This article is more than 1 year old
20,000 proteins expressed by human genome predicted by DeepMind's AlphaFold now available to download
Plus: Facial-recognition upstart Clearview raises $30m
In brief Deepmind and the European Bioinformatics Institute released a database of more than 350,000 3D protein structures predicted by the biz's AI model AlphaFold.
That data covers the 20,000 or so proteins made in the human body, and is available for anyone to study. The proteomes of 20 other organisms, from Zebrafish to E.coli bacteria, are also in there, too, and hundreds of millions of more structures will be added over time, we're told.
“In the hands of scientists around the world, this new protein almanac will enable and accelerate research that will advance our understanding of these building blocks of life,” said DeepMind’s CEO Demis Hassabis. He hopes that it will be a valuable resource that will be used in the discovery of new drugs and our understanding of diseases.
All of this depends on how accurate the AlphaFold’s predictions are, of course. It topped the protein-folding competition CASP last year. It is about 95 per cent accurate at figuring out the locations of the individual atoms in the protein within a precision of an Ångstrom for the most well-studied proteins, we’re told.
Investors are still giving controversial facial recognition Clearview AI millions of dollars
Clearview AI, the controversial upstart that was sent multiple cease-and-desist demands by tech companies, has bagged $30m in series B funding from investors.
Best known for scraping three billion photos from people’s public social media accounts to develop face-matching software, Clearview is often used as a case study why unregulated use of the technology may be harmful.
It is being investigated by data and privacy watchdogs in the UK and Australia, and Canada has cracked down on the upstart.
Talking to the dead through GPT-3
Here’s a gripping tale of a man who, stricken with grief over the death of his fiancée, decided to fine-tune a GPT-3-based chatbot to mimic her personality so he could talk to her beyond the grave.
Joshua Barbeau came across Project December last year, a website that allows anyone to spin up a chatbot, trained on supplied snippets of text, that writes in a distinct type of way. The application uses OpenAI's text-generation models GPT-2 and GPT-3 in its backend. People type in a sentence and the machine replies.
Out of curiosity, Barbeau gave it old texts and Facebook messages that his fiancée, Jessica Pereria, sent him during their relationship. Although he knew it was make-believe and the chatbot sometimes said things that were unJessica-like, there were instances in their conversation that felt surprisingly very real.
It’s interesting to see how he used the chatbot to feel better about her death. It's worth a read to find out what he did knowing that there was a finite amount of time he could spend talking to her before it was deleted. ®