Artificial intelligence & the content writing revolution, really?
On March 14, 2023, I was invited by the SMX Paris to give my opinion and debate around the topic "Will AI revolutionize content creation in SEO?". Laurent Rignault moderated the discussion between Thomas Leonetti, Paul Sanches, Philippe Yonnetand myself.
In the first place, these elements are only the beginnings of a long reflection. The question of artificial intelligence and its applications in the creative field is a fascinating one. It renews our point of view on our professions.
As a service provider, Assonance is naturally influenced by the emergence of artificial intelligence tools accessible to the general public. First of all, we need to study all the opportunities brought by this technology. This is a research that we have been doing on a dilettante basis for several years now. We are testing all the automation devices for recurring tasks. We've already made a number of presentations along these lines, including one at SEO Camp'us on the industrialization of e-commerce content for demanding brands, and another on workflow optimization at the latest Women in Tech SEO Festival (UK) 2023.
Secondly, our role as consultants requires constant monitoring. We need to be able to answer our customers' questions: about using AI in their projects, and about the risks of the advent of AI that could also be used by competitors. The latest product launches such as ChatGPT and OpenAI's announcements about GPT-4or Google's announcement of Bard, encourage us to take a very serious look at this issue. The cards are being reshuffled not only for our own businesses, but also for all the markets we serve.
Do not reduce content creation to writing alone
I'm starting with this topic on purpose. Today, the arrival of ChatGPT has made many web marketers realize that we can indeed automate and/or delegate content creation to a machine. This idea is far from new, as authoring assistance tools are already making a name for themselves on the market. In the field of SEO, of course, tools like the ones we use, such as YourText.Guru, Babbar, Kill Duplicate and Semrush, are already providing us with plenty of data to shorten the time we spend in front of our screens wondering where to start.
But our work, and the time we spend striving to be more efficient and creative, also gives us the opportunity to divert tools from their original purpose. That's why we like to delve into information databases that we don't necessarily consult for the reasons you might think.
We, and all professionalWeb writing didn't wait for OpenAI to automate our processes and save time. I think we can even say that we didn't wait for the arrival of IT, if we look at the analogical organization methods that already existed. The diagram above shows an example of a project flow broken down into tasks. There are bound to be omissions or details to be added. We can clearly see the very small role played by the "writing" stage within this creative process. The preliminary phases of audit and strategic reflection have not been included in this diagram. I'd even go so far as to say that the "writing" part isn't always the most interesting.
Writing and obtaining a finished text are not the results of a magical, spontaneous action. As if all it took was a few minutes or a few hours at the keyboard to come up with something satisfying. For oneself or for a third party. The writer is rarely alone. Although they may lock themselves away in their office for a while, it is exceptional for them to be the sole decision-maker. This is a situation we might find in self-publishing, but only if we maintain the illusion that an author is never influenced by the opinion of another person, who in turn becomes a kind of contributor.
Image generated by artificial intelligence (source: https://yourtext.guru/)
That's why I recommend taking a step back from all content generation tools. On the one hand, writing the text itself is only a tiny part of the whole creative process. On the other hand, it would be a mistake to attribute to these (current) tools the ability to guess and/or merge all our needs and requirements.
Using these tools also begs the question "how much automation do I really need? I have the feeling thattoday we fantasize about omniscient artificial intelligences who could satisfy all our desires. There isn't a single step in the above diagram that we don't already carry out, at the very least with the help of a tool, software or, more simply, a script developed for a specific need.
We therefore need to differentiate the needs and capabilities of AI between autonomous and assisted creation.
Finally, there is the question of thestatus of creator. Is it only the user, the designer of an AI, the person who trains the AI(1 ) or simply the company that owns the AI? Or the user's employer? Traditionally, we tend to recognize an author or creator in general by the traces they leave in their creation. This is even how we can appraise works of art, or attribute works to artists despite the absence of an obvious signature. It is also these traces that help define a notion of originality, of singularity. It is the unique aspect of a creation that makes it so precious(2).
We could also try to differentiate in the writing of a content the levels of intervention of each actor:
the designer,
the publisher,
the editor,
the contributor,
the proofreader,
the integrator...
