Investor, serial entrepreneur and network philosopher Naval Ravikant wondered how it was possible that no company had taken all the academic articles from Sci-Hub, created a great language model and not “built the best doctor in the world.” world”.

Sci-Hub is a website that contains around 90 million freely accessible academic articles and scientific documents, of which more than 25 million are medical. Ignoring the fact that the articles have intellectual property that Sci-Hub does not respect—it has servers in Russia—the question makes some sense.

If AI algorithms trained with millions of texts extracted from the web can decide the next word based on the previous ones, why can’t these same algorithms decide a treatment? At the end of the day, a treatment is a text generated based on symptoms—expressible in text—and all the prior medical knowledge contained in Sci-Hub texts. We train a model with the documents and we already have the best doctor in the world. Sure?

If you are bothered by the part about using content without respecting intellectual property, you should know that ChatGPT q or image generation models such as Midjourney or Stable Diffusion do precisely this: generate content based on texts and images extracted from the web, many of them. them with intellectual property rights. Q

Specifically, the Associació Professional d’Il·lustradors de Catalunya has written a manifesto (available at apic.cat) denouncing that “certain uses of AI can be a threat to the professional illustration sector.” The Aitana Bonmatí that Barça generated with AI to congratulate the player are an example of this. The fear is justified and the manifest one is necessary.

But this is not so. Doctor and supermarket cashier jobs are both exposed to AI (about 60% of our jobs according to the IMF). But while the former benefit from it; with diagnostic, prediction or simulation tools; the seconds do not. AI complements the capabilities of doctors and replaces those of cashiers. Same level of exposure, different level of complementarity.

On the other hand, there is an issue of social acceptance: we all feel more or less comfortable selecting a menu on a giant iPhone at a McDonald’s, but would you like to receive your diagnosis via WhatsApp from the “best doctor in the world”?