In six months AI has gone from an idiot savant intern to a crappy consultant. I'd call that progress.
AI accelerates complex search 10x or maybe 100x, but still will occasionally respond to recipe requests by telling you to just substitute some anti-matter for extra calories.
or emit (or spew) pages of training data or output when you "please change all headers to green", which I experienced recently.
What you’re referring to is popular opinion. AI has become so pervasive in our lives that we are used to it and the magnitude of achievement has been lost on us. The fact that it went from stochastic parrot to idiot savant to crappy consultant is from people in denial about reality and then slowly coming to terms with it.
In the beginning literally everyone on HN called it a stochastic parrot with the authority of an expert. Clearly they were all wrong.
So can parrots. They'll gladly generate neologisms. I'm interested in how academics define "knowledge that doesn't exist".
> "knowledge that doesn't exist".
I said that term. So there’s no official definition but you already know that.
Basically it’s clear among everyone academics included that LLMs can rudimentarily do what humans do. That means composing knowledge and working things out to form new knowledge that doesn’t exist.
> That means composing knowledge and working things out to form new knowledge that doesn’t exist.
That's not a terribly useful criteria, though. A Markov chain can produce novel sentences, hell a bingo machine can if you write words on the balls. "Knowledge" is kind of meaningless but also seemingly profound.
I don’t know why you came up with this pedantic example. Perhaps you’re autistic? If so then I apologize for assuming you aren’t.
Everyone knows that we are talking about more than just knowledge consisting of a random sting of letters. We are talking about actual useful knowledge.
A certain amount of pedantry is required for these discussions, otherwise we're left in a place where we can't define "actual useful knowledge". At this moment I assume you're defining "actual useful knowledge" as simply anything you find convincing, which is a criteria that could be easily gamed. How are you determining that knowledge is actually novel?
I get it, you like AI, so much so that you're willing to throw out personal attacks to defend it, but it's important to be critical or it's easy to be suckered.
I don’t love AI. I hate it. But im not deluded for what it is.
* where "polly-want-a-cracker" is some form of existing, common fizz-buzz-ish code.
The term stochastic parrot has nothing to do with usefulness and everything to do with the existential meaning of whether this ai is repeating what it is taught or creatively forming new knowledge from logic and composition from previous knowledge.
It is categorically unequivocal that LLMs do not just parrot previous knowledge stochastically. They form new ideas from scratch.
AI code completion is god mode. While I seldom prompt for new code, AI code autocompletion during refactoring is 1000x faster than plumbing fields manually. I can do extremely complicated and big refactors with ease, and that's coming from someone who made big use of static typing, IDEs, and AST-based refactoring. It's legitimately faster than thought.
And finally, it's really nice to ask about new APIs or pose questions you would normally pour over docs or Google and find answers on Stack Overflow. It's so much better and faster.
We're watching the world change in the biggest way since smartphones and the internet.
AI isn't a crappy consultant. It's an expansion of the mind.
Tech is useful, how it's built is very unethical, and how it's worshiped is sad.
Maybe some folks need this, but the way I use this tech doesn't rely upon that so much. By the time results start appearing, my brain is already fast at work processing the output to qualify whether the information the LLMs return is accurate, whether it's a good leaping off point, whether I can keep drilling deeper, expand my prompt scope, etc.
I'm using it as search. Just as old search had garbage results we had to filter out, so do LLMs. But this tool is a way more advanced query language than Google ever supported. These tools are like "Google 9000".
It feels like I'm plugged into the Matrix rather than getting SEO'd garbage. I know the results have issues, but that doesn't matter - I can quickly draw together the pieces and navigate around it. Compared to Google, it feels like piloting a star ship.
Seems unnecessarily tiring. Instead I use a SEO spam and ad-free search engine. It's called Kagi. It allows me to further refine my search via lenses and site prioritization. Also, it has zero hallucination chance, because it's a deterministic search engine.
