With the latest incarnation of AlphaFold, researchers at DeepMind and sister company Isomorphic Labs – both overseen by cofounder Demis Hassabis – have mapped the behaviour for all of life's molecules, including human DNA.
"Fuzzy Cumulonimbus Cloud"
Can ChatGPT tell me how to make supermeth or what
that means nothingggggggggg
will have to read their actual paper about it to see what it actually does
by the way I am working on a whitepaper I'm trying to publish tentatively called "I want to cure cancer with LLMs: how to scope your AI use case". Hardest thing will be how to not make it very insulting in tone, but I did recruit a coauthor (but she's also mean, so)
Steam, LoL: credeiki
+13
Options
WeaverWho are you?What do you want?Registered Userregular
edited May 9
laying in bed watching tiktoks and suddenly it's Kyle Mclaughlin and Lara Flynn Boyle at the 1990s Emmys
edit ok @Neco the above was followed up closely by a random Shelly Duvall appearance this is wild
If you do a search for AlphaFold server the search results show the website has existed for only 12 hours. An age check of their main site gets no results.
The main site links to a couple of the papers from a few years ago but that field's beyond my ken.
The job I currently do is too stupid to replace with AI. It’s kind of why I took it.
Professional Penny-Arcade Poster?
Ha, no I’m pretty sure an AI could do that quite easily.
I support government acquisitions, but because nobody in the chain wants to trust anyone else in the chain, people review each piece of it. And none of those reviews is automated, and they are making it even less automated because of some circular self dealing in the current process..
old-timey fisherman wisdom was that it was foolish to learn how to swim because then it would just take longer for you to drown
One year when I was in scouts, we had an instructor from idk Norway or Finland or such and his dad was a fisherman of some sort.
And we were talking about survival stuff and he said "the ocean is so cold, they tell you if you fall into the water don't try to swim"
With some further explaining, we found out you are supposed to just float to keep from losing extra heat waving you limbs through the water. Not just let yourself drown.
so right what I want to get across in this whitepaper is a few things, among which: yes, biotech is futuristic and AI is futuristic. You do realize that like...that doesn't mean AI solves biotech, right?
AI does specific things. Any given biotech problem involves some specific steps. You gotta ask some actual fucking working scientists what those steps are, and then figure out if AI will in any way make them faster or better (or enable you to circumvent the steps or paradigm shift but let's be real if you have to ask the scientists what the process is you do not have the knowledge for a paradigm shifting ai concept) and you have to be ready for the answer to be no. Like. I just read some garbage Time article someone posted at work being like, the age of biotech AI is now! Biotech uses CRISPR, so AI will solve that! And it's like. What the fuck do you think CRISPR is? (it is a biochemical process where you put some proteins and some nucleotides in an animal. You design what nucleotides in order to decide what specific part of the genome you will edit. You are editing the genome because you are looking to create a mouse with a specific genetic condition. You are making a mouse with a specific genetic condition for any number of reasons but you have chosen which genetic condition in accordance with your specific experimental goals (e.g., design a fat mouse so you can test diabetes drugs)) Ok now that you've taken the time to learn in the just vaguest terms what CRISPR is, why do you think AI would help do this in any way? like what part of the process would AI do...? Actually, maybe you better go ask someone who understands AI what AI can do, cause otherwise you won't be able to answer that either.
What is your actual goal? Your goal can't just be 'introduce AI to my org'. Like what are you actually fucking trying to do? You want to cure cancer? no you do not want to cure cancer, that is not a real thing. Maybe what you mean is you are looking to find a small molecule that inhibits a protein involved in a common cancer pathway? That sounds more like a real thing. How do scientists go about that? Well, one thing that they might do is virtually throw a bunch of small molecules against 3D structures of proteins, and see if any of them nestle in nicely. But, if you don't have a 3D structure of a protein you care about, you can use alphafold to generate a plausible one, most of the time. Hey that's AI, good job! And hey honestly, if you want to identify some promising proteins in the just absolutely horribly cluttered bazillion cancer papers out there, could to worse than to ask chatGPT to give you a list. Hey you used AI again, good job! Ok cool so now you did this docking simulation and used AI twice, do you have a drug? Lol no now you have to test it in cells and in mice and in people. How are we gonna use AI to do that? ummmmmm I mean. You could use a pipetting robot...
--anyway
i have feelings
about this
hopefully I will write this all out and get it published in our stupid consulting magazine, we will see. I have a whole like categories of what AI does and where in the process and how you can break your process down and see if any of the categories make sense. maybe I'll make some ~process graphics~ and such
I read some papers recently about using neurostimulation to disrupt the feedback loop between the cancer cells and the normal neurons, because electrical activity makes the glioblastoma grow faster. It does seem to increase survival time although it is not curative. Maybe see if there's any clinical trials you dad's eligible for? I don't totally know how to do that (getting patients for clinical trials is a valid use of AI in a drug development process, also, like if you have a big pharmacy or health system dataset)
Im thinking about writing a sort of postmortem of my experience in the last 5 years about best practices and how to avoid pitfalls of tightly coupling /calcifying incompletely specified semi-whitebox test environments during certain kinds of tentative exploratory design work based on the projects Ive been working on (most papers about it are toy projects so itd be a good internal thing and might get me some attention) but Id have to like find a work audience and do some research on existing work or writings????
