| The good and the bad of ML 09:06 - Feb 17 with 759 views | StokieBlue |
So a pharmaceutical company wanted to find molecules which would work against drug resistant E.coli so they trained a ML model to assist them in finding molecules which could work and be the lowest level of toxicity to humans. It found a number including one which was only considered for diabetes previously but was found to work against dozens of the most problematic strains of E.coli.
All good and impressive.
Their model essentially penalised toxicity for humans whilst rewarding therapeutic activity.
They were then invited to a conference on his ML could be misused in things like chemical or biological weapons research.
They hadn't considered it before they simple flipped the reward/penalty logic so that toxicity was rewarded in their model.
In less than 6 hours the model generated 40,000 molecules which were above the toxicity threshold set by the researchers. The algorithm correctly "designed" VX and many other known chemical warfare agents as well as some with predicted toxicity higher than any publicly available agents.
Pretty scary stuff and shows why usage of ML needs to be carefully considered in many fields, although there is little to stop a rogue individual or government training a ML algorithm to do something like this if they wished. Obviously it's not a weapon until the various other stages are done but speeding up all the R&D into the compounds in a 6 hour window is nothing like we have seen before.
https://www.theguardian.com/commentisfree/2023/feb/11/ai-drug-discover-nerve-age
SB [Post edited 17 Feb 2023 9:15]
|  | | |  |
| The good and the bad of ML on 09:14 - Feb 17 with 709 views | SuperKieranMcKenna |
I’m still not convinced of the potential of machine learning given Harry’s non-sensical scattergun posts. |  | | |
| The good and the bad of ML on 09:15 - Feb 17 with 702 views | giant_stow |
....Gawd.
There are so many ways in which we're potentially screwed, its almost mind-boggling. How long do you give the human race Stokie? |  |
| |
| The good and the bad of ML on 09:19 - Feb 17 with 670 views | StokieBlue |
| The good and the bad of ML on 09:15 - Feb 17 by giant_stow |
....Gawd.
There are so many ways in which we're potentially screwed, its almost mind-boggling. How long do you give the human race Stokie? |
Climate change is the biggest problem but it's not existential for the all the human race, just a good percentage of us.
Unless there is a pathogen/event with 100% mortality rate there will be humans around for a long time.
SB | | | |
| The good and the bad of ML on 09:51 - Feb 17 with 593 views | giant_stow |
| The good and the bad of ML on 09:19 - Feb 17 by StokieBlue |
Climate change is the biggest problem but it's not existential for the all the human race, just a good percentage of us.
Unless there is a pathogen/event with 100% mortality rate there will be humans around for a long time.
SB |
Reassuring in a funny sort of way. | |
| |
| The good and the bad of ML on 09:56 - Feb 17 with 587 views | Guthrum |
| The good and the bad of ML on 09:19 - Feb 17 by StokieBlue |
Climate change is the biggest problem but it's not existential for the all the human race, just a good percentage of us.
Unless there is a pathogen/event with 100% mortality rate there will be humans around for a long time.
SB |
We are possibly the most adaptable of creatures, thanks mainly to our ability to shape our environment and to scrape a living from the most marginal of locations (deserts, deep jungles, icy regions - so long as there's something to eat or it can be bought in). |  |
| |
| |
By continuing to use the site, you agree to our use of cookies and to abide by our Terms and Conditions.