> It feels like I'm plugged into the Matrix rather than getting SEO'd garbage. I know the results have issues, but that doesn't matter - I can quickly draw together the pieces and navigate around it. Compared to Google, it feels like piloting a star ship.
Same for Kagi, without selling my data or trawling information obtained without consent or disregard of ethics, and many other things.
Note: I don't use any of the Kagi's AI features, incl. proofreading.
As I said, the network doesn't carry citation/source information. IOW, when it doesn't use a tool, it can't know where it ingested that particular piece of information.
This is a big no no, and it's the same reason they hallucinate and they'll continue doing that.
As a tangent, I see AI agents hit my digital garden for technical notes, and I'll probably add Anubis in front of that link in short order.
On the opposite end of the spectrum of worshippers there are naysayers and deniers. It’s easy to see why there are delusional people at both ends of the spectrum.
The reason is that the promise of AI both heralds an amazing future of machines and a horrible future where machines surpass humanity.
For the third time [1] [2], I'll divide the line between core network and tools that core network uses. Agents are nothing new, and they expand capabilities of the LLMs, yes that's true. But they still can't answer the question "how did you generate this code and which source repositories you did use" when the LLM didn't use any tools.
The core network doesn't store citation/source information. It's not designed and trained in a way to do that.
geez.
[0]: https://notes.bayindirh.io/notes/Lists/Discussions+about+Art...
Second the question you brought up can’t be answered even by a human. It’s a stupid question right? You blindfold a human and prevent him from using any tools and then ask him what tools he used? What do you expect will happen. Either the human will lie to you about what he did or tell you what he didn’t do. No different from an LLM.
The core network doesn’t store anything except generalization curve. Similar to your brain. You didn’t store those references in your brain right? You looked that shit up. The agentic LLM will do the same and the UI literally tells you it’s doing a search across websites.
Geeze.
> "Agentic AI is a class of artificial intelligence that focuses on autonomous systems that can make decisions and perform tasks without human intervention."
I can work with that. So we have agents that autonomously react to their environment, changes, or what we can say impulses. They sit there and do what they are designed to do, and do that autonomously. Makes sense. However, this sounds a bit familiar to me. Probably me hallucinating something, so let's dig deeper. There seems to be an important distinction, though:
> "Agentic AI operates independently, making decisions through continuous learning and analysis of external data and complex data sets."
So, we need to be able to learn, evolve, and analyze external and complex data sets. That's plausible, but my hunch is still lingering there, tingling a bit stronger. At this point, for Agentic AI, we need an independent "thing" which can decide, act, learn, and access external data sources to analyze and learn from them. In short, I need to be able to give this Agentic AI a goal, and it accomplishes it automatically with the things at its disposal. Fair enough.
We were discussing (software) agents and their history. So let's pivot more to agents. Again, turning to Wikipedia, we find this sentence:
> "In computer science, a software agent is a computer program that acts for a user or another program in a relationship of agency."
Again, a piece of software that acts for a user or another program. Hmm... They have five basic attributes: 1)are not strictly invoked for a task, but activate themselves, 2)may reside in wait status on a host, perceiving context, 3) may get to run status on a host upon starting conditions, 4)do not require interaction of user, 5)may invoke other tasks including communication. That hunch, though. It feels more like mild kicking. Where do I know these concepts? Somewhere from the past? Nah, I'm hallucinating. You told me that they are new.
As I skim the article and pass "Intelligent Agents" past, I see something very familiar line under "Notions and frameworks for agents" title: "Java Agent Development Framework (JADE)". I know this. Now I remember!
I have used this framework to code a platform where an agent gets orders from a client for a set of items, and submits them to another agent, where other agents send their best prices, and another agent calculates the best combination for the cheapest price. Doing a "combinatorial reverse auction" for a set of items. We had no time to implement feedback-based price adjustment strategies, but the feedback and announcement code were there, so every agent knew how the transaction went. They all were autonomous. A single agent acted on behalf of the user, and the whole platform responded to that without any humans at any step, including final decisions!
That was my Master's thesis. I have also presented it at the IEEE Symposium on Intelligent Agents, IA in Orlando in 2014 [0]!
