God why do LLM comics look so fucking uncanny
Because perspectives get misaligned and there’s no continuation.
There’s a whole desk with books that just disappears. The poster in the background can’t decide what it wants to say. Nor can it decide on its size. The shelves in front of it can’t decide what they contain or where they are positioned And the wall it’s on can’t decide how it’s angled relative to the room.
The “NO AI” dude is taking a giant step forward with his left foot behind a garbage bin for some reason.
In the third panel the doctor is missing a ring-finger while his thumbs have grown 3-4cm
Just to name a few.
The stupid texts all over the place that are made to sound deep yet are somehow devoid of any useful information make me irrationally angry. You can really tell these models have been optimized to generate powerpoint presentations for creatively bankrupt low-level managers
LLMs have discovered deepities. It’s so nauseating.
My favourite is the googly-ass binocular microscope. Maybe two scientists are supposed to use it at the same time like 50s teens sharing a milkshake.
But even when they get everything right they just trigger alarm bells in my head, like the way their faces and eyes look freaks me out.
The “uncanny” part I think isn’t because it’s done badly, it’s because it’s done well - but not in any of the ways you’d get from a human. By the time a human artist has learned to get that good at shading, colouring, line work, human figures, posing etc, they’ll have also learned about composition and perspective and good storytelling and when to not pointlessly overload the background with weird shit.
Yeah that scans, its like when it’s taken as a whole it looks off. The people and poses look like they’re copied from this platonic stamp somewhere rather than drawn.
Dumb fucks really think LLMs are the kind of AI involved in medical and scientific research.
Yeah. At this point LLMs are what people are referring to 95% of the time when they say AI. If one comes up with the cure for cancer, it’s because we already have it and it picked it up in the last data scrape.
And someone wins the coveted Nobel Prompt Prize.
And in 4.5% of the remaining 5% they reference generative AI
Then 0.45% of the remain they’re referring to NPCs in games.
0.045% of the time they’re talking about Weird Al.
0.0045% of the time they’re talking about a ouija board
Removed by mod
LLMs is a subset of generative AI, but an image generator as an example is not a LLM
Removed by mod
Transformers and diffusion models are not the same, even if they use related math
Edit: By the magic of advanced mathematics, every Turing complete structure can be rewritten into any other Turing complete structure with no loss of function or change in behavior
So the logic “remove one” will remove all of modern computing because the AI companies are not giving up their models. If they have to start pretending they aren’t neural network anymore, then they will.
And if you DON’T rewrite, then you’re back to them actually being different again
Removed by mod
And that’s why science research should go private. Can’t risk slop machines stealing the breakthrough.
ChatGPT, cure cancer for me.
No, that didn’t work, try again
That one gave me a rash, try again.
That last one did not work, also the rash is starting to look a little gross.
Sorry for the delay, I had to go to the ER to get my arm amputated from the rash, can you try again, also look for a cure for amputated arm.
Hyper scalers have so little for their models and compute to do they are just brute force hacking things or mining crypto.
Not quite, but some of them do run on the same principle.
But they are?
Uses LM to define new peptides to bind targets.
Removed by mod
So, the manuscript I referenced which no one read uses a language model. you just input text protein sequence and it will generate binding peptides sequences, using only the language model. A version of an LLM.
Removed by mod
They absolutely are, but most people on this platform would prefer to hear their comfortable lie instead.
No, chatbots are not used in medical or scientific research.
Yes they are? Literally have to listen to a friend bitch about them every single day. She’s a biochemist. They do literally use specialized AI models, general LLMs and and a number of other transformer based technologies.
And while the specialized ones are useful the stupid chat bot ones are fucking over processes all over the place.
I’m not such why you’d utter such a lie so confidently. This reeks of fear, ignorance or a mix of both.
No, chatbots are not used in medical or scientific research.
So you’re a scientist?
I use LLMs every day, they search deep into supplemental data and manuscripts that would take months to sift through. Can’t find specific information, reagents or data using title and abstract searches.
LLMs can be useful tools, but this entire thread is people with bullshit jobs scared of being replaced by an algorithm because they know they have a bullshit job.
I think it’s more that the way the chatbots work is the same. They assign a bunch of numbers to tokens, and the numbers determine how interrelated different tokens are. You put in a bunch of tokens and then it spits out what it thinks the next one should be based on all the interrelatedness numbers and some math things like transformers, and the math things have knobs on them to determine how strong they are. It doesn’t matter if the tokens are words in a sentence or peptides in a protein. The underlying technology is the same.
This is an exciting use for a tool that I mostly despise for probably all the same reasons you do.
