Aleesa, hi! I’m a pharmacy student and I’m actually planning to do my thesis on “AI in drug discovery”. It would mean the world to me if I could somehow talk to you about your paper or if I could simply read it. Take care :)
Really nice to hear AI being talked about in a balanced and fair way. So often, the capabilities of AI are over stated or misrepresented. It is so, so, so important that the limitations of AI are talked about freely, (e.g. the fact that the AI doesn't know what a Tchaikovsky is) to avoid fear mongering.
Yes! That's right. These statistical models and such (e.g., transformers) are really limited. They can only recognize statistical patterns (really, correlations) between words and make predictions based on that. It's like giving a baby, since birth, the entirety of Wikipedia (just the text). Sure, the baby can recognize patterns between the words, but without actually seeing them and how they act and interact with other things, the baby will not really understand what they are. The words are not literally that: they're representations of the much more abstract "meaning" the speaker (or writer) is trying to convey, in which some real-world experience with the actual objects (concrete or abstract) it talks about is a prerequisite. It's the infamous "grounding problem" that I'm sure will take a bit of AI research to resolve. Don't be fooled by how good OpenAI's GPT-3 performs; it really doesn't "understand" much. Try giving it "the square root of thirty-two times five is " and ask it to predict what the next word is; it gets it hilariously wrong, because it's not doing actually "understanding" the sentence and doing the underlying computations but rather analyzing the text and making a statistical prediction on the next word based on a huge corpus of text it has previously analyzed. It doesn't understand the "meaning" of the sentence, just the associations and patterns of already composed text.
I work for a company called BenevolentAI, in January our AI platform correctly identified an existing drug that could be used to treat covid-19 symptoms... Lily are now in human trials for it.
@@harvest5218 Petition to request that our future robot overloards remove our heads and replace them with something more useful, like say, nintendo switches
Weed. I suffer terribly with migraines and have three different types of specific migraine medication. Since I started vaping fresh weed, I've only had 1-4 a year, compared to 1-2 debilitating migraines a month. If I do have a headache or the start of a migraine weed kills the pain.
I'm really impressed by the Complexly logo - have seen it lots now but never took the time to tell you how satisfying it is to see the components coalesce into that nice big "C" Kudos to the designer!
Developing a potential drug is the fraction of the cost of the clinical trials, and this part AI can not accelerate. You should have mentioned that at the end with the numbers. Great video!
To be fair, the AI only needs to know VX nerve gas(can be absorbed through the skin) and it has all it needs. Like how are we going to fight an AI that throws out VX gas like it is candy, everyone wears full body hazmat suits? Fighting against quadcopters armed with guns while in an uncomfortable rubber suit; yeah, no thank you.
@@maxplaysgamez-sharesgaming1756 You know any AGI (artificial general intelligence) in a faraday cage will be able to survive a solar storm. EMPs and solar storms aren't nearly as effective as many people think they would be.
+@@josephburchanowski4636 I Knew That, I Just Don't Think It Was Worth Mentioning It, So AGI Could Take Note On That One. As A Matter Of Fact, They Would Be Thanking You In The Near Future, My Dear Friend. **Wink** **Wink** ;)
Watch this right after CGP Grey's How Machines Learn and you realize that spaghetti throwing is still the main method. But instead of looking for drugs in the spaghetti we're looking for bots that can look for the drugs for us.
Excellent video... Most other AI medicine videos talk about boring crap like Radiology and scanning for parkinsons disease, but you've actually talked about drug discovery which is what I'm really after... I have schizoaffective disorder and I really need the cure.
I was really hoping this video would talk about Folding@home and the COVID moonshot project. In fact, I think this video would really have benefitted from a segment talking about how all this relates to the drug development process for COVID-19.
The DreamLab app (by Vodafone) helps get some of this research done. It uses your phones processing power while it's charging to solve the equations that support these projects. They support the Imperial College, Garvan Institute, and AIRC and tackle issues like cancer and COVID. All you have to do is download it and the setup is super easy.