Revolutionary new medical study encourages patients to stop being poor
This is a factor that almost all of these studies ignore and it's kinda infuriating.
"We studied a bunch of non-affluent people (who only signed up for the study because they needed the money) and we were shocked to find they're in poorer health than affluent people (who eat far smaller amounts of processed foods because they can afford to and also can afford to do things like regular checkups with their doctor because they fucking have a doctor and not whatever person happens to be on duty at the one urgent care that their HMO actually lets them go to on the other side of the city)"
Im thinking about writing a sort of postmortem of my experience in the last 5 years about best practices and how to avoid pitfalls of tightly coupling /calcifying incompletely specified semi-whitebox test environments during certain kinds of tentative exploratory design work based on the projects Ive been working on (most papers about it are toy projects so itd be a good internal thing and might get me some attention) but Id have to like find a work audience and do some research on existing work or writings????
who writes papers anyway, eggheads?
I've been told that in order to get promoted a requirement is either I write a whitepaper or I contribute significantly to a proposal. While I've very much enjoyed working on proposals by myself while I was at the defense company (or by myself with one other guy to provide the technical idea, and I write it all up), I will DIE if I have to somehow contribute to a proposal on like doing an IT transformation at a government agency, that is cowritten by 30 other people and involves about 300 meetings. So whitepaper it is. Especially cause for me it's easy to write an article; I write a million shitposts a day after all
Revolutionary new medical study encourages patients to stop being poor
This is a factor that almost all of these studies ignore and it's kinda infuriating.
"We studied a bunch of non-affluent people (who only signed up for the study because they needed the money) and we were shocked to find they're in poorer health than affluent people (who eat far smaller amounts of processed foods because they can afford to and also can afford to do things like regular checkups with their doctor because they fucking have a doctor and not whatever person happens to be on duty at the one urgent care that their HMO actually lets them go to on the other side of the city)"
I mean yes. Being poor is bad for your health.
But also something you can have some controls on for larger studies like this.
Also trying to figure out which food is killing us faster is an ongoing thing. The study even mentions multiple studies across non-US health systems.
I mean the Spanish study that basically is causing issues in the EU isn't about poor people eating jamon every day.
This feels reductive. And this is coming from me who hates most dietary studies. At least this one covers though people and years to cross section a lot of the populace.
Revolutionary new medical study encourages patients to stop being poor
This is a factor that almost all of these studies ignore and it's kinda infuriating.
"We studied a bunch of non-affluent people (who only signed up for the study because they needed the money) and we were shocked to find they're in poorer health than affluent people (who eat far smaller amounts of processed foods because they can afford to and also can afford to do things like regular checkups with their doctor because they fucking have a doctor and not whatever person happens to be on duty at the one urgent care that their HMO actually lets them go to on the other side of the city)"
If I’m reading correctly, that’s essentially what the study is arguing without using the actual words.
“The studies advise eating unprocessed meats and fresh fruits and vegetables,”
Cool. We try. But it costs an hour of my time and $15+ to cook a serving of chicken and veggies, and a can of beans and sausage takes 10 minutes and costs $4. Or, you know, I’m at work and only have time for a deli sandwich.
Posts
We just went to lunch
And never came back
Yeah, so like, deli meats and sausages and salami
We can’t all be out there having our chef make us veal chops nightly
Revolutionary new medical study encourages patients to stop being poor
Look I already changed classes once and it totally fucked up my meta
Made for the auger
fact: I have no favorite tattoos
This is such an A+ response I have no way to top it
that means nothingggggggggg
will have to read their actual paper about it to see what it actually does
by the way I am working on a whitepaper I'm trying to publish tentatively called "I want to cure cancer with LLMs: how to scope your AI use case". Hardest thing will be how to not make it very insulting in tone, but I did recruit a coauthor (but she's also mean, so)
edit ok @Neco the above was followed up closely by a random Shelly Duvall appearance this is wild
https://www.tiktok.com/t/ZTL4YxP5b/
i have patented all the meths
The job I currently do is too stupid to replace with AI. It’s kind of why I took it.
Professional Penny-Arcade Poster?
The main site links to a couple of the papers from a few years ago but that field's beyond my ken.
I support government acquisitions, but because nobody in the chain wants to trust anyone else in the chain, people review each piece of it. And none of those reviews is automated, and they are making it even less automated because of some circular self dealing in the current process..
All molecules of life has real "you can online with your friends, or your enemies," energy
or your enemies
online!
Maybe on mobile but not on desktop.
oh man this is what I want for Christmas
One year when I was in scouts, we had an instructor from idk Norway or Finland or such and his dad was a fisherman of some sort.
