The Explosion of AI Models and the Repudiation of Startups

At the heart of artificial intelligence (AI) development lies the desperately simple yet profound realization: AI is the product of the sum of its data. AI laboratories acquire data from various sources, but a recent phenomenon has highlighted a controversial practice: the acquisition of data from defunct startups to feed AI models. This practice has serious implications for culture, artistic creation, and human thought, pushing us to re-examine our ethical considerations regarding intellectual property and respect for the individual.

The Exploitation of Startups’ Digital Archives

When Keelan Miskell decided to shut down his crypto startup, he didn’t see his data as a significant asset of value. The users were anonymous, and the startup was only a few years old. However, the company transformed its digital archives into a substantial financial capital. Miskell earned around $25, 000 from his source code, and the future looks promising for other lots still in negotiation.

Ethical and Philosophical Implications

The acquisition of data from disappeared startups raises crucial ethical and intellectual property questions. If anonymous users of these startups inadvertently contributed to the creation of this data, what rights do they have over these data once the startup is liquidated? The principle of “property of data” raises deep questions about privacy protection and the management of personal data.

The Circle Against Artistic Creation

The culture and creation of art are also impacted by this practice. Artists and creators often use raw data to fuel their works, from music to digital art. The acquisition of data from vanished startups questions the origin and legitimacy of these data. This can lead to tensions between creators and AI labs, which seek to optimize their models.

The History of AI Development

By examining the history of AI development, we gain a better understanding of the importance of this practice. The early AI models relied on large and diverse datasets, often collected in opaque ways. Over time, transparency and legality began to guide data collection. However, the current practice of acquiring data from disappeared startups represents a step backwards in this ethical process.

Language Model Confrontation

Language models, such as GPT-4, are exemplars of this practice. These models are trained on a variety of texts and content, including data from vanished startups. Their performance is impressive, but the transparency and legitimacy of the training data are often overlooked. A link to the article of Technology Review.

Conclusion: An Ethics for the Next Generation of AI

In conclusion, the acquisition of data from disappeared startups poses significant ethical challenges. It is crucial to develop ethics and regulations that protect user privacy and intellectual property, while still enabling the advancement of AI. The AI community must reflect on these issues to ensure that AI contributes positively to society and respects the rights of all.

Cet article est rédigé par Blaise Grenier, Journaliste reporter d'images et auteur. Découvrir le parcours de l'auteur et la charte éditoriale »