[ExI] ‘The game has changed.’ AI triumphs at solving protein structures

[Rafal Smigrodzki]

On Tue, Dec 1, 2020 at 4:11 PM spike jones via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

>
>
>
>
> *…*> *On Behalf Of *William Flynn Wallace via extropy-chat
> *Subject:* Re: [ExI] ‘The game has changed.’ AI triumphs at solving
> protein structures
>
>
>
> >…Do y'all think that this is the best thing AI has ever done?  bill w
>
>
>
> Entirely possible that it is the most significant advance AI has made with
> respect to how it will impact humanity’s future.
>
>
>

### Some caveats are in order - the errors in the AI-simulated structures
were on the order of 10 nm. This is not yet a practical solution of the
protein folding problem, since to be able to use a structure to correctly
predict ligand binding, enzymatic function and other functional aspects you
need an accuracy of a fraction of a nanometer. Still, it's a huge leap
forward and quite reminiscent of the first AphaGo and other game AIs -
going from nothing to middling (protein folding is here now) to very good
to smashing the best human to vastly superhuman in just 3 - 4 years. It
everything goes well we might indeed have an AI capable of whole-cell
predictive metabolic and maybe even structural modeling using only genome
in the next few years.

If this capability became available it would be the biggest inflection
point in the whole history of biology and medicine, greater than
antibiotics, anesthesia and any other single advance. It would allow
stunningly fast development of highly advanced pharmaceuticals,
biotherapeutics and implantable devices, orders of magnitude faster than
now.

Old-timers here may recall that twenty-something years ago on this list I
proposed the solution of the protein folding problem as a way for an UFAI
to quickly defeat humanity. The UFAI could use the predictive capability
afforded by protein structure modeling to create both a self-replicating
computational substrate to live in and the self-replicating weapons (or
bioweapons) to kill us. Looks like no UFAI is needed to solve protein
folding.....  hopefully AlphaFAI, the grandson of AlphaFold, will be here
to protect us when the UFAI is (almost inevitably) created.

Rafal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20201209/c4332b2d/attachment.htm>