- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
Absolutely needed: to get high efficiency for this beast … as it gets better, we’ll become too dependent.
“all of this growth is for a new technology that’s still finding its footing, and in many applications—education, medical advice, legal analysis—might be the wrong tool for the job,”
Each winter marks the beginning and end of a generation of AI. We are now seeing more progress and as long as there is no technical limit it seems that its progress will not be interrupted.
What progress are we seeing?
In what area of AI? Image generation is increasing in leaps and bounds. Video generation even more so. Image reconstruction for games (DLSS, XeSS, FSR) is having generational improvements almost every year. AI chatbots are getting much much smarter seemingly every month.
What’s one main application of AI that hasn’t improved?
Which chatbots are getting smarter?
I know AI has potential, but specifically LLMs (which most people mean when talking about AI) seem to have hit their technological limits.
Copilot, ChatGPT, pretty much all of them.
Smarter how? Synthetic benchmarks?
Because I’ve heard the opposite from users and bloggers.
Advanced Reasoning models came out like 4 months ago lol
Advanced reasoning? Having LLM talk to itself?
Yes, which has improved some tasks measurably. ~20% improvement on programming tasks, as a practical example. It has also improved tool use and agentic tasks, allowing the llm to plan ahead and adjust it’s initial approach based on later parts.
Having the llm talk through the tasks allows it to improve or fix bad decisions taken early based on new realizations on later stages. Sort of like when a human thinks through how to do something.
Lul yes but no, but they are clearly better at many types of tasks.
For example? Citations?
Pretty sure these “tasks” are meaningless metrics made up by pseudo-scientific grifters.
Small bits of code, language related tasks, basic context understanding, not metrics I have literally measured simply noticed has improved compared to non reasoning models in my homelab testing. 🤷♂️
AlphaFold 3 which can help in the prediction of some proteins. Although it has some limitations, it cannot be used in all cases, only in what it can perform without any problem.
deleted by creator
They’ve been a boon for medical diagnoses as well, I believe.
Has anyone made AI powered accounting software yet? I’d love to tell my computer ‘Here’s all my financial information in a big heap. Do my taxes.’ The numbers and tax laws are all known things. It shouldn’t be hard.
Any strictly rule-based system, like accounting and taxes, is a job for traditional software, not AI. Particularly when the laws change every year.
Once it has the information in a recognisable format. Reading and recognising random receipts, bank statements, payment slips, and whatever and sorting it into a coherent format is what I’m trying to avoid.
I see. So AI for gathering the information to put into the accounting/tax software?
That’s a more reasonable ask, but I wouldn’t personally trust AI with that. I’ve done something similar in games where I take a picture of something on screen and ask AI to collect all the information from many similar pictures into a table. It’s definitely good enough for gaming, but it makes mistakes often enough I wouldn’t sign my name attesting to the truth of anything it produced, you know?
Fair point, but i feel like that’s something that’s technologically solvable, and this is dealing only with text, a lot of which is already digital, just in multiple formats, and all easily checkable against the final figures if anyone so desires.
As a random aside, I saw a clip recently where someone had asked an ‘AI’ model to reproduce a photo with zero changes one hundred times. There were more than zero changes.
Surprisingly, the mistakes ChatGPT made weren’t related to picture processing. Every time I’ve sent a picture, it has flawlessly analyzed the text (even if it’s a screenshot of a massive Linux log or a screenshot with multiple windows / arbitrary text placement). The problems were more like the markdown table I created would not be reproduced perfectly with the new changes/additions. It’s pretty reliable early on, but either as the chat gets longer or the table does, fidelity can be lost. Not very often, but it does happen.
Just to clarify. But I find as long as you’re paying close attention and can catch mistakes or verify the output, AI does make such tasks much less tedious.
NVL72 will be enormously impactful on high end performance.