The Thinking Game now has +300M views on YouTube. From DeepMind's documentary, the most replayed scene shows a meeting where someone tells CEO Demis Hassabis that AlphaFold can predict all known protein sequences, around 1 to 2 billion, in about a month.
Aren't there a lot of dangerous proteins. Like really, really dangerous. Apocalyptically dangerous. Are they going to review to make sure they don't give the formula for the zombie virus to the public? Are they even capable of reviewing in such a way? Do any of these people have even a high school biology level understanding of what they are doing?
If i had billions, id be setting up think tanks with specific goals in mind. Energy forums to discover or create new things, ocean think tanks, to clean and use the ocean for power. I'd have such a blast handling major problems every single day. And id totally have a whole group of people inventing batman-esk awesome crap for me too.
Well, this is just WILDLY optimistic due to the limitations of our models. There are huge gaps in our understanding of how linear protein sequences wind up in their functional 3D forms.
This was a great documentary… although they had funding from Google .. they were at a stage of development where the CEO didn’t have to worry about making quarterly numbers .. the goal was growth and exposure so this was the right move. I got the same vibe as when they sold the patent for insulin for $1 to ensure accessibility
( it’s the right move anytime .. but AI advancements will be highly proprietary imo .. just like modern insulin is now )
Sad truth right there. They did this because they needed publicity, and buzz, and funding. Once they become profitable, the altruism dies and they become monofocused on profit at the expense of everything (and everyone) else.
This scene doesn't make much sense out of context. Step 1: identify every protein structure. Step 2: ? Step 3: problem solved! So now you have 200 million items, some may not even exist and there is no guarantee all of them can exist because this was done with AI.
And while I'm sure there's more to it, a scene that shows some CEO going "just do it" and that's supposed to be the big thing that NO ONE ELSE has ever thought of, that leads to success, is wildly misleading.
Because proteins obey such rigid rules for their construction, it's not like LLMs that spit out language or visual images. By every account, it's *incredibly* accurate. It won't replace real experimentation, but version 3 will be able to predict protein interactions with non-protein molecules, and it is hard to understate how important that information is.
This is google, they are not short on cash. But the researchers at universities that developed the Protein and sequence databases that these people used are underfunded as hell, and they won't be getting any money from Google. This problem was solved very quickly, simply because google put their wallet behind it. It's a great achievement, imagine what ample funding of science could achieve. I am a scientist and my plumber earns more than me, and I fight for funding daily.
I remember that, I was part of it for a moment, running my pc at night. I think I ended up stopping because my gaming pc at the time was too loud and had too many lights, so it interfered with sleep.
This got me back into gaming, weirdly enough. First you have to know that the PS3 got a Folding@Home client (and was one of the most energy efficient systems you could have for running a client thanks to the CELL processor) and second you have to know what the PS3 was the cheapest Blu-ray player for like a year after its launch. Those things together convinced me to drop cash on it without even worrying if I'd find games to play. Of course, I found plenty of games too.
As an extra bonus it's also what got me interest in indie games. The PS3 did such a great job promoting indie games and their devs and of course the top #1 most deeply impactful gaming experience of my life came from one such game originally released on PS3 (Journey).
IDK if my participation in Folding@Home meant anything to anyone but I am grateful for a handful of deeply meaningful experiences I've had thanks to various PlayStation consoles over the years and would've missed out on a bunch if F@H hadn't helped pull me back in.
Spent 10 years of my life doing this and feel it accomplished nothing and cost me a ton. Watched my Mom and little sister die from cancer during that time.
The protein folding thing helped create the machine learning models that turned into this major one. Without the effort of millions it would've taken much longer if it happened at all.
Despite what some politicians seem to want, science is moving forward. Hell, recently I learned about a vaccine for cancer that's being worked on. They tailor make a vaccine based on your own cancer cells, then your body can wipe it out on its own.
Proteins are a sequence of amino acids, which we can discover. However, in order to fully understand how they work and interact with the world, you also need to know the structure those sequences of amino acids fold into, which can be complex. This can be done with computer simulations, but it's been computationally intensive to do that in the past. Alphafold used AI to create predictions of how proteins fold, and it did it really fast and fairly accurately. They were talking about how to...
deploy it to the world, and they were thinking of a service where people submit protein sequences, and then they'd spit out a prediction. However, in the meeting, while talking about the speed of their AI, one person, to illustrate how fast it is, said, as an example, it could predict all known proteins (about 2.2 billion) in about a month, and the CEO was like, "wait, why don't we just do that?"
Kinda hard to explain, cause there are a lot of applications for knowing all possible proteins, especially in relation to medical science to do with the curing of diseases, drug production, and cancer. Knowing how your drug will interact with proteins, before proceeding with any kind of test trial, does wonders for determining potential side effects. Its not a replacement for test trials, but rather a tool to improve them. Alphafold 3 is now doing the same with RNA, and DNA.
