Biocomputing: the promise of biological computing

Advances in neur­os­cience are pav­ing the way for the con­struc­tion of bio­lo­gic­al com­puters using neur­ons and brain tis­sue. This inven­tion was both a break­through for com­put­ing and a tool for fun­da­ment­al research and the study of human diseases.

They are some­times called mini-brains. These cereb­ral organoids are in fact 3‑dimensional cul­tures of cereb­ral tis­sue. They could lead to the next revolu­tion in com­put­ing, the birth of organoid intel­li­gence (OI).

As a remind­er, and con­trary to what its name sug­gests, arti­fi­cial intel­li­gence (AI) does not mim­ic human intel­li­gence. In fact, it works in a com­pletely dif­fer­ent way. Just watch it play chess: it sac­ri­fices far more pieces than any human play­er¹. Anoth­er dif­fer­ence is its energy con­sump­tion. In June 2022, the record capa­city of the Fron­ti­er super­com­puter, housed at Oak Ridge Nation­al Labor­at­ory in the United States, was 1.1 exa­flops (i.e. it per­forms up to 1.1 1018 oper­a­tions per second), a per­form­ance equi­val­ent to the human brain. Except that the human brain oper­ates on 20 Watts, where­as the Amer­ic­an super­com­puter requires 10MW…

Organoid intel­li­gence (OI), on the oth­er hand, prom­ises to bring the two sys­tems closer togeth­er, by repro­du­cing the com­pu­ta­tion­al per­form­ance of the best com­puters and the energy effi­ciency of a human brain. This tech­no­logy is being developed at the cross­roads of three tech­no­lo­gic­al break­throughs: elec­tro­physiology, arti­fi­cial intel­li­gence, and cereb­ral organoids.

The first is essen­tial for com­mu­nic­at­ing with cereb­ral organoids. The chal­lenge? Find­ing a non-invas­ive sys­tem that takes into account the mul­tiple elec­tro­chem­ic­al sig­nals that occur every second in the small mass of tis­sue in cul­ture. Research­ers at Amer­ic­an uni­ver­sit­ies are pro­pos­ing the use of cage-shaped elec­trodes², an ini­tial solu­tion for dir­ect com­mu­nic­a­tion with the organoid in culture.

The devel­op­ment of the bio­lo­gic­al com­puter is still in its infancy.

The ques­tion of elec­trodes is also import­ant for the devel­op­ment of brain tis­sue. Without sig­nals, cereb­ral organoids can­not be con­struc­ted in three dimen­sions and remain inop­er­at­ive. Nerve tis­sue has to work to func­tion and brain mech­an­isms require more than a simple elec­tro­chem­ic­al sig­nal. Memory involves the reor­gan­isa­tion of neur­on net­works and the inter­ven­tion of oth­er brain cells, such as those that make up microglia, the immune sys­tem of the brain. To devel­op, organoid intel­li­gence must integ­rate all these parameters.

The second tech­no­lo­gic­al break­through that makes OI pos­sible is AI. It is essen­tial for explor­ing what these bio­lo­gic­al sys­tems can do. Brain tis­sue cul­tures pro­duce very large volumes of data, both spa­tial­ised and struc­tured over time. Inter­pret­ing this data is a chal­lenge that recent advances in algorithms will be able to meet. 

The last tech­no­logy on which the devel­op­ment of OI is based has yet to prove itself, and that is the one that will enable brain organoids to change scale. At present, the largest brain tis­sue cul­tures meas­ure just a few mil­li­metres and con­tain a max­im­um of 15,000 neur­ons. Mak­ing them big­ger means pro­tect­ing them against oxy­gen depriva­tion, to which neur­ons are very sens­it­ive. To do this, they need to be per­fused, to ensure that each cell is con­nec­ted to a sup­ply of oxy­gen and nutri­ents, just as the blood­stream does in vivo.  Micro­fluidics seems cap­able of provid­ing this per­fu­sion. But this tech­no­logy has not yet been trans­posed to cereb­ral organoids. This tech­no­lo­gic­al trans­fer will make it pos­sible to go from a cul­ture the size of a fly’s brain to one com­par­able to that of a mouse.

Will these sys­tems replace our com­puters? Not in the near future. In 2019, a Japan­ese team suc­ceeded in get­ting two cereb­ral organoids to com­mu­nic­ate³. It is dif­fi­cult to pre­dict what will hap­pen next, giv­en the large num­ber of teams work­ing on the sub­ject. Nev­er­the­less, there is hope that the first OI sys­tems will be tools for neur­os­cience research. Bio­lo­gic­al com­puters could help to explain how a brain man­ages to pro­cess incom­plete inform­a­tion, for example.

They could also help to elu­cid­ate the mech­an­isms of demen­tia, Asperger’s syn­drome, or oth­er com­mon human brain con­di­tions. It is cur­rently very dif­fi­cult to have accept­able labor­at­ory mod­els for these issues. Eth­ic­al con­sid­er­a­tions, for example, nat­ur­ally for­bid study­ing the effect of a molecule that dis­rupts memory on humans… Cereb­ral organoids offer an altern­at­ive for such research.

OI also raises eth­ic­al ques­tions in its own right: Which sys­tems could be at risk of suf­fer­ing pain? What pro­to­cols will be used to assess the intel­li­gence of a cul­ture? These issues need to be anti­cip­ated, and OI spe­cial­ists have com­mit­ted to incor­por­at­ing them into the devel­op­ment of tech­nic­al aspects by sign­ing the Bal­timore Con­ven­tion in 2022⁴. The devel­op­ment of the bio­lo­gic­al com­puter is still in its infancy.

Agnès Vernet