im personally excited for it.
Has anyone done anything yet with ‘agentic ai’ that couldn’t be done better/faster/easier in other ways? So far, I’ve mostly seen ‘I spun up 20 agents to google how to make a tuna casserole on twenty search engines and one more to summarize the results. Please, ignore that this burns as much compute power as our entire office does on a random Tuesday and produces pretty much the same result as if I had just searched once and read the free AI summary, which in turn is slightly worse than if I had searched it with no AI involved and just read the first link.’
It’s a great way to encourage people to burn more tokens. If you want more reliability you need to focus on deterministic feedback and bubble wrapping everything.
Nothing to be excited about. All what those “agents” do is to Connect to LLM. There has been added more logic as in the past.
But still. It is a LLM. They get fundamental set ups. After launch they have to undergo dozens of risk assesment to meet governance Standards - that are actually pretty basic for the time being.
“Agentic AI” is nothing more than a new fancy word for investors and stock market. Those “agents” a very far away from being actually ready (as if those companies Care). Even creating this around a LLM needs a lot of know how and from my persoective it is just Not sustainable as of now.
I’ve written a couple.
My job requires a lot of troubleshooting, almost to the point of reverse engineering. I’ve spent many hours reading and getting through logs.
Now I have an agent who can one-shot stuff half of the time, and n-shot the 4/5 other times. And it learns from experience.
I enjoyed experimenting with it, and now it’s a timesaver for my team.
I don’t even like giving deterministic software access to my system. Now imagine all the bad things that can happen when you give access to non-deterministic software.
Cool concept in theory, SUPER annoying and risky because in federal use because there is no “undo button” for all the mistakes the agent will make though
Seems to be a fair amount of shilling for ai and LLM’s on Lemmy lately. This happening on other platforms?
how is talking about ai “shiling” ? its merely another tech.
You may aswell have came along and said you like eating shit. You likely have some lessons to learn in life.
I’m not impressed.
AI agents are a dream.
One that’s pretty far from reality.They don’t work at all right now, and there’s no evidence they’ll work reliably enough to be useful anytime in the next decade or more.
why should i feel bad?
It’s a joke in the shape of a picture. We call that a “meme” on the internet. We’ll explain what the internet is later, then I’ll make a 10 hours YouTube video explaining why agents suck big time.
lmao i understand what a meme is. i just dont get why the above commenter responded with basically ai hate. im more neutral or supportive of ai so im not really used to anti ai povs.
What’s good about AI or agents? I’m waiting.
i like learning new concepts from it,or discussing philosophy with it.
i want an ai agent so i tell it to do a task i would rather not do. like re encode a lot of movies then move them to another folder,or make specific art for me on gimp for example. so i have the original file not just an image,i can’t draw so i like ai art. i sometimes like to generate art to match something about philosophy im feeling and print it. stuff like that
You would rather ask an extremely unreliable piece of crap that accidentally destroys software engineering as a whole rather than find it yourself or read a book?
Note that I am NOT against every technology. Every new piece of software brought something useful in the past. LLMs is the only one that has only negative parts. LLMs only solves the laziness problem, and it does it wrong most of the time.
books are complex,ai i can ask it to simplfy it and it does.
- Excessive resource use for what it can produce thereby negatively impacting the broader environment
- Excessive noise and heat pollutants (among others if generating power locally) thereby negatively impacting local environments
- Used for trivial things that could generally be done faster or more accurately by a human
- Places the levers of available knowledge and social influence in the hands of a few powerful men
- Excessive use supplants your trained ability to critically think which can make you more dependent upon them in the future
- The generic LLMs we have today are being invested in at the expense of investments in other technologies which could have a greater positive impact
- People are losing their jobs to a machine that does it worse in a society with no recourse for protecting human life
Laugh as business spend billions on software that doesn’t work. Cry as those same businesses fire staff to cover the cost of broken software.
Like, they can’t even run pass maths tests or run a vending machine business. How do you expect one or a suite of them to manage your product lifecycle?
openai for example claims chatgpt can perform close to or above phd level math. so your claim about it not passing math tests isn’t really correct.
The math lie was a lie. It turned out that it’s some guy who did all the job. You’ve been lied to like every other sucker.
lol, sorry…the company selling a product made an outlandish claim and you believe it? I got a bridge to sell you!
I think the fundamental problem is that you haven’t actually used these tools enough to see that the companies peddling them are snake oil salesmen. Nothing they say is true, they’re completely full of shit, and they’re terrified they’re not going to make any of the money they promised their investors they would make.
And like…if you think about it for a moment it would make sense, given how these tools work, that they can’t do much in the way of reasoning. They just get better and better at replicating the language they’re supposed to mimic, which sometimes even carries the right content! Wow! But that’s only a secondary effect, right?
openai for example claims
Claiming something doesn’t make it true. If it were true. It wouldn’t be a claim it would be a statement of fact which is directly opposed to the point of marketing.
Seeing as you don’t support your claim that it can do permanent head damage maths, I will support my statement of they suck at maths.
so your claim about it not passing math tests isn’t really correct.
You base this statement on claims from the company that’s trying to sell the product. Maybe try to make chatgpt do some sums first before believing these claims.
That is a testable claim. I just had it calculate the force of gravity between 2 objects and it was wrong by virtue of rounding incorrectly. That’s like highschool level math.
what model did you use? you should use pro with thinking for math
Why should we have to pay extra for a computer to do maths correctly? That is basic functionality of a computer.
Chatgpt, as made available by the first search result for “chatgpt.” I chose this approach to match your claim you chose to repeat without verification.
I am not going to cross shop stochastic regressions with an English LUT in order to find one that works better for math just to appease your suspiciously wheel ladened goal post.
ok. but the free models are dumber than pro version. so chatgpt isn’t defined by its base model.
idc if you buy chatgpt sub or not. im just asking tech questions here
Its a buzz term they’ve been using to obfuscate the fact that every piece of LLM slop needs a human to review it, thus costing more time and money than hiring a competent human straight out.
i think if they used a strong model for local ai it wouldn’t need much reviewing if at all. but it would need a beefy computer.
I have to use them for work. Even the cloud hosted ones (which are arguably better than self hosted as they can have more resources thrown at them) need all of their output reviewed. They just aren’t up to the level of code generation as a senior dev yet (and if you think they are, I feel sorry for how unskilled your team is).
Locally hosted LLMs are a better option, but it doesn’t change the fact that the technology is designed outright to give a statistically likely result (based on the training data), instead of a correct one. LLMs make “guesses” about what the output should be and are not capable of internal regulation. They do not understand concepts and merely regurgitate jumbled data back at you with no consideration for content beyond that what can be analyzed statistically. It’s not a question of if they will make a mistake, but where and when.
https://youtube.com/shorts/ZMP8_jD-y0s These kinds of responses are an inherent feature of how they function, they cannot be fixed or tweaked out of the system.