Chat bot is the interface. Its like saying, “NO, life saving drugs arent made with keyboards and mice!”
Why are AI simps so fucking weird?
It sure is weird how AI boosters have made a hobby out of deepthroating strawmen & pure speculation.
The latest trend on my LinkedIn is people comparing AI water usage favourably to almond farming.
You know, the thing that famously uses a scandalous amount of water and which also has environmental groups campaigning against it…
They’re constantly being told they’re correct about everything, what do you expect?
generative ai? I dont think so.
Most research is done using PREDICTIVE ai not generative ai.
And no one’s against machine learning, but fruit romance videos do not cancel out surveillance, inequality, and destruction of nature.
fruit romance videos do not cancel out surveillance, inequality, and destruction of nature
What?? Of course they do, I saw it in an AI-generated video!
Removed by mod
One of the infuriating things I’ve seen in the industry in a bunch of cases. Folks working on valuable things getting pulled away to work on chatbot. Not even to try to make the chatbot game changing, just to make it more popular.
Also, there was pressure in the industry to “be green”; to be efficient. That is all so far out the window now too…
Google knows AI will not cure diseases. They have services for pharma for two years now and no one is finding it useful, despite the cost.
That being said, I wouldn’t have complained if they’d left that AI genome project running, just in case
I hate how LLMs have co-opted the term AI. Google was using AI for this just a different and I’d argue better technique. It’s super frustrating to see other AI approaches be totally discarded just for everyone to put all their chips on LLMs being the ‘right’ approach
They’ve said something to that effect a long time ago. Back when they were first getting into the self diving car game they said that more people die of car accidents than cancer and it was more important to them. They’re number driven, if the numbers say they’ll make more money somewhere else, that’s where they’ll go.
Making money is not mutually exclusive. Google is already bloated with cash (it pays dividends). The are able to invest in both self driving cars and cancer cures at the same time.
So where’s that miraculous cure, then?
Dude, didn’t you read that comic? The anti-AI psycho destroyed it. The vibe scientist probably tried to type in the same prompt again but this time the answer was in Mandarin.
You’re so right, HIV infected plasma wasn’t the cure for cancer. Maybe we should step back and look at this from a new perspective. I will generate a new vaccine for you and find a new batch of test subjects.
Google discontinued development of alphafold, broke up the team, let many go and transferred the rest to general LLM development except for some that were moved to their vertically integrated own pharmaceutical research (to cash in on the whole value chain and lock everyone else out of their tools).
Mind you, that is the tool that got people a Nobel Prize.
https://thenextweb.com/news/deepmind-alphafold-team-dismantled-gemini-anthropic
Luckily there’s also OpenFold as an open-source alternative for protein folding (maybe also nucleic acids? I dunno)
There are numerous models available nowadays, even alphafold still is but Google took itself out of the game by letting it die due to lack of further development, focusing on its “be evil” company philosophy.
I’m convinced “Don’t be evil” was a canary.
It certainly was, and it didn’t take long for the canaries to start dying, which immediately alerted everyone that there was a possible canary-murderer on the loose within the organization, which they solved by following the canary-murderer’s instructions to pursue profit at any cost and using a tiny sliver of the budget for regularly buying more canaries.
letting it die due to lack of further development
You are incorrect. see above.
deleted by creator
Doesn’t work for mid-large proteins. AlphaFold now models using >50 post translational modifications and full size proteins.
Now we see the tech being used for the whimsical and capricious, but also the potentially destructive, like that hacking incident.
Google discontinued development of alphafold
No they did not, it’s on version 3 and they just added >50 PTMs to modelling this month. Another version is being developed and sold to pharma on subscription on a large scale batch modelling platform.
broke up the team
The team continued on to other projects, including the commercial version mentioned above. There was no point in continuing when the development was just adding more features.
the tool that got people a Nobel Prize
2 people from Deepmind, Jumper and Hassabis, one person at U Wash , Baker, is still developing AlphaFold with Google and is a academic lab.
You put a lot of faith on one vague article.
You partially correct my statement but weirdly enough largely confirm it, just with a PR twist. “There was no point in continuing when the development was just adding more features.” But it is not, AlphaFold 3.0 is still far from perfect. We need more than just a few more added features. But the next big step will be either some closed thing, developed by a largely changed team, sold for a fortune as subscription model to pharma multis or be done by entirely different groups that have nothing to do with Google or Deepmind.
There is a reason why Hassabis is not working on it anymore and of course he will say it was all as it should be, given how he was moved upwards.
Giving LLMs credit for all machine learning is problematic
This is their go to motte-and-Bailey argument fallacy. Say something outlandish, get called out on it, retreat behind something more reasonable and how no one notices before sallying forth with unreasonable claims again.
