LLMs aren’t gonna give us any of those things. Thats just snake oil.

I believe at most, coding, scripting, and text processing.

Exactly, there’s a few specific use cases that the current generative AI is good at. Reading medical imagingand medication development are things I’d add to your list.

However, outside of those specific tasks AI is not particularly useful. An incremental improvement in productivity in some cases.

Medical imaging and medication development would be bespoke models no? Not standard LLMs?

Probably. I mis-spoke.

You can add transcription to your list though, it’s saving providers a lot of time on charting

Generally speaking, yes, but standard LLMs now have their own built-in image recognition so we might not be too far off from LLMs being able to describe to the doctor what they should look at on an image that’s been flagged. Take a capable LLM of a few hundred billion parameters, train it on tons of medical images and such, and it might actually be helpful. Or it might not, no way to find out until tested.

AI is not just LLMs. Much of AI is machine control, and bioinformatics.llms are the foam on top of the ocean.

Yeah, but the current hype is around LLM, the other kinda of AI are the same as they were before this started.

Alphafold almost certainly benefited from genAI research, since 2 and 3 are somewhat genAI inspired architectures, they use their own versions of transformers and also diffusion

I’m aware, and I wish people were more clear about the difference, but its LLMs that are driving these insane buildouts, not the other disciplines. This article is clearly talking about the LLM “AI”, not bioinformatics or anything that may actually be fruitful.

The LLM companies like to claim they’ll solve cancer, but its all snake oil hedged on a mythical AGI popping up and fixing everything.

I’ll believe that these things are possible when there’s some empirical evidence that they are.

Why don’t you buy my amazing car. Yes it looks like that car that your nan drives to the shops but actually it can fly and time travel as well.

Those results are still tightly controlled by NDAs. They need time for FDA approval, patents, etc. Then time to get to market. These are not released in scientific papers anymore. It’s protected because it cost so much up front.

Surely in the age of self-promoting narcissists we would hear about these break-throughs in all the media?.. Not just “it can potentially cure cancer if you give us another $tn”, the real results under review by relevant authorities/academics. Moreover, these breakthroughs do require a shitload of compute, but I do not see how a chatbot could be helpful with it. Other than, of course, stealing the preprinted work of some academic and claiming it their own, as recently with OpenAI Navier–Stokes equations conundrum. Can you imagine how many problems like that would already be solved if all these billions were spent on it, rather than chatbots drawing glossy pics?..

You may be missing my point. THEY aren’t being kept in the dark. YOU are. Anyone involved in the project signs an NDA. Then the relevant experts can’t talk about it.

This is big with legislators too. They are taken on tours under NDA. They see there is no water use, they listen to the noise levels, they see the new substation provided to the community. But they cannot tell the constituents those specifics. Only “it’s not like the news is claiming”.

And the news is not the customer. No one in the industry cares about bad press. Customers won’t stop buying. So the build out accelerates.

AI has been mainstream for what, 4 years now?

if AI had such powerful results for developing novel solutions to problems such as medicine or material science, why have we not seen major improvements in those sectors?

the way AI is spoken about is that it’s “world impacting” and “sector disrupting”. the only disruption and impacts I’m seeing are in human rights violations, price gouging, privacy negation, and corruption.

if these are the results after only 4 years, I can only logically assume it will only get worse with more time and support.

The PC became mainstream in 1981. Why didn’t the Internet become big for another 10 years?

bullshit. you’re talking about two different technologies.

the PC didn’t promise to cure cancer. it promised to help you run your business more efficiently.

Takes years to get new drugs to market. AI is speeding up drug discovery by predicting protein folding ridiculously fast, but there’s a lot of other work that needs to be done. Perhaps there will be models to speed up some of those processes too, but the real bottleneck is approvals and while AI might be used to speed up the paperwork part of it, I can’t think of any way it’ll be able to speed up actual trials. What I suspect will happen is we’ll get an even faster response time with the next global pandemic that requires a vaccine, and we’ll start getting more and more mRNA vaccines, including some for some types of cancers.

This paper details 9 processes in which AI is already helping or is predicted to be able to help in material development. Material development also takes time to get anything to market, for various reasons. But there’s a nonzero chance that the recent battery tech boom from China has been aided by AI. They just tested a very high density solid state cell not long ago.

if it wasn’t maliciously designed it’s absurdly negligent. both of which are reason enough to not use AI at all.

wouldn’t it make more sense to streamline the existing processes first before creating a traffic jam of “potential” cures? where at least the models could be trained on existing cutting edge medical research instead of flying blind and developing novel treatments and cures.

the more I think about it the more is seems plausible that the actions AI implementation takes is malicious negligence. incompetence derived from a malicious intent to do harm on someone or something.

is this the disruption AI execs were talking about?