinds- has quickly gone from simple assistance with a draft and express classification of content to the expansion of an
artistic avant-garde
that may soon be the norm.
To create this industrial elephant, we need about twenty indications because the performance of each application varies enormously depending on the expertise and know-how of its operator.
In a few months,
distrust
of the advancement of these
technologies
-in the production of content- is already creeping into legal and media discussions.
The reason is the
unusual speed
of its progress and improvement to limits in which even the very concept of what we understood as "
reality
" is being reconfigured.
In this matter, everything is dizzying: the image requested a week ago has greatly improved thanks to those who continued to run the same App. The first fear had, once again, dystopian rims.
In December 2022, when the startup
OpenAI (funded, in part, by Microsoft) made
the ChatGPT tool
available to the general public
, the old dilemma about the threat that robotization poses to the “human” labor market gained new resonance.
Given the abilities that this new generative AI app demonstrates (with which
you can chat, ask questions, or ask it to write essays
and explain complex concepts), the
new ones threatened
by the technological stampede would now be those who work in trades or tasks.” creative” and content selection, such as journalists, curators, editors and designers.
The speed with which
translation applications
have improved their performance based on what they learn, even in very challenging cutting-edge texts,
today dramatically reduces the volume of work
that "human" translators receive.
And meanwhile, students, in almost any field of knowledge,
believe their long hours of research have been resolved -or at least greatly abbreviated-
: a kind of omniscient PanWikipedia has arrived that is also capable of writing.
Although many voices rushed to clarify that
ChatGPT does not "believe" in the strict sense
, others were added to that first impact, very worrying in this era of globalized economies and sensibilities.
What would a field of sunflowers be like as Van Gogh thought, but in the style of Antonio Berni?
Miodjourney says so.
Texts, photos and works of art?
Perhaps one should start by saying that
ChatGPT
, essentially a
linguistic tool
, is not alone.
Other generative AI, such as Stable
Diffusion, DevianArt, Midjourney or DALL-E
apply similar computer models and procedures, but no longer for the production of texts but
images
.
If in the 90s of the 20th century the data search engines on the web (such as the primitive AltaVista and Netscape and then the omnipresent Google) became a first computer vector of the infinite ocean of navigable information
on the web
, the 21st century finds in generative AIs a new classification phase that consists of detecting patterns and
recombining information to produce “new” knowledge
.
The images of Pope Francis, according to Pablo Xavier.
In other words: under certain directives, a generative AI selects, assembles and “polishes”, with much more refined parameters, all the information available on the network, to yield new and increasingly accurate results.
A generative AI selects, assembles and “polishes”, with much more refined parameters, all the information available on the network, to yield new and increasingly accurate results.
Here it is worth making a clarification.
Since in 1957, at MIT at the University of Massachusetts, the scientist John McCarthy coined the term "Artificial Intelligence" to refer to
the science of "making intelligent" machines
, the collective fantasy that one of them imitates or reproduces the ways in which the human being learns or creates has caused more than one sting.
Although
we have lived with AI for a long time
(Amazon's Alexa or Apple's Siri smart assistants are nothing else), these new chatbots are presented as
battering rams of a disruptive future
that their predecessors did not bring to the fore.
Now, we are told, they have the power to "create."
However,
the precautions of yesteryear still apply
.
By replacing the biological process of creation, typical of the human brain, with a series of algorithms that make increasingly complex associations, the machines
continue to be far from being able to imitate the divergent thinking
that characterizes our species.
OpenAI and the ChatGPT logos turned traditional pedagogical strategies around the world upside down.
In New York, educational institutions are prohibited from accessing it.
REUTERS/Dado Ruvic/Illustration/
Even so, the distances are shortening at a dizzying rate,
the algorithms learn and jump every week
... It is enough to give a series of instructions or parameters to DALL-E or MidJourney -the Apps used to illustrate these pages- to verify the
impressive ability of these tools
to produce high-impact images, some of which may pass as “original” and give the impression that these chatbots are actually creating their own works.
The reality of the matter, of course, is more complex.
Strictly speaking,
they are remixing already provided footage
into a new unpublished collage as per our request.
This is not a Titian, although we asked Midjourney to "paint" an apocalyptic picture like him.
Create or copy, that is the question
While millions of people around the world have fun playing and experimenting with these new generative AIs, many of these developer companies -which are the technology giants-
are sued and prosecuted for massively violating intellectual property protection laws
.