Yeah machine learning for medical research has been around for a lot longer than LLMs. LLMs do not do medical research, they are just really expensive and fancy “next word” auto-correct generators.
I’d argue they made scentific research harder due to the many fake papers you can find online
that’s been a problem since 2005 at least (first time i did serious medical resesarch)
That’s been a problem since start of humanity
Isn’t it worse now? AI slop can be produced waaay faster
no? when you’re doing real research you limit your sources to “reliable” ones. being able to write fast has never been the obstacle to publishing bad research.
I remember way back when with the Folding at Home project. That’s a type of machine learning distributed. People happily contributed their computer time. Hell, there were contests around how much compute time a particular group could give.
I feel like anything involving electricity will be called AI fairly soon.
It makes steak taste better.
I cant believe how many people give strong opinions on generative ai when they cant even distinguish it from the much more general term machine learning
Wait, these things still can’t draw hands?
Or proper perspective
I’ve given up. I know this because I didn’t even notice until I saw this and went back to check.
That one frame… It was like the AI is going “humans have 10 fingers” then drew three on one hand, and crammed the rest onto the other.
The “AI” we are obsessing over and building entire power plants around… Yeah, that stuff sucks, a lot. When we get to real AI, aka AGI, then it’ll quickly turn into sky net and we’re all fucked.
They won’t stop until that happens, and nobody has the means to stop them… Anyone who has the means to stop them, won’t.
Start digging your Doomsday bunkers now folks.
This shit can’t even keep a consistent background over 4 comic panels
The scientist’s desk design changes in all four panels as well.
One thing I have to ask is why the AI insists on having slogans and lists on every object in its “illustrations.” You can see this everywhere. Hell, I went to a close friend’s funeral and they had AI generated memorial stickers. Even there, there are signs filling the negative space saying shit like “dedication, service, a safer community.” Is this an AI proclivity or do normal people just like that crap?
It’s essentially by design. To keep AI from destabilizing, the way AI is tuned is to essentially ‘go the same safe way’ on similar prompts. LLM models are essentially a giant field of weights for chains of tokens. Prompts are turned into tokens and the closest result is the output. The more room you give it to vary, the more different styles and variation it can express with a propmpt, but it raises the chance of destabilizing the ouput, producing things like additional fingers or complete garbage. It’s tuned to walk the same way to produce something we perceive as “passable” but it comes at the cost of taking away variation, which is why AI art/writing will consistenly converge towards the same narrow style set.
(This explanation is not super accurate. I just tried to convey the jist of it)
Still with the missing fingers too
It’s hilarious what it does with a room full of table and chairs like a classroom. Cannot figure the legs, what a chair looks like, how people sit at desks, none of it.
I am pretty sure the “AI” systems making medicine aren’t the same as generative AI.
They are not. They are purpose built systems that leverage specialized knowledge.
The problem is you have people asking LLMs to find a cure for cancer and have no way of knowing if the output is any good. Then those people start emailing everyone saying they cured cancer.
Random dipshits asking ChatGPT, at best, would just give you a story about a cure for cancer. It wouldn’t actually be able to cure cancer even under the best circumstances since it was made to be a chatbot, not a medical research tool. It’s like going to Burger King and asking for a pizza. 🤣
I don’t have experience with cancer but my email has seen an increase in people claiming to have used an LLM to design a perpetual motion machine
“Chat GPT said if I put two repelling magnets onto a skateboard, the force of the magnets repelling each other will move the skateboard forward!”
Not knowledge… Data. X, Y is one Cartesian point composed of two parameters but they can process billions of parameters. It’s not knowledge. Knowledge assumes that a thing has intelligence and realizes what the data means. Not just regurgitating the data but actually extracting the meaning.
I was referring to the people that build the MLMs not the MLM themselves.
There are actually some, but you wouldn’t recognize it as generative AI.
Math is a structured language for conveying relationships, a mechanically verifiable system, and a system for verifying those statements.
This doesn’t work for all math, but a lot of it.So you can train an LLM on math the symbolic language and then mechanically check the output for validity and immediately feed the lesson back in.
Mechanically shitting out math and then throwing away the junk has been a thing for a while without llms. What they bring is a system for picking the next step for the derivation that’s less likely to end up being junk.
So if you can model it with math you can use that model to mimic the pattern of symbols that’s often valid.It’s moderately more efficient at searching infinity for value than other systems.
So it doesn’t give you a medicine, it gives you the mathematical description of a molecule, with an abstract proof that under certain models it might have utility and be physically possible, which another program can feed into simulations to see if it’s something worth looking at.
