I never said it was unique though, you’re projecting. You’re the one using whataboutery.
AI and ChatGPT
- Thread starter JB08
- Start date
-
The World Cup 2026 Forum is now open.
Pickle made a distinction earlier in the thread between the unnecessary nature of AI and the apparent necessity of eating/travelling (ignoring the fact we don’t need to eat specific foods that are particularly bad for the climate, and we don’t need to travel to every part of the planet for tourism or always drive a gas guzzling car instead of walking or cycling). That was what I was referencing in my first post.
But I do see this particular focus on AI energy usage a lot, seemingly ignorant of the rest of the energy consumption we do as a species, at far larger scales. Not saying you specifically, as I know your initial post was a joke, just a general point.
Demis Hassabis co founder of DeepMind, behind all of Google/Gemini efforts and Noble prize winner for AlphaFold responded to the dog story
He seems to think we will cure all diseases by 2030, Alpha Fold helped alongside GPT.
We are in the golden age of mRNA now and custom drug development for specific genomes
I'm of the mind AI will help make some breakthroughs in energy with time, we're just in that pain point moment where we're building all these data centers, using a lot of energy and the initial capital and energy costs people deem bad, but they'll bring about huge changes in society.
You're reading things into my response that weren't there. I wasn't talking about flying for foreign holidays when I was talking about essential travel, I was talking about journeys to see loved ones, to work etc. Day to day stuff. You're the one that waded in with a bunch of assumptions and deployed whataboutery.
Like it or not, much of what AI is used for is non essential at the moment and whether people eat meat, travel or whatever other strawman you want to build won't change that
I'd assume making up a number here but 99% of usage of AI will be non essential.
With GenAI people will be making custom music, writing a prompt into a netflix style AI platform to generate new movies from scratch at some point in the next 10-15 years. Realistically people won't be solving climate change and the energy crisis, or diseases because those are a one off thing, (meaning we might have 5000 iterations of those but eventually they'll be solved problems) peoples need for entertainment won't be. People will always want fresh content.
Hoping we just get a breakthrough in fusion or something of the like because we can't scale up AI without fixing the energy problem at some point (the compute and energy demands of AI and people will only grow)
In principle, designing a custom vaccine is not particularly difficult or time intensive. The main constraint is the experimental work that follows. Wet lab validation, manufacturing, and clinical testing require substantial time and money. Many candidates also fail at one step or another.
AI may help improve candidate selection (e.g. by predicting which antigens are more likely to be immunogenic) but these predictions still need experimental verification. In practice, the wet work remains the main bottleneck.
Lots of current trials reflect this. For example:
https://www.economist.com/leaders/2026/01/22/britains-good-idea-for-custom-genetic-medicines
https://www.england.nhs.uk/2024/05/...ccess-trials-of-personalised-cancer-vaccines/
These programmes focus on building infrastructure to scale personalised therapies, including identifying patients, recruitment, regulatory pathways. Rather than solving design problems.
Generating candidate vaccine designs computationally is already relatively fast; a single researcher can produce dozens in a short time (similar principle as with Covid vaccine design, completed in 24-48hrs). What takes years and costs millions is manufacturing, clinical trials, and demonstrating safety and efficacy. AI may improve candidate selection/prioritsation and even trial design, but it does not remove the requirement for experimental validation.
Last edited:
Designer viruses are a concern, targeting genetics etc... How do you protect against that?
That's one of my 'bad actor' fears, a virus hyper contagious with a high fatality rate.
I assume bio terrorism will rise quickly if we get to that stage.
I don't know how can can prevent it at all, it's not like you can stop a virus that attacks the heart or brain unless you mobilize local labs (clinics) have the ability to produce vaccines overnight and people go for a vaccine, but even then people can just ideate quickly and create another strand.
If a person goes through an LLM (private openAI etc) could it be flagged from the person using it? or perhaps the likes of AlphaFold will have flags to intelligence authorities for people using proteins associated with harm?
I'm guessing the hardest part is people using open source tools to achieve the same aims since you can't control it. Or could we have a 'vaccine' that presents known threats in the immune system similar to how a cancer vaccine would happen?
Assuming we could also have a digital twin of our DNA in a healthy state that we have a comparison against for real-time monitoring of if we get infected not sure the cost viability of that though.