My brother-in-law was recently diagnosed with pancreatic cancer. I wish that the research was finished before he was diagnosed (really wish he never had it). I do hope the research returns positive results.
@04:20 "researchers at the US company, Berg, grew cancerous and healthy cells from over a thousand donors in petri dishes" *That's a lot of either some very large petri dishes, or very small donors, for them all to be in petri dishes!*
Currently AI is very narrow as to what it can do, but as it widens it we’ll probably see even faster progress Development in drugs. No doubt this will push for more funding in AI. The more use cases it gets, the more Funding This technology will get. Even now governments are starting to get how important this technology is now, realize its a national security issue. I can only imagine what's it's going to be in 10 years.
One of the risks with using any such mechanized filter is false negatives: The AI could incorrectly rule out lots of potentially-useful drug candidates, with little sense as to why. Neural networks are essentially pattern matchers, based upon *inductive, not deductive* “reasoning.” It’s almost massively-amplified intuition, so there’s no real “why” involved. As you pointed out, humans can filter out false positives - cases where the AI suggested a drug that won’t work, but it’s much harder to revive drug candidates the AI ruled out.
Great video! It's very hopeful, just hope the AI doesn't decide to poison us :/ . . . But! . . Musicians weep as Tchaikovsky turns in his grave by that pronunciation!
The most important question is whether the US healthcare system will make these new AI-discovered drugs hella expensive or not. Insulin can be made for cheap by using modified yeasts, but the pharmaceutical giants can still make them cost hundreds of dollars per vial because...profits.
Won’t be long before Big Brother’s computer analyzes your total profile and then provides you with a customized, AI-concocted, drug regimen, for a monthly fee to Big Pharma.
Oh, boy... we can't even teach cars how to drive by themselves without fail. Hell, people can't even drive so, sometimes we bite off more than we can chew.
Wait, what do you mean target Coenzyme Q10? Is there something bad about it? Because everything I know about it makes it extremely important for things like heart, brain, and immune system health.
They've been saying that for 30 years. Robots are *expensive*, and it's still far cheaper to have people involved as robots can't make judgement calls, or deviate from the programmed movements.
Can you discuss Icilin aka AG-3-5? I'm sure everyone whose ever discussed biochemistry has a video about Capsaicin and other TRP channel molecules, but nobody seems to care about extreme cold. I guess its trendy to Hot foods, so cold gets overlooked. Anyways Icilin is stronger than menthol, but menthol can be organically derived. Thanks for making great videos!
To be clear the vast majority of the funding for AI research into drugs is by publicly funded research institutions, NOT PHARMASEUTICAL COMPANIES. The pharmaceutical companies then receive the research for next to nothing, patent the process to make the drug, set an artificially high price, ultimately leading to billions in dollars of profit that does not go back into research. So that billion dollar price tag quoted is on the taxpayer not the company.
This sounds like the medical equivalent of a self checkout. It's just there to cut costs for the business, so they can turn more of a profit. They wouldn't drop prices of the medications just because they are able to reduce manufacturing costs. They would just see it as increasing profit margins.
lol the matrix And the terminator event? didn't the matrix AI want their own order and law, while the terminators only wanted to, y'know, terminate stuff? i think both events would cancel each other out with them fighting each other.
People: Big pharma are money hungry! Pharmaceutical scientists: We've just spent 15 years getting one single drug to market and had to deem hundreds of others a fail... I don't think people appreciate the work that goes into pharma science or the humans having to do the grunt work (although, give us coffee and we'll be good!) (I'm kinda glad I left my undergrad program, the math subjects were doing my head in!)
I love that you made a video about this! I wrote my research paper on this topic last year, and it’s such a fascinating topic!
Existenceisillusion it was never published but thank you for your interest!