It’s not just
What an absolute triumph of human intellect! By asking for a pharmacological cure for cancer, you have single-handedly pushed the boundaries of computational oncology, and it is a breathtaking honor to co-create history with you today!
It’s viewed as having potential by people who aren’t delirious largely because of the “mechanically verifiable” part.
Their plagiarism machine can’t even consistently render that guy’s desk from panel to panel. Somebody’s going to trust it to make drugs?
Ha and the posters and bottles. Maybe AI could be useful for generating spot the difference books.
But a bit boring when you can just circle the entire panel and get it right, no?
Have you considered that the scientist maybe works for a good organization that can afford him 4 desks that look different?
Ah, yes. Perhaps they are mounted on some manner of conveyor belt.
A rotating floor plate would probably be more efficient
Use ai to generate a build plan and BOM
I think you left an “O” off of your bill of materials, there.
There we go, some healthy heavy craziness to keep the crazy going, a crazy train if you will
Not sure why the organization would allow medicine-smashing man onto the premises though.
he’s the founder
I’m curious what purpose that clearly custom microscope has
That’s the least realistic part of this
The kind of crappy art generation is vastly different than the types of AI used in drug discovery.
In fact drug discovery is a pretty good problem for AI to help solving.
They aren’t using chatbots to do it. It’s almost a completely different technology.
Exactly. Things in advanced analtyics or ML or whatever are totally smeared over with a broad brush, they give everything the landlord special.
It more or less is a different technology in the same way a arquebus and an M16 are different technologies, same principles different execution.
Yes, but the problem is that the techbros pushing crappy AI and text generation are doing so while disingenuously implying (or often outright stating) that their product is eventually going to revolutionize the world by also magically curing cancer or whatever the hell else if we provide them just one more datacenter, bro. Except that it categorically cannot.
Specialized machine learning tools are specialized tools for the specific clearly defined use cases for which they are designed and intended (edit). The commercial LLM and diffusion image generators, meanwhile, serve no beneficial purpose whatsoever other than burning electricity, spreading lies, and making people dumber.
I mean this is where we have to disagree, I think there are loads of benefits from the technologies. But capitalism is the wrong motivator… “How can we make money out of this?” don’t make me puke. It should be free for all, to cure cancer, to decipher ancient languages, to model the climate etc.
How the fuck is an ass kissing chatbot going to help you model the climate?
Not everything is a fucking chat bot.
“AI”, the way everybody uses it today, means “ass-kissing chatbot”. That’s what Jensen means, that’s what Satya means, that’s what the investors think it means, that’s what the people reading scripts on the news mean. Oh, excuse me, I meant to say “super intelligence”.
We have neural nets, machine learning, adaptive systems, automated translation, context completion, all kinds of terms for the stuff that actually does useful stuff.
That’s not what the trillion dollars is getting spent on
There’s no point having a discussions with people like me that are intractable when it comes to the technology and are acutely aware of the USA capitalist ruling class and the atrocities they commit and enable, with people that are unable and unwilling to understand what a model is, and are approaching computer science like a luddite.
The USA capitalists are having you lick their boots when you think that “AI” = ChatGPT. You’re not ‘chatting’ with anything. You, me, and many others, are constantly falling for their marketing shit all the time. Don’t bite!
If you can differentiate between a stack of PS3s that made a ‘super computer’ to try to predict what the weather will be tomorrow and the USA building a datacenter, poisoning your water, denying you electricity, stealing your capital, surveiling you, and influencing your decisions, then we can definitely sit together and use our brains to dismantle the infrastructure of oppression. Your enemy is not ‘conversation simulators’, its capitalism.
I know a Doctor at an Nvidia backed biotech company that is working with generative AI to find new drug formulations to test. Allegedly they have some promising ones that are in testing. The thing is even if they are successful and the models generate a viable drug this shit won’t be free. They’ve sunk so many billions into processing power they’ll justify charging $100k a dose to pay back their Nvidia loan.
or they can use it and declare bankruptcy then make it open source how are they going to monetize it when he made it intentionally open source the elites epstein class can sue him but he declares bankruptcy thus saving humanity trillions in cost from the epstein class.
and unicorns are real.
Breakthroughs like this WILL come from AI. But…
- They’re not going to come from LLMs or image generating AIs.
- They’re not going to come from the numbnuts generating comics in his bedroom - they’ll come from real scientists.
As demonstrated by AlphaFold, which is neither an LLM nor generative AI. Coincidentally, I’ve never heard any actual complaints about what AlphaFold does for research, and the complaints seem to focus on LLMs and generative AI instead. Go figure.
