Scientific knowledge is supposedly something that nourishes all human beings to continue advancing, but the problem is that in many cases the articles that contain this knowledge are in tools that require a subscription to read them. This limitation in access to universal knowledge has led to the emergence of different platforms that bring together all these articles, such as Sci-Hubwhich now improves with his AI called Sci-Bot which promises to put an end to ChatGPT’s “hallucinations” in the scientific field.
How it started. At the end of this same month of April, a message on networks published by Mushtaq Bilal began to go viral, and no wonder, since it gave a notice in which, ironically, it invited us to use a new Sci-Hub tool that allowed access to scientific advances for free. Something they do through the back door and that already it almost cost them closure forced by the famous ‘Pirate Bay’
But logically this publication had the opposite effect, going viral, and also revived the eternal debate about the paywalls in science they can block access to this knowledge. But now Sci-Hub’s new tool has arrived to change this (partly).
A great library. To understand the magnitude of Sci-Bot, you must first look at the size of its brain, since since Elbakyan founded the web in 2011, Sci-Bot has become in a headache for scientific dissemination giants such as Elsevier or Springer, which are behind the publication of thousands of top-level articles.
Here, according to the official data of the platform itselfSci-Hub hosts 88,343,822 research documents and books, so we are talking about 100 TB of human knowledge covering more than 95% of the publications of the main scientific publishers. And with free access and without going through the checkout, as happens on the websites of some of these publishers.
The jewel in the crown. As Sci-Hub’s own page reveals, Sci-Bot is an AI that is designed to be able to search within the titanic database to select the most relevant studies and compose articulated responses. Its main attraction is that compared to generalist AIs like ChatGPT or Claude there are hardly any hallucinations, such as its creators pointed out in a scientific article in which tests were carried out in this sense.
And this is something very important because I have been able to experience with my own eyes how AI invents bibliographical references or assigns research to authors who have nothing to do with it. But Sci-Bot, being anchored to a real database from which it draws the information, means that there are direct references to the original papers, allowing users to jump over the hated paywalls to access scientific evidence.
Still needs improvement. At the moment it is starting in its alpha phase and that is why it has different limitations, such as that it can only answer one question at a time and does not maintain the thread of chained queries, even if they are on the same topic. But the truth is that it is quite promising to have access to the vast majority of human knowledge.
They put obstacles in his way. Here, logically, the magazines have a lot to say, since they do not like having the articles freely available when they request a subscription to access them. This means that right now Sci-Bot has the most recent scientific articles as its blind spot, since due to the new and aggressive security measures implemented by large publishers in recent years to avoid scrapingthe database has some gaps in articles published in the most recent months. This makes the AI unable to respond regarding the most recent evidence.
But without a doubt we are facing an advance that began with the arrival of Sci-Hub with the promise of democratizing science, although through the back door by freely publishing articles that are actually ‘private’. And the only thing this will do is create a new front between open access and large publishers seeking financial returns.
GIPHY App Key not set. Please check settings