A professor of neuroscience reveals the development of an artificial intelligence algorithm that he says could give humans "infinite perfect memory," in a move he sees as a potential leap forward in understanding human cognitive abilities.
Gabriel Kriman, an artificial intelligence and neuroscience researcher at Harvard Medical School, launched a startup company last month (called Engramme) with the aim of marketing this new technology, which he describes as a "battle against forgetting," as it allows – according to his claim – the storage and retrieval of memories without time constraints.
The idea is based on what is known as "big memory models," a concept inspired by "big language models" used in artificial intelligence applications such as ChatGPT, which are used to analyze and retrieve an individual's digital life data when needed.
In an official statement, the company's founders said that this technology could reshape many professions, from medicine and law to arts and engineering, by enhancing human cognitive abilities.
In a LinkedIn post, Kariman wrote that humanity has "fought against forgetfulness since the dawn of history," adding that, in collaboration with his partner Spandan Madan, he developed an algorithm that allows humans to remember everything, describing it as a "memory singularity moment" in which forgetfulness may cease after thousands of years of human suffering from it.
Despite these ambitions, the mechanism for retrieving memories remains unclear, although previous research has explored the possibility of using brain-computer interfaces to achieve this.
Engramme is seeking to raise around $100 million, according to Bloomberg, with a valuation estimated at around $1 billion.
This initiative is not the first of its kind; previous attempts to use artificial intelligence to recover digital memories have been made. Among these is the "Story File" platform, which allows the creation of interactive virtual versions of individuals based on their audio and video recordings.
Major technology companies are also working on similar ideas; Meta has obtained a patent that allows for the "simulation" of user activity through language models trained on his data, allowing him to be represented digitally even in the event of his absence or death.