It’s going to come from scientists who know how to use ML to supplement their work. Also I don’t think we are finding cure for cancer as a drug. Only cure I think may work 100% we master fully controllable nanotech that can also read DNA non invasively.
Never going to happen.
Cancer isn’t one disease, it’s a series of different diseases for each organ.
There are basically infinite cancers if you think about it since mutations and damages are totally casual
(To be noted: i know very little about cancer so i may be wrong and basing this comment on a probably misconception we laymen have)
Cancers are one thing: cells dividing beyond the limit. This is a DNA problem, made worse by existing mutations. Turns out LLMs are getting really good really fast at biology. So why then is it “never going to happen”? What’s the blocker?
I think I may not have written my comment in the most coherent way possible.
My first sentence was a reply to “nanotechnology”. It’s one of these sci-fi technologies that simply can’t happen. You won’t suddenly have atomic-scale supercomputers floating around being able to somehow unravel DNA AND have the capacity to edit it.
It’s just not how matter behaves at those scales.
My second sentence still stands. I do agree that some kind of large scale data attack to the problem will help.
But the solutions won’t be magical “Fantastic Voyage” nanorobots shooting Ant Man lasers at cells.
It will likely be a whole panoply of attacks on cancer cells, cis-platin with special spike proteins that need all kinds of external inputs to work (you’re not downloading them and 3D printing that), or some other cancers may respond to pumped-up antibodies.
etc
I do hope that our new data processing heavy-lifters will also help with anti-aging and life extension.
ya feel me?
I agree. I was responding to the cancer thing rather than nanobots.
The thing is, when you look small enough, cancer and aging share a cause: DNA issues. You don’t need nanobots for this, we already have methods of editing DNA today. All we need to know is what to change, something easier said than done.
Honestly, there is potential to ‘cure cancer with (help from) “AI”.’
1: You take a DNA test (simple cheek swab or whatever) to identify your particular genetic risk factors might be. Or a small biopsy is done to sample some of the tumor itself and figure out exactly what kind of cancer it is. Just a needle would work, assuming it’s somewhere a needle could reach. Only need a microscopic sample of it. Or if it has already metastasized into the bloodstream, a simple blood sample would do.
2: This genetic information is fed into a purpose-built AI (not an LLM) that’s trained to identify the particular strain of cancer and match it with a custom-designed drug cocktail that will target that very specific cancer mutation perfectly.
3: The drug combination is synthesized, and you now have the perfect medication to combat exactly the strain of cancer you have.
The problem with ‘curing cancer’ is that there’s millions of different kinds of cancer, millions of different mutations and genetic risk factors that can combine in different ways to produce different cancers that respond to different treatments. We already have ‘cured’ a few specific types of cancer – a few specific strains that we already have found solutions to. But researching and developing a solution like that takes monumental effort and a ton of money and time. It’s just not feasible to do for all the different kinds of cancer out there. If an AI could do that part quickly and easily, that might give us access to ‘cures’ for nearly every type of cancer, instead of just a few out of millions.
2: This genetic information is fed into a purpose-built AI (not an LLM) that’s trained to identify the particular strain of cancer and match it with a custom-designed drug cocktail that will target that very specific cancer mutation perfectly.
The “trained to identify the particular strain of cancer and match it with a custom-designed drug cocktail” is the magical thinking part of your idea - basically the “???” step of the three steps to Profit of the underpant gnomes.
Exactly, so I’m not saying AI will not be useful, AI will be useful in reducing cancer related deaths, early detection and management, but for a true “cure” it would have to be some other technology.
Well, as I explained in a post above, AI (so, not LLMs or other generative shit) is pretty limited in what it can research and as of right now can’t do anything for which there isn’t a single concrete research direction with some kind of algorithmic validator for the solutions or where results can’t be reached in a couple of days.
Further, there is no evidence that it ever WILL - maybe it will, maybe it won’t: it’s at best a possibility with an unquantified probability of happening.
Ai techniques (more on the ml side or just the math behind it) are starting to be applied to image and genome analysis to determine the boundaries from different types of cells based on gene expression. This is useful for identifying cancer where the gene expression is similar in many genes but the amount/ratio of expression might be different, and for identifying the location and size is the tumor
Problem is that some of the research is coming out of Israel which doesnt exactly have the best human rights of ethics track record, and i have no doubt the data for testing was gathered through human experimentation
Of course they’ll come from ai, but that’s more a linguistic thing. We have a tendency to call new computational technologies ai. But yeah, neural networks are being used in medical research and will likely bear fruit.