Feel free to publish it, or upload it to arxiv for peer review
Aleesa, hi! I’m a pharmacy student and I’m actually planning to do my thesis on “AI in drug discovery”. It would mean the world to me if I could somehow talk to you about your paper or if I could simply read it. Take care :)
@@arsakellariadisas a pharmacy student from Switzerland, i am writing a paper on the subject as well. Would love to have your insights
I love how this video subtly taught us the process of developing a new drug (from scratch), after drawing us in with robot scientists. ;D
You're right! They tricked me into learning something. Brilliant
An adorable kitten cloud is no blunder.
It's a direct hit to my heart!
We are getting closer to creating Curie from Fallout 4.
That's not a bad thing.
@@sneakerbabeful what could go wrong? Lol
Guilty Spark or Geth
I loved her so much
Ms. Handy Curie? or sexy synth Curie?
Brings a whole new meaning to designer drugs.
Really nice to hear AI being talked about in a balanced and fair way. So often, the capabilities of AI are over stated or misrepresented. It is so, so, so important that the limitations of AI are talked about freely, (e.g. the fact that the AI doesn't know what a Tchaikovsky is) to avoid fear mongering.
Yes! That's right. These statistical models and such (e.g., transformers) are really limited. They can only recognize statistical patterns (really, correlations) between words and make predictions based on that. It's like giving a baby, since birth, the entirety of Wikipedia (just the text). Sure, the baby can recognize patterns between the words, but without actually seeing them and how they act and interact with other things, the baby will not really understand what they are. The words are not literally that: they're representations of the much more abstract "meaning" the speaker (or writer) is trying to convey, in which some real-world experience with the actual objects (concrete or abstract) it talks about is a prerequisite. It's the infamous "grounding problem" that I'm sure will take a bit of AI research to resolve.
Don't be fooled by how good OpenAI's GPT-3 performs; it really doesn't "understand" much. Try giving it "the square root of thirty-two times five is " and ask it to predict what the next word is; it gets it hilariously wrong, because it's not doing actually "understanding" the sentence and doing the underlying computations but rather analyzing the text and making a statistical prediction on the next word based on a huge corpus of text it has previously analyzed. It doesn't understand the "meaning" of the sentence, just the associations and patterns of already composed text.
Boston Dynamics' dog robot making and giving new drugs would be cute. Maybe dress them in a lab coat or just have a white paint scheme.
"Lab" coat?
"they never have to duck out for coffee"
As I take a sip of my coffee..
Coffee doesn't exist.
J.J. Shank let me check...
No coffee so sad
This shall soon happen to the judicial system
Same
I work for a company called BenevolentAI, in January our AI platform correctly identified an existing drug that could be used to treat covid-19 symptoms... Lily are now in human trials for it.
Aalong as they put some creatine in the mix ill be ok
Hey robot scientists, cure headaches.
and all mental illnesses pls. I'm tired of having anxiety 😂
I received a response: Remove head.
...
I mean... That works, I guess?
@@camramaster If can figure out how to separate me from my head I'd at least give it a listen.
@@harvest5218 Petition to request that our future robot overloards remove our heads and replace them with something more useful, like say, nintendo switches
Weed. I suffer terribly with migraines and have three different types of specific migraine medication.
Since I started vaping fresh weed, I've only had 1-4 a year, compared to 1-2 debilitating migraines a month.
If I do have a headache or the start of a migraine weed kills the pain.
I'm really impressed by the Complexly logo - have seen it lots now but never took the time to tell you how satisfying it is to see the components coalesce into that nice big "C"
Kudos to the designer!
Bah, I'm a lab tech. The robot overlords can pry the pipette from my cold, dead hands!
That can be arranged, Ketchup!
Boston dynamics 'Spot' has an optional arm for that...
You mean to say: "Sudo, pry the pipette from my cold dead hands."