As with a film, where we have a director, a scriptwriter, a cinematographer, actors, editors... web content should probably be considered a collective work, especially with the assistance of AI.
But the workings of an artificial agency call for even deeper thinking. Specifically, it's about identifying the real source. Indeed, when we are traditionally in the research phase, we are supposed to link ideas together, give birth to new ones, process them and put them down in words. However, when the contents of a training set are used like ingredients in a recipe... Can the authors of the contents in the training set claim to be a co-author of the AI-generated creations?
There is a right to use resources available to the public for scientific research purposes:
"Union law provides for certain exceptions and limitations relating to uses for scientific research purposes that may apply to acts of text and data mining. However, these exceptions and limitations are optional and are not fully adapted to the use of technologies in the field of scientific research. In addition, where researchers have lawful access to content, for example through publication subscriptions or open access licenses, license conditions could exclude text and data mining. As research is increasingly carried out with the help of digital technology, the competitiveness of the Union as a research area is likely to suffer unless measures are taken to remedy the legal uncertainty surrounding text and data mining. "
At the same time, an exception and limitation of this use would be possible. For that, the hand is left to the owner of the rights of the content likely to integrate a training set (corpus of data used by the AI). The latter must signify its "opt-out" in a way that the robot can understand. "opt-out :
"Digital copies and reproductions made during text and data searches are stored with an appropriate level of security and may be kept exclusively for scientific research purposes, including for the verification of research results." [...] Without prejudice to the provisions of II, digital copies or reproductions of works that have been lawfully accessed may be made with a view to text and data searches carried out by any person, whatever the purpose of the search, unless the author has objected in an appropriate manner, in particular by machine-readable processes for content made available to the public online. "
What would a revolution in SEO content creation look like?
Between the one that is on the horizon and the onewe would like to see, is it the same revolution?
While some players hope to find a spontaneous and applicable solution to all their content issues thanks to AI, others fear for their profession and their material future. These questions are entirely legitimate. They are also issues that concern us, as service providers whose business revolves around editorial... and inevitably editorial.
We have decided to consider artificial intelligence as an additional tool. On the one hand, because we already use it to automate specific tasks or to broaden our horizons by accelerating research phases. On the other hand, because we realize that saving time on operational and repetitive tasks does not mean that we can work less on intellectual tasks. On the contrary, artificial intelligence or any automation process tends to give us more time to think strategically, and to design content. It also means more time for monitoring and prototyping, in order to move towards editorial innovation. This is why the automation gradually integrated into our processes since 2018 has even rather resulted in our rates going up, rather than levelling our offers downwards as some fear.
The costs of :
licenses (software, SaaS tools, APIs),
development of scripts to adapt these technologies to our own needs,
user training,
evangelizing to employees, customers and partners,
process adaptation (change management),
and surely other points I'm not thinking of now as I write these words.
For us, artificial intelligence acts as a springboard springboard for moving upmarket and design content with which we can be even more satisfied. From our point of view as an agency, economies of scale encourage the emergence of new ideas. Our human brains gain in availability. Thank you, dear machine.
You can also find more thoughts on this subject in :
an interview I answered for the Blog du Modérateur on March 09, 2023 on the occasion of SMX Paris 2023,
an article in Reacteur.com "Automating content writing: overview and perspective" February 15, 2021.
(1) Deltorn, Jean-Marc and Macrez, Franck, Authorship in the Age of Machine learning and Artificial Intelligence (August 1, 2018). In: Sean M. O'Connor (ed.), The Oxford Handbook of Music Law and Policy, Oxford University Press, 2019 (Forthcoming) , Centre for International Intellectual Property Studies (CEIPI) Research Paper No. 2018-10, Available at SSRN: https://ssrn.com/abstract=3261329 or http://dx.doi.org/10.2139/ssrn.3261329
(2) ADER Basile, " L'evolution de la notion d'originalité dans la jurisprudence ", LEGICOM, 2005/2 (N° 34), p. 43-49. DOI : 10.3917/legi.034.0043. URL: https: //www.cairn.info/revue-legicom-2005-2-page-43.htm
