A Harvard neuroscientist has launched a startup claiming to have developed an AI algorithm that gives humans “perfect and infinite memory”. Gabriel Kreiman, who researches artificial intelligence and neuroscience at Harvard Medical School, says the technology will transform cognitive capabilities by allowing memories to be stored indefinitely.
Kreiman describes his work as a “fight against oblivion”. The system uses “large memory models”, a twist on the large language models (LLMs) behind tools like ChatGPT, to retrieve data from a person’s digital life. The startup, Engramme, was founded last month with co-founder Spandan Madan.
In a manifesto on the company’s website, the founders claim the technology will reshape every profession from medicine to engineering. “Humanity has been fighting the problem of forgetting since the dawn of time,” Kreiman wrote on LinkedIn. “This is the MEMORY SINGULARITY: after 300,000 years, this is the moment that humans stop forgetting.”
It remains unclear how memories would be recalled, though previous studies have explored brain-computer interfaces. Engramme is seeking around $100 million (£74m) in funding, with a valuation as high as $1 billion, according to Bloomberg.
Other firms are also working on AI-driven memory. StoryFile offers virtual avatars trained on personal videos, while Meta was granted a patent in December to simulate users via LLMs trained on their activity, even after death.