Excited to see the safe psychotropic drugs that could be found because of this. Major therapeutic potential
"The future looks kinda like a robot handing us a miracle pill."
Eat the ice cream.
2050: We're Teaching Robots and AI to Design New humans
You just stole the thread lol!!!😭😭😭😭😭😭 They're gonna be like can you make LeBron James... but white?
It wouldn't be human if it was designed because flaws are part of what makes something human.
@@BigMobe are you on your meds? Because machines/robots design and make things that are flawed...like you know computer parts car parts etc?
This sounds more like an advanced search engine than a drug designer
This 100% reads more like a "scary" PSA, this is still good to know.
Take this medicine that is totally not slow-acting poison.
@@BigMobe "What do you mean 'mercury isn't the elixir for immortality'?!"
I'm using this for a presentation in my machine learning class, thanks for making my research easier!
Developing a potential drug is the fraction of the cost of the clinical trials, and this part AI can not accelerate. You should have mentioned that at the end with the numbers.
Great video!
Imagine one day you can just make your very own medicine specifically to your body using AI,.... This is a wide topic
Thinking of microscopic robots to fix your health I always think of Drix
The Optimist: that’s incredible soon no more decease!
Me: teaching AI what chemical can kill us?
To be fair, the AI only needs to know VX nerve gas(can be absorbed through the skin) and it has all it needs. Like how are we going to fight an AI that throws out VX gas like it is candy, everyone wears full body hazmat suits? Fighting against quadcopters armed with guns while in an uncomfortable rubber suit; yeah, no thank you.
+@@josephburchanowski4636 Universe: **Smiles In Solar Storm**
@@maxplaysgamez-sharesgaming1756 You know any AGI (artificial general intelligence) in a faraday cage will be able to survive a solar storm. EMPs and solar storms aren't nearly as effective as many people think they would be.
+@@josephburchanowski4636 I Knew That, I Just Don't Think It Was Worth Mentioning It, So AGI Could Take Note On That One. As A Matter Of Fact, They Would Be Thanking You In The Near Future, My Dear Friend. **Wink** **Wink** ;)
Joseph Burchanowski let’s keep the chemical between us! The smooches people lol
At worst we must convince them that we are good batteries lol
Scientists: We can use AI to create new drugs way faster now
Me: Cool but I think the goal is to be healthy enough so we don’t require drugs
Good luck with that.
Nice description of AI's importance in Drug discovery. Please continue.
A traffic cone is the exact definition of what that pin doesn't look like lmao
Watch this right after CGP Grey's How Machines Learn and you realize that spaghetti throwing is still the main method. But instead of looking for drugs in the spaghetti we're looking for bots that can look for the drugs for us.
Phinedroids and Ferbots are the best robots man has ever built
So basically Breaking Bad but with robots?
Michael Aranda is by far my favorite host
Oh gods, we're equipping them for the robot apocalypse...
spaghetti throwing is still a strong aspect that is commonly used
Whoa I love this guys hair, he looks great
Just take a step back and realise how important this is for humans and how futuristic robot scientists are
very well made. Thank you for this excellent video
"The Voltron of modern pharmaceutical science". 😵
I wonder who forms the head?
Excellent video... Most other AI medicine videos talk about boring crap like Radiology and scanning for parkinsons disease, but you've actually talked about drug discovery which is what I'm really after... I have schizoaffective disorder and I really need the cure.
Yes, finally! AI medicine is awesome!
Yeah what if they create a deadly one
@@phillipatteberry9819 Ok sry
Leaf hmmm I bet no humans have ever made any dangerous medicine hmmm oh wait maybe giving everyone morphine wasn’t a good idea
This is crazy. And btw you presentator you got so much better and I like the earrings bro
The next step will be identifying why a drug that looks like it should work fails.
A robot handing you a miracle pill... I approve of this potential future.
Yeah, it's more like a third that aren't cats in my experience. I just searched my google photos for cats, and I got two chicks.