I wish we were using a different word; the one the plagiarism machine is actually, which is 'LLM'. It's not AI, because it doesn't learn, it just predicts. Like imagine if instead of reacting to sentences you read by coming up with your own sentence, you looked at the words in the sentence and then predicted the most likely word that will follow the question mark by running a probability equation against every question anyone has ever asked lol. No wonder they use so much damn energy.
Current AI is being used by a individuals with various disabilities, to improve their overall quality of life, like giving mute people a voice that isn't limited to specific words and phrases, but actually being able to convey even emotions, through text to speech. I suppose another third kind of application, could be translation services for people in foreign countries.
My dad had a stroke a couple years back and has trouble articulating his thoughts, so it does help him communicate via email or text for important things like banking and medical stuff.
If i may though suggest a 3rd thing. Using it to study and scan deep space telemetry and information. There's so much to go through and a machine can do it very accurately, saving researchers years of work.
A friend of mine who was a... ahh I forget the proper name... A nuclear engineer for cancer treatments, saw this and spent like 2 weeks pouring over all their research, and wished that this stuff was available 20 years ago when he started, it would have saved so many lives.
Warlizardfromtheforum
Aren't there a lot of dangerous proteins. Like really, really dangerous. Apocalyptically dangerous. Are they going to review to make sure they don't give the formula for the zombie virus to the public? Are they even capable of reviewing in such a way? Do any of these people have even a high school biology level understanding of what they are doing?
Agatsu74
There's still some good in this world, Mr. Frodo.
Bizarkly
I liked the von Neumann poster -- *where can I get one*?
GreenMnM
If i had billions, id be setting up think tanks with specific goals in mind. Energy forums to discover or create new things, ocean think tanks, to clean and use the ocean for power. I'd have such a blast handling major problems every single day. And id totally have a whole group of people inventing batman-esk awesome crap for me too.
Tenzil
Well, this is just WILDLY optimistic due to the limitations of our models. There are huge gaps in our understanding of how linear protein sequences wind up in their functional 3D forms.
hwatL4bloopy
Perfect, I remember the early versions of this being shown at my college with a interactive screen
Khanamana324
This is the type of shit our taxes should be funding, not giving trillion dollar handouts to oligarchs and their enforcers.
Deonan
APES TOGETHER STRONG.
Osiricus
Breaking: White House blocks access to AlphaFold
iteachcalculus
I knew there was a reason I got out of x-ray crystallography twenty years ago to get into teaching. So many hours spent purifying proteins…
Crispo
Veritasium also did a great video about it https://youtu.be/P_fHJIYENdI?si=7-T04YChOd8hkcGh
TheGreatMani
Demis Hassabis is my favorite scientist since Carl Sagan
taulogy
https://www.youtube.com/watch?v=d95J8yzvjbQ
gtollie
This was a great documentary… although they had funding from Google .. they were at a stage of development where the CEO didn’t have to worry about making quarterly numbers .. the goal was growth and exposure so this was the right move.
I got the same vibe as when they sold the patent for insulin for $1 to ensure accessibility
( it’s the right move anytime .. but AI advancements will be highly proprietary imo .. just like modern insulin is now )
CallMeCourierSix
Sad truth right there. They did this because they needed publicity, and buzz, and funding. Once they become profitable, the altruism dies and they become monofocused on profit at the expense of everything (and everyone) else.
echonite
See, this is what AI is good for. B
Prototypo
This scene doesn't make much sense out of context. Step 1: identify every protein structure. Step 2: ? Step 3: problem solved! So now you have 200 million items, some may not even exist and there is no guarantee all of them can exist because this was done with AI.
Nathanyel
And while I'm sure there's more to it, a scene that shows some CEO going "just do it" and that's supposed to be the big thing that NO ONE ELSE has ever thought of, that leads to success, is wildly misleading.
ThingsThatDontJustifyGenocide
This is medical-AI, not OpenAI corporate LLM generative bullshit AI.
Don't cede "AI" to the generative shitboxes.
This is good, and the accuracy of it in above 90%. It is good enough for "make the thing and test it to see what happens"-work.
inboxmefoxes
Because proteins obey such rigid rules for their construction, it's not like LLMs that spit out language or visual images. By every account, it's *incredibly* accurate. It won't replace real experimentation, but version 3 will be able to predict protein interactions with non-protein molecules, and it is hard to understate how important that information is.
TinyOctopus
This is what billionaires should be doing. Because they could have made billions from selling access, but instead, they gave it to the world.
albaboss
Plus you know teams like this are scrounging for funding. If they were easily fully funded, this could have been a reality much sooner
DrDownunder
This is google, they are not short on cash. But the researchers at universities that developed the Protein and sequence databases that these people used are underfunded as hell, and they won't be getting any money from Google. This problem was solved very quickly, simply because google put their wallet behind it. It's a great achievement, imagine what ample funding of science could achieve. I am a scientist and my plumber earns more than me, and I fight for funding daily.
PaperinoVB
Before that, there was folding@home: https://it.wikipedia.org/wiki/Folding@home
Jamerperson
I run the folding@home imgur team. Come join us with team 238444
Atomic2
I remember that, I was part of it for a moment, running my pc at night. I think I ended up stopping because my gaming pc at the time was too loud and had too many lights, so it interfered with sleep.
