ChatGPT: "Faced with technological upheaval, Europe must not be afraid of disruptive innovation"

Victor Storchan is an entrepreneur and artificial intelligence engineer

.

The prowess of ChatGPT is certainly technological and scientific, but it is first and foremost that of the successful deployment of the disruptive technology underlying this conversational agent.

It therefore constitutes a valuable case study for European standards policy.

To be able to absorb the incessant technological upheavals, the production of the standard must arise from intimate experience in the field rather than from the technocratic skills of the Brussels laboratory.

The success of ChatGPT lies singularly in OpenAI's inherently empirical handling of the fine balance between validity and velocity.

Validity of the technology by controlling the uncertainty and risks inherent in breakthrough innovation when it leaves the laboratory and the speed of its deployment, it

Read also Mathieu Bock-Côté: “Artificial intelligence and falsification of the world”

This success invites the European institutions of reference (legislators, regulators, standardization bodies, etc.) which overlook the technological ecosystem to agree to replace part of the technocratic decision-making with a system of broad responsibility towards the community of entrepreneurs, developers and researchers.

ChatGPT is the illustration in action of the fact that these actors in the field benefit from a unique position to embrace

the technological, operational and organizational complexity required to deploy reliable and trustworthy AI systems (embodied by the model itself but also by the intuitive user interface, the efficiency of updates or the measured communication company in particular).

Admittedly, ChatGPT users have noted errors (hallucinations or inaccurate information) but the overall robustness of the service's deployment is the Californian start-up's biggest differentiator.

it is the first of its kind after the failures of Meta in 2022 (Galactica) or Microsoft in 2016 (sexist and racist chatbot).

Regulator and institutions consistently lag behind the brutal and unpredictable paradigm shifts with which AI history is replete.

Victor Storchan

Europe has lost the habit of building breakthrough technologies (cloud, AI, synthetic biology, etc.) and deploying them on a global scale to derive the exponential economic gains they induce while assuming responsible for the hazard of the risks inherent in innovation.

By de facto abandoning the development of AI and products to other powers, Europe is significantly weakening its ability to understand on the ground the unique management of multiple sources of uncertainty and the consubstantial consubstantial with the deployment of a technology. disruptive like AI.

By preferring to situate its action in the regulatory field, Europe risks locking itself into a rigid, vertical and centralized framework.

With the texts of the IA Act or the DSA,

it aims to export its technological standard by relying on the strength of its domestic market.

It enacts binding and cross-cutting rules governing the deployment of AI and the use of products via a mechanism that is certainly well-oiled but extremely vertical and centralized: the appointment of the regulator with its group of associated experts, the vote by parliament, standardization by designated bodies, transposition by Member States, etc.

Read alsoChatGPT: “The great threat”

With AI, this logic of rule centralization comes up against the demanding reality of technological brilliance.

Regulator and institutions consistently lag behind the brutal and unpredictable paradigm shifts with which AI history is replete.

Note, for example, that the OECD definition of AI does not contain the notion of generative AI, which is at the heart of ChatGPT.

Whether for AI or for anti-Covid vaccines, the time that elapses between the research phase and the dissemination of innovation in society through commercial applications has been considerably reduced.

All the kept proportions,

the Suffren submarine requires 30 years of design for an equivalent lifespan when 5 days are enough for ChatGPT to acquire a million users.

Society's ability to develop its social “antibodies”, that is to say to debate the impact of technology, to become aware of the changes taking place, is consequently affected.

As a result, the intentionally incremental deployment of ChatGPT first as a restricted demo and then as a controlled (API) service not only allows OpenAI to test its technology in a secure

but also to society as a whole to reflect on the transformations, opportunities and risks of such technology (education, health, human rights, etc.).

ChatGPT proves to us that Europe must think of an event policy favorable to breakthrough innovation

Victor Storchan

It is therefore fundamental for Europe to accept the element of uncertainty inherent in disruptive innovation and not to suggest that incomplete knowledge of the future effects of AI is intrinsically negative.

Uncertainty is also a tremendous source of potential.

In his book

, the Dutch diplomat Luuk Van Middelaar describes the ability of the European machine to improvise in order to face the unforeseeable consequences of financial or migratory crises, to adapt in real time by moving from the policy of the rule, a brilliant factory mechanism of a consensus in a given and fixed system, to the politics of the event, the art of having a grip on unforeseen and unprecedented events.

ChatGPT proves to us that Europe must think of a policy of the event favorable to breakthrough innovation, otherwise it is doomed to take on the role of timid moderator in the technological confrontation between the great powers.