As a collateral and extra-legal effect, some "Class Actions" - as lawsuits that bring together a group of victims to represent a much greater damage caused to society are called in the US - served, in turn, to clarify the way in which that these chatbots “create”.
Stability AI, the developer of the "Stable Diffusion" tool,
was sued for infringing copyright laws
when it was found that it had been "trained" using databases from LAION, a German company that has been archiving millions of images without paying royalties of use to the artists responsible for their creation.
This is not a Titian, either: Midjourney offered mixed results to our request for a Titian-style apocalyptic image.
To the extent that generative AIs
make use of the recombination of pre-existing images
to "create" new ones (through a complex process of "noise" and rarefactions applied to the original image, called "diffusion"), already it is easier to understand that
what appears to be an original creation
of the tool
is, neither more nor less, a sophisticated digital collage based on images, texts and information available on the web
, but whose true authors (let's say, from its "base paste ”) are made invisible and, therefore, violated in the economic rights related to the exploitation or use of their works.
What appears to be an original creation of the tool is, neither more nor less, a sophisticated digital collage based on images, texts and information available on the web.
The lawsuits erupted
at the end of January and the effect of their revelations soon spread geographically throughout the United States;
now they threaten to move to Europe.
Deepfake Pope Francis.
As the lawsuit against Stability AI continues in Delaware, the New York Copyright Office
refuses to issue intellectual property certificates
for fully digital comics created by the MidJourney engine, and whose "human" authorship cannot be proven.
By establishing that legal protection will not apply to aspects or "parts" of the work created by these AIs, they are endorsing the arguments that support the lawsuits filed in court: to the extent that these engines were trained using
images and data obtained from gigantic databases
, the originality of the artistic result produced is doubtful.
If some specialized publications estimate, for example, that DALL-E has been fed more than
half a billion referential images
, it is clear that the legal door is just closing ajar.
While the new phase of technological development rearranges once again our capacities for wonder -and
ChatGPT passes exams for Law and Medicine courses in the US and Spain-
, some
governments begin to take action on the matter
to avoid greater evils and try to regulate the advent of this first generation of generative AI.
Elon Musk is one of the more than a thousand signatories of a public letter in which businessmen and experts from the technology sector ask for a six-month stoppage in the development of AI: REUTERS/Dado Ruvic/Illustration
The European Union (which already launched a first
"Artificial Intelligence Law"
in 2021) intends to regulate them, limiting their use in the selection of personnel or in intelligence tasks.
The Joe Biden administration
in the US presented an "AI Bill of Rights"
to regulate its use and preserve the rights of citizens within increasingly autonomous computerized systems governed by non-human intelligence.
The New York Department of Education has
banned the use of ChatGPT on its networks and devices,
as have some financial institutions in the country.
As part of the new cold war between the US and China, the latter has already begun to block devices and apps that lead to GPT chat, just one of the many devices that promise to revolutionize the ways in which we create art and knowledge.
Deeper still, the very concept of “
originality
” is finally diluted.
China's quick reaction is neither accidental nor devoid of strategic sense.
Ernie, the AI-powered chatbot from Badiu (the “Google of China”) promises to add to this cultural, economic and political fabric soon, as Google, Microsoft and other techno companies warn Biden that excessive regulation of generative
AI
can give their eastern rival (not constrained by the institutional demands of Western democracies) a paramount
advantage
in the technological escalation that has been facing them for years.
A response from ChatGPT, an AI chatbot developed by OpenAI, is seen on its web page in this illustration image taken February 9, 2023. REUTERS/Florence Lo/Illustration
But this Wednesday, more than a thousand leaders in the sector have called for an urgent stoppage.
Among the more than a thousand main signatories, according to the Catalan newspaper
La Vanguardia
, are Elon Musk, the CEO of Stability AI, Emad Mostaque, DeepMind researchers -owned by Alphabet-, Apple co-founder Steve Wozniak, Professor Yuval Noah Harari and industry leaders such as Yoshua Bengio or Stuart Russell.
As they agree (and have published) "powerful artificial intelligence systems should only be developed once we are sure that their effects will be positive and their risks will be manageable."
And they anticipate that these systems could generate economic and political disruption, for example.
For this reason, now it is they themselves who are
demanding regulatory measures
.
look also
The creator of the viral deepfake of Pope Francis told the truth behind the image
look also
Elon Musk crossed Bill Gates: "His understanding of artificial intelligence is limited"
look also
Elon Musk and more than a thousand experts ask to pause advances in artificial intelligence: "There are great risks for humanity"