I think it's very hard to prevent but I'm no geneticist, biologist, virologist or medical practitioner.
https://www.nature.com/articles/s41746-025-01874-x
https://pmc.ncbi.nlm.nih.gov/articl...,limitations and advancing precision medicine.
Last edited:
Yeah it's a worry and I suspect it's being looked at by many agencies globally.... Like you I don't know where we start with it but rapid local response will definitely be a part of it.
A digital twin, Jesus Christ, I can't wait until Silicon Valley crashes and burns. A record of your DNA has nothing to do with digital twins, we've been able to sequence DNA for donkey's years. There was a documentary about it in 1993 called Jurassic something or other I seem to remember.
I get triggered by the words digital twin , it's the biggest steaming pile of bollocks of a concept imaginable. It's a simulation guys. We've been doing those since before we had computers. Calling it a digital twin just means you want to separate a dozy business or government from their money.
Most likely. In fact, it is close to impossible to get info on making a biological weapon from public Western LLMs. These models have several levels of defense, including one or more external classifier that checks if the LLM generated text is 'harmful', in fact, for Claude that adds 5% in inference.
Not sure how this will be handled with open-source LLMs though, you can very easily remove their defenses.
That would be almost certain. It exists, analogue, ish, already. All you'd be doing is putting health data into virtual solution space so that actual data, like blood tests, would update it and it could run all manner of tests if you explicate it in pure mathematics.
I mean, I think this inevitable with the only variable being what kind of architecture is used when it becomes "standard". As someone else said, though, it's not really a twin, it's just a bunch of information which is modeled based on your various tests (blood and whatever). If you have enough data, you can mathematically test, in VSS, for all kinds of things, though it may take a while to get there, so that one blood test, could, in THEORY, be used to check a thousand diagnostic boxes. The diagnostic utility of something like this, whatever form it takes, would be immense.
Hearing that phrase a lot recently. It really is classic tech bro snake oil bollox. A new name for something we’ve already been doing for ages.That use case might be just a simulation (not looked) but digital twins as a concept has been around for a long time it's certainly not a new conceptual AI fad. I did some research work on it as a grad a long time ago now around city planning.
Used to be called mirror worlds (which is even worse as a name), I think the digital twin naming came about due to city planning which makes sense given the existing terminology of twinned cities. I recall the US even back in the mid 2000s was looking to go a step further and build out a live world model for warfare, feck knows what they have now.
The whole point is to separate it conceptually from standard data fed simulations, if it's actually an integrated simulation with live data then digital twin seems fair. It's basically just
live modelling though yes.
From a geography/GIS perspective, digital twins is quite old. It refers to mapping more than just roads. Things like buildings, both inside and out being mapped and you have a digital twin version of a city. Esri has been promoting it for quite a while.
Last edited:
Indeed, it is now such a vague term as to be meaningless. Is it a 3D map, or a throughput model for a factory, or a model of a vehicle used for CFD, or a performance prediction for a battery? Or do all those things in fact have their own names already and the general collective term for them would be "models" or "simulations" unless you want to pull someone's pants down in exchange for them?
Nvidia DLSS5 announcement..... What the feck is this shit.
I genuinely thought it was a joke but it isn't. Digital Foundry must have been paid for their analysis of it because they think it looks amazing. It just looks like AI slop.
I genuinely thought it was a joke but it isn't. Digital Foundry must have been paid for their analysis of it because they think it looks amazing. It just looks like AI slop.
Can’t believe I ever doubted the posters in here who argued it wouldn’t just mostly be used for total meaningless, harmful shite.
A more balanced look at "ai cures dog cancer"
si=FDXTVBaRiOe7FkrN
si=FDXTVBaRiOe7FkrN
Nvidia DLSS5 announcement..... What the feck is this shit.
I genuinely thought it was a joke but it isn't. Digital Foundry must have been paid for their analysis of it because they think it looks amazing. It just looks like AI slop.
Woah, this actually looks pretty good minus some of the faces, but it's the worst its gonna be and I'm guessing you can have it on/off in each case to fit your graphical preference / art style
Nvidia DLSS5 announcement..... What the feck is this shit.
I genuinely thought it was a joke but it isn't. Digital Foundry must have been paid for their analysis of it because they think it looks amazing. It just looks like AI slop.
I don't think you understand what AI slop means. That most certainly isn' AI slop.