And we were talking about survival stuff and he said "the ocean is so cold, they tell you if you fall into the water don't try to swim"
With some further explaining, we found out you are supposed to just float to keep from losing extra heat waving you limbs through the water. Not just let yourself drown.
Anyways I started gliobasting
AI does specific things. Any given biotech problem involves some specific steps. You gotta ask some actual fucking working scientists what those steps are, and then figure out if AI will in any way make them faster or better (or enable you to circumvent the steps or paradigm shift but let's be real if you have to ask the scientists what the process is you do not have the knowledge for a paradigm shifting ai concept) and you have to be ready for the answer to be no. Like. I just read some garbage Time article someone posted at work being like, the age of biotech AI is now! Biotech uses CRISPR, so AI will solve that! And it's like. What the fuck do you think CRISPR is? (it is a biochemical process where you put some proteins and some nucleotides in an animal. You design what nucleotides in order to decide what specific part of the genome you will edit. You are editing the genome because you are looking to create a mouse with a specific genetic condition. You are making a mouse with a specific genetic condition for any number of reasons but you have chosen which genetic condition in accordance with your specific experimental goals (e.g., design a fat mouse so you can test diabetes drugs)) Ok now that you've taken the time to learn in the just vaguest terms what CRISPR is, why do you think AI would help do this in any way? like what part of the process would AI do...? Actually, maybe you better go ask someone who understands AI what AI can do, cause otherwise you won't be able to answer that either.
What is your actual goal? Your goal can't just be 'introduce AI to my org'. Like what are you actually fucking trying to do? You want to cure cancer? no you do not want to cure cancer, that is not a real thing. Maybe what you mean is you are looking to find a small molecule that inhibits a protein involved in a common cancer pathway? That sounds more like a real thing. How do scientists go about that? Well, one thing that they might do is virtually throw a bunch of small molecules against 3D structures of proteins, and see if any of them nestle in nicely. But, if you don't have a 3D structure of a protein you care about, you can use alphafold to generate a plausible one, most of the time. Hey that's AI, good job! And hey honestly, if you want to identify some promising proteins in the just absolutely horribly cluttered bazillion cancer papers out there, could to worse than to ask chatGPT to give you a list. Hey you used AI again, good job! Ok cool so now you did this docking simulation and used AI twice, do you have a drug? Lol no now you have to test it in cells and in mice and in people. How are we gonna use AI to do that? ummmmmm I mean. You could use a pipetting robot...
--anyway
i have feelings
about this
hopefully I will write this all out and get it published in our stupid consulting magazine, we will see. I have a whole like categories of what AI does and where in the process and how you can break your process down and see if any of the categories make sense. maybe I'll make some ~process graphics~ and such
I read some papers recently about using neurostimulation to disrupt the feedback loop between the cancer cells and the normal neurons, because electrical activity makes the glioblastoma grow faster. It does seem to increase survival time although it is not curative. Maybe see if there's any clinical trials you dad's eligible for? I don't totally know how to do that (getting patients for clinical trials is a valid use of AI in a drug development process, also, like if you have a big pharmacy or health system dataset)
who writes papers anyway, eggheads?
This is a factor that almost all of these studies ignore and it's kinda infuriating.
"We studied a bunch of non-affluent people (who only signed up for the study because they needed the money) and we were shocked to find they're in poorer health than affluent people (who eat far smaller amounts of processed foods because they can afford to and also can afford to do things like regular checkups with their doctor because they fucking have a doctor and not whatever person happens to be on duty at the one urgent care that their HMO actually lets them go to on the other side of the city)"
I've been told that in order to get promoted a requirement is either I write a whitepaper or I contribute significantly to a proposal. While I've very much enjoyed working on proposals by myself while I was at the defense company (or by myself with one other guy to provide the technical idea, and I write it all up), I will DIE if I have to somehow contribute to a proposal on like doing an IT transformation at a government agency, that is cowritten by 30 other people and involves about 300 meetings. So whitepaper it is. Especially cause for me it's easy to write an article; I write a million shitposts a day after all
I mean yes. Being poor is bad for your health.
But also something you can have some controls on for larger studies like this.
Also trying to figure out which food is killing us faster is an ongoing thing. The study even mentions multiple studies across non-US health systems.
I mean the Spanish study that basically is causing issues in the EU isn't about poor people eating jamon every day.
This feels reductive. And this is coming from me who hates most dietary studies. At least this one covers though people and years to cross section a lot of the populace.
If I’m reading correctly, that’s essentially what the study is arguing without using the actual words.
“The studies advise eating unprocessed meats and fresh fruits and vegetables,”
Cool. We try. But it costs an hour of my time and $15+ to cook a serving of chicken and veggies, and a can of beans and sausage takes 10 minutes and costs $4. Or, you know, I’m at work and only have time for a deli sandwich.
I didn’t get into genetic engineering to NOT turn every turtle on earth into a neutron bomb!
Is dracula really a vampire???????
*cut to him fangily eating people*
spoopers
I have played so MUCH Hades 2, there is so much in there already (at least as much as there was in Hades 1 at 1.0)