When did I complete my Master's thesis?
Oh. 2010. 15 years ago.
Alright. This solves it.
Now, on to your second question. Let's put it right here:
> You blindfold a human and prevent him from using any tools and then ask him what tools he used? What do you expect will happen. Either the human will lie to you about what he did or tell you what he didn't do. No different from an LLM.
You're mangling my question here. The question I ask is different:
> Generate me a Python code for solving problem X, then tell me which source repositories you used to generate this code. Cite their licenses, if possible.
All of this information is in the core network for the first part of the problem. LLMs without tool capabilities can generate code, and generate it well. The source of this knowledge came from their training set, which consists of at least "The Stack", and some other data sources on top of that. So, the LLM can generate the code without any tools, but it can't know where the source came from. It's just there, in the core network.
You think the question is stupid, but it's not. This is where all the ethical questions regarding LLM training are rooted. LLMs hallucinate licenses, don't know where the code came from, and whatnot. If you ask me about a code piece in my source code, I can give you the source, the thought process, and design, citing the originality or how I found it elsewhere and got into my codebase. Lying about it would be a big problem in the light of licenses, but LLMs get scot-free because they're just fair using it. Humans can't do the same thing, why LLMs? Because their owners have money and influence? Seems so.
> You didn't store those references in your brain right? You looked that shit up.
No, when I looked that shit up, I recorded where I read it alongside other contexts, including the weather that day in some particular cases. I don't answer "I just know, I don't know how" when people ask me about the source of my knowledge.
This is the difference between humans and LLMs; this thin line is very important.
We are talking about agentic LLMs which have been around for about a year only. Not some bs pre LLM ai chatbot or some useless thing like that? Are you autistic? No joke and more respect to you if you are but to a non autistic person I am obviously talking about AI which in modern contexts means LLMs and agentic AI obviously means agentic LLMs
Once you started getting into your masters thesis I stopped reading. Conversation is over. Good day.
The gist is, Agents and ideas underpinning Agentic LLMs are 20+ years old, and agents were managing systems and keeping things up autonomously for decades now. JADE has been developed by Telefonica to keep tabs on the telephone infra, also since the agents can migrate, it was also the original edge computing, but I digress...
You don't have to give a damn about my research. The point is not that. You challenged my knowledge, and I shown you what I know, how I know, plus you read a small history of intelligent agents, to boot.
I don't know what you are trying to achieve with asking me being autistic or not. I'm not, and it doesn't matter. The way it comes is bluntly insulting regardless of my situation.
So yes, Agentic LLMs are new, but the Agents itself is not, and the agents I'm talking about are not dumb chatbots. They can wander distributed systems, process data, learn from that data, report their findings and optimize themselves as they operate. They are not just parrots, but real programs which keep infrastructures intact.
Since you're losing your temper, and getting into ad-hominem category, and seeing it's tea time here, I'll prefer to sip my tea and continue my day.
Thank you for the chat and insults, and have a nice and productive life.
Unless you know Vim!
or the IDE (or text editor for that matter) well. People don't want to spend time understanding, appreciating and learning the tool they use, and call them useless...
I don't bend the tool, even. It's what it's designed to do.
Even on a greenfield project, I rarely spend more than a day setting up the scaffolding and that’s for something I’ve not touched before. And for refactoring and tests, this is where Vim/Emacs comes in.
I've been surprised by this for a long time, having seen coworkers spend days typing in things manually that they could have put there with IDE capabilities, search-replace, find -exec or a five minute script.
I've used Vim bindings and strongly typed languages with IDEs that have strong AST-based refactoring my entire career.
Nothing comes close to changing one condition of a test and having the AI autocomplete magically suggest the correct series of ten updates that fix the test. In under the blink of an eye, too.
Everything is truly changing in big ways.
But why are you doing this? Granted, you may have a longer career than I do, but I never once think: The test condition is wrong, let's update it. Oh, I wish I could update the code alongside it!.