Assasssins Creed Shadows has some good lighting improvements. It can save a lot of resources, calculations and time trying to do this in engine and devs can adjust the amounts.
Sounds a bit counterproductive.
Completely changing the original design of a character's features so it looks like any other generic AI created face? AI slop doesn't just mean creating daft videos of babies and dogs at the beach.
If it's modifiable then absolutely, but it's a worry that they thought this looked great for their showcase. The Starfield one in particular just looks lazy to me.
Call it overreacting but it's a damning indictment of what Nvidia thinks of art direction if it can just practically erase months of work at the flip of a switch.
Which bits are completely changing the original design because nearly every face was an obvious enrichment in my view. All but one bit of that video looked a vast improvement.
AI slop doesn't just mean generative content you don't like, it's low quality oand/or mass produced content farming.
Game characters are by design pretty generic and base. If they can use a model to enrich features how is that a bad thing? If it turns out the tech standardises in game characters to a more limited range of features then that might be problematic but I don't see that evidenced.
I think you've come in with a bias reviewing that video.
If you think there's no major differences between the original designs and how this new DLSS changes them, or that game characters by design are pretty generic already, whilst claiming I'm the one with a bias then I really don't see the point in discussing it with you. We clearly have different opinions on art design, or at least the importance of it.
Last edited:
Robot in California knocked over utensils and dishes, started to dance uncontrollably and had to be restrained by workers.
Last edited:
The cops and judges involved in allowing this woman to spend 6 months in prison with nothing beyond an AI facial match should all be fired.
Viewing AI as a tool whoever was the responsible party should 100% face the same kind of fine/legal action as any other case IMO.
This is one of the biggest questions for our supposed near future, where does liability lie when a person isn't the origin of a decision or an action? AI companies aren't going to take responsibility, and the people using it also won't, and if they're made to, then they may just not use AI. It will definitely just end up with the consumer being fecked, as usual.
Somewhere along the line there will be a human in the loop and when things go pear shaped they will likely be culpable. Anyone for full automation without human sign off is a loon, let that be decades in the future if ever....
Oh no, whatever would we do if we didn't have AI?
The liability lies with the firm adopting AI, the same as utilising any vendor for critical processes. Yes there's a looser arrangement over contracted functionality but it's still a tool adoption. Even if you think of it as an employee the employer has the ultimate responsibility to train, proceduralise and in general ensure proper performance.
Nvidia DLSS5 announcement..... What the feck is this shit.
I genuinely thought it was a joke but it isn't. Digital Foundry must have been paid for their analysis of it because they think it looks amazing. It just looks like AI slop.
Does this not just look way better? Surely machine learning and computer game graphics is a no-brainer.
Improving visual fidelity I don't mind at all.
I used to play the likes of Skyrim and look for mod packs for better texture, lighting etc. This is just an out of the box way to get better visuals.
I saw an example where Harrison Ford went through DLSS 5 on Reddit
I can understand peoples grievances here but I'm pretty sure it's still experimental and NVIDIA will look at feedback, and it's also optional so it can always be turned off.
I think why not free up developers to make the gameplay as good as possible and things like DLSS5 just be something they can plug in and just work.
In some cases I can see why people are annoyed, also people are annoyed at the price of GPUs (I want an upgrade but I cant afford it) so I think there's a big reason for people who game to hate AI too.
Improving visual fidelity I don't mind at all.
I used to play the likes of Skyrim and look for mod packs for better texture, lighting etc. This is just an out of the box way to get better visuals.
I saw an example where Harrison Ford went through DLSS 5 on Reddit
I can understand peoples grievances here but I'm pretty sure it's still experimental and NVIDIA will look at feedback, and it's also optional so it can always be turned off.
I think why not free up developers to make the gameplay as good as possible and things like DLSS5 just be something they can plug in and just work.
In some cases I can see why people are annoyed, also people are annoyed at the price of GPUs (I want an upgrade but I cant afford it) so I think there's a big reason for people who game to hate AI too.
I think that's all reasonable.
That said, the Indy shot is fake.
To a certain extent, this feels a bit like the Internet's reaction to a helmet in The Odyssey trailer.
Improving game graphics is the last thing I'd think to use as an argument against AI and machine learning. Developers lean too much on it. That's an argument.
I was thinking it probably is fake since the NVIDIA model basically infers based on the frame / developer settings of color grading, frame details etc.