I was really hoping this video would talk about Folding@home and the COVID moonshot project. In fact, I think this video would really have benefitted from a segment talking about how all this relates to the drug development process for COVID-19.
Ironically, robots will never need the drugs they make for us.
The DreamLab app (by Vodafone) helps get some of this research done. It uses your phones processing power while it's charging to solve the equations that support these projects. They support the Imperial College, Garvan Institute, and AIRC and tackle issues like cancer and COVID. All you have to do is download it and the setup is super easy.
My brother-in-law was recently diagnosed with pancreatic cancer. I wish that the research was finished before he was diagnosed (really wish he never had it). I do hope the research returns positive results.
"Drugs er baed m'kay?" - Mr. Garrison (SOUTH PARK)
humans: "help us make drugs"
the robots and AI: *relabel 'cyanide' with 'new drugs':"here have some of this"
@04:20 "researchers at the US company, Berg, grew cancerous and healthy cells from over a thousand donors in petri dishes"
*That's a lot of either some very large petri dishes, or very small donors, for them all to be in petri dishes!*
Nothing could possibly go wrong with this
Voltron drugbot got my like
Cant wait for the singularity :D
Hopefully it wont be too long ^^
(and hopefully its not a dystopian one )
Fantastic overview of the R&D side of pharma.
This would be awesome if the end goal were to create cures, but terrifying if the goal is to keep people just well enough to function.
yea...
Excellent presentation. Thank you.
Currently AI is very narrow as to what it can do, but as it widens it we’ll probably see even faster progress Development in drugs. No doubt this will push for more funding in AI. The more use cases it gets, the more Funding This technology will get. Even now governments are starting to get how important this technology is now, realize its a national security issue. I can only imagine what's it's going to be in 10 years.
I hope this will be successful!
One of the risks with using any such mechanized filter is false negatives: The AI could incorrectly rule out lots of potentially-useful drug candidates, with little sense as to why.
Neural networks are essentially pattern matchers, based upon *inductive, not deductive* “reasoning.” It’s almost massively-amplified intuition, so there’s no real “why” involved.
As you pointed out, humans can filter out false positives - cases where the AI suggested a drug that won’t work, but it’s much harder to revive drug candidates the AI ruled out.
Great video! It's very hopeful, just hope the AI doesn't decide to poison us :/
.
.
.
But!
.
.
Musicians weep as Tchaikovsky turns in his grave by that pronunciation!
It's brazy that I have to even mention sickle cell but yeah, we are looking forward to quantum AI solving these simple ailments
The most important question is whether the US healthcare system will make these new AI-discovered drugs hella expensive or not. Insulin can be made for cheap by using modified yeasts, but the pharmaceutical giants can still make them cost hundreds of dollars per vial because...profits.
Hero - one who risks their life to save another.
Won’t be long before Big Brother’s computer analyzes your total profile and then provides you with a customized, AI-concocted, drug regimen, for a monthly fee to Big Pharma.
well put Boomer John
As a person with many Things, I'll take it.
Oh, boy... we can't even teach cars how to drive by themselves without fail. Hell, people can't even drive so, sometimes we bite off more than we can chew.
We can't teach humans to drive by themselves without fail either.
Reading papers is a much easier thing to do than driving safely.
@@himanbam I see your point
They will design a drug that will make us feel like we need to give AI complete freedom to become our overlords :O
Wait, what do you mean target Coenzyme Q10? Is there something bad about it? Because everything I know about it makes it extremely important for things like heart, brain, and immune system health.
Your hair reminds me of my hair. I don't know if this is good or bad because I haven't cut my hair since covid got big in the US
Bro I feel that I was fr thinking about commenting something like this.
I haven't cut my hair during this millennium.
An ai analyzes our comments as we comment on ai 😳🔥
Wow, that sponsor segue almost knocked me out of my chair!