JackieTreehornProductions
Wasn’t that on the PS3 too and you could let it run while in sleep mode?
LaughAtMemeInjectReason
Folding@home is still active. There's no need for the past tense here. :)
PaperinoVB
"It was already there" doesn't imply that it no longer exists, right? Or there is some other quirk in the english language that Im not aware of?
CouldntCakeLess
For extra credit try: fitted sheets
PaperinoVB
Hey, man... we're talking about science, not science fiction.
anononanonanona
This got me back into gaming, weirdly enough. First you have to know that the PS3 got a Folding@Home client (and was one of the most energy efficient systems you could have for running a client thanks to the CELL processor) and second you have to know what the PS3 was the cheapest Blu-ray player for like a year after its launch. Those things together convinced me to drop cash on it without even worrying if I'd find games to play. Of course, I found plenty of games too.
anononanonanona
As an extra bonus it's also what got me interest in indie games. The PS3 did such a great job promoting indie games and their devs and of course the top #1 most deeply impactful gaming experience of my life came from one such game originally released on PS3 (Journey).
anononanonanona
IDK if my participation in Folding@Home meant anything to anyone but I am grateful for a handful of deeply meaningful experiences I've had thanks to various PlayStation consoles over the years and would've missed out on a bunch if F@H hadn't helped pull me back in.
shammael
A few places where I worked donated plenty of computer time to that.
Unknowingly, of course.
PaperinoVB
Bro ;-)
phalanxausage
I had half the computers in my department running this when I worked an odd shift at IBM.
NoneOfYoBiseniss
Spent 10 years of my life doing this and feel it accomplished nothing and cost me a ton. Watched my Mom and little sister die from cancer during that time.
bamblebimble
The protein folding thing helped create the machine learning models that turned into this major one. Without the effort of millions it would've taken much longer if it happened at all.
Despite what some politicians seem to want, science is moving forward. Hell, recently I learned about a vaccine for cancer that's being worked on. They tailor make a vaccine based on your own cancer cells, then your body can wipe it out on its own.
roflipop
I’m so sorry.
NoneOfYoBiseniss
Thx, me too, still not over it.
MattDude
Please tell me what the fuck I'm looking at. Sincerely, A confused Imgurian.
marsilies
Proteins are a sequence of amino acids, which we can discover. However, in order to fully understand how they work and interact with the world, you also need to know the structure those sequences of amino acids fold into, which can be complex. This can be done with computer simulations, but it's been computationally intensive to do that in the past. Alphafold used AI to create predictions of how proteins fold, and it did it really fast and fairly accurately. They were talking about how to...
marsilies
deploy it to the world, and they were thinking of a service where people submit protein sequences, and then they'd spit out a prediction. However, in the meeting, while talking about the speed of their AI, one person, to illustrate how fast it is, said, as an example, it could predict all known proteins (about 2.2 billion) in about a month, and the CEO was like, "wait, why don't we just do that?"
dailysuicidalthoughts
Kinda hard to explain, cause there are a lot of applications for knowing all possible proteins, especially in relation to medical science to do with the curing of diseases, drug production, and cancer. Knowing how your drug will interact with proteins, before proceeding with any kind of test trial, does wonders for determining potential side effects. Its not a replacement for test trials, but rather a tool to improve them. Alphafold 3 is now doing the same with RNA, and DNA.
dailysuicidalthoughts
It is, IMHO one of only two currently acceptable applications of our current level of "AI".
inboxmefoxes
I wish we were using a different word; the one the plagiarism machine is actually, which is 'LLM'. It's not AI, because it doesn't learn, it just predicts. Like imagine if instead of reacting to sentences you read by coming up with your own sentence, you looked at the words in the sentence and then predicted the most likely word that will follow the question mark by running a probability equation against every question anyone has ever asked lol. No wonder they use so much damn energy.
Omicron416
Okay, I'll ask: What's the other one?
dailysuicidalthoughts
Current AI is being used by a individuals with various disabilities, to improve their overall quality of life, like giving mute people a voice that isn't limited to specific words and phrases, but actually being able to convey even emotions, through text to speech.
I suppose another third kind of application, could be translation services for people in foreign countries.
RyvaTheRenamon
My dad had a stroke a couple years back and has trouble articulating his thoughts, so it does help him communicate via email or text for important things like banking and medical stuff.
If i may though suggest a 3rd thing. Using it to study and scan deep space telemetry and information. There's so much to go through and a machine can do it very accurately, saving researchers years of work.
dailysuicidalthoughts
A friend of mine who was a... ahh I forget the proper name... A nuclear engineer for cancer treatments, saw this and spent like 2 weeks pouring over all their research, and wished that this stuff was available 20 years ago when he started, it would have saved so many lives.
Junktrunkjunkie
Some people are doing the right thing. Some people are making the world a better place. At a time like this that’s worth remembering.