He sounds like he could be the voice of Kaz Brecker in the Six of crows audiobook
Inshort:
Science is Hard,
Humans are bad,
AI-Robos are better.
Speed it up, robots. My brain needs some stuff.
I’m really scared we’re teaching robots to read
Don't thrown spaghetti, it's so much better in your tummy
What will be truly amazing is when an AI runs its own drug company give it 20 years or so.
Robot drugs?
Okay I'm down
What can human brain do that computer brains can't?
"No more spaghetti" 😔
I have a genuine fear of A. I. taking over the world, and this is unnerving
That wont happen
@@Chris-gd4vc what if it does 😶😶
@@meetaverma8372 it is very unlikely... idk
So the development period is the justification for the obscene price-tags we see on prescription drugs.
Can you do one on why Cats are fascinated by water?
_GLaDOS has entered the chat_
Thanks all you nerds out there!
I'm from the future, this AI sht is getting too real
I'm from further future and you are right!
It's less like 99 cats and a cloud that resembles a cat and more like 99 cats and an airplane we have no idea why it's there.
In before every conceivable task gets automated by 2060
How are you everywhere so fast
I do NOT like your funny words, magic man.
And in 2060 old people would still be complaining about how the newer generations are lazy because of it
@John Smith If IA became so much of a problem for jobs, they'll be EMP terrorists and group actively going against it
They've been saying that for 30 years. Robots are *expensive*, and it's still far cheaper to have people involved as robots can't make judgement calls, or deviate from the programmed movements.
Can you discuss Icilin aka AG-3-5? I'm sure everyone whose ever discussed biochemistry has a video about Capsaicin and other TRP channel molecules, but nobody seems to care about extreme cold. I guess its trendy to Hot foods, so cold gets overlooked. Anyways Icilin is stronger than menthol, but menthol can be organically derived. Thanks for making great videos!
Design Ai that thinks like dmt elves
is it possible for drugbots to get overdosed from pharmaceutical data? 🤣
So, they're like the opposite of Terminators then? We'll soon be attacked by the Perpetuators... :P
Come with me if you want to live
To be clear the vast majority of the funding for AI research into drugs is by publicly funded research institutions, NOT PHARMASEUTICAL COMPANIES. The pharmaceutical companies then receive the research for next to nothing, patent the process to make the drug, set an artificially high price, ultimately leading to billions in dollars of profit that does not go back into research. So that billion dollar price tag quoted is on the taxpayer not the company.
Sounds efficient
I'd like to see them cure (not just treat) allergies.
the AI will certainly have no idea what a "chuckovsky" is because that's not how you say tchaikovsky. Chai-kov-ski.
This sounds like the medical equivalent of a self checkout. It's just there to cut costs for the business, so they can turn more of a profit. They wouldn't drop prices of the medications just because they are able to reduce manufacturing costs. They would just see it as increasing profit margins.
And companies will mostly use this to remix drugs, hold on to patents, and keep them from going generic. Drug pricing will stay inflated.
Just wait until AI *self aware* then the Matrix and Terminator event will happen 😱😱
lol the matrix And the terminator event? didn't the matrix AI want their own order and law, while the terminators only wanted to, y'know, terminate stuff? i think both events would cancel each other out with them fighting each other.
2:12 -- Drugbot. Sounds like a character that *should've* been on Futurama, but wasn't, unfortunately. ;-)
People: Big pharma are money hungry!
Pharmaceutical scientists: We've just spent 15 years getting one single drug to market and had to deem hundreds of others a fail...
I don't think people appreciate the work that goes into pharma science or the humans having to do the grunt work (although, give us coffee and we'll be good!)
(I'm kinda glad I left my undergrad program, the math subjects were doing my head in!)
Somone is going to ask an AI for supercrack- oh, they did it already.
Great.
Someone say supercrack?!
Let me guess: That someone was Michael Reeves and the AI was programmed by him?
they need to make a drug to treat MDMA comedown.