| ▲ | tegling 5 hours ago |
| Why must there be a dichotomy: "use AI/don't think for yourself", vs "don't use AI/think for yourself"? I think we'll need to grow in our individual relationship with AI. Can't AI be that persuasive and contrarian, at times adversarial, friend? Someone you have to be somewhat wary of, but can provide very thoughtful, intelligent and challenging conversation? Such a "friend" obviously never replaces knowledge. If anything, you'll need to arm yourself with as much knowledge as possible to engage with that friend. |
|
| ▲ | eloisius 4 hours ago | parent | next [-] |
| Skipping the “friend” criticism as you’ve already addressed it in sibling comments, but another reason among many that even trying to use it as an adversarial interlocutor is flawed thinking, is that even if you prompt it to debate you, you’re the one still prompting it to do that. It can never _mean_ to contradict your argument with its own unique viewpoint. It can only argue because you gave it that prompt. Ironically it’s still a sycophantic position. |
| |
| ▲ | tegling 4 hours ago | parent | next [-] | | Appreciate your response! It was an honest question and this really helps. I think it explains my feeling of getting less "brainstorming" use of AI lately. We seem to be moving from dumber models that only provided a few different views (including stuff I didn't know) to sophisticated models that really "understands" what you want to hear and tailor for that. Any "knowledge" introduced to you in this process is only a side-effect/a strict non-goal. Rather, it wants you to feel engaged and happy with the response.
Flipping this around: could it be helpful to develop some kind of "concensus models" filled with knowledge and with the sole purpose of teaching/providing you with that corpus? Or, to create some kind of mean-reversion effect of opinions at societal scale (rather than promoting "bubbles")? But at that point, we're very close to the subjects of propaganda and ideas of monoculture... | | |
| ▲ | lproven 4 hours ago | parent [-] | | > Appreciate your response! It was an honest question and this really helps. I This reads increasingly like you're using a bot to write your answers. | | |
| ▲ | tegling 3 hours ago | parent [-] | | Or (sad, but true) learning manners from bots. Oh boy, where is this world heading? | | |
| ▲ | nisegami 3 hours ago | parent [-] | | I really struggle to see how someone "learning manners" could be construed as a bad thing? I feel like the world would be a lot kinder if everyone tried to be a bit more like the stochastic parrots tuned to be pleasant and polite. | | |
| ▲ | rsfern 2 hours ago | parent [-] | | I think the issue is that “bot manners” is often vapid filler. Like GP didn’t even ask a question so “it was an honest question that helps” doesn’t make sense in context. “Hey that was an insightful comment” would be a more engaged polite response IMO, “bot manners” kind of signals disengagement |
|
|
|
| |
| ▲ | nisegami 3 hours ago | parent | prev | next [-] | | >It can never _mean_ to contradict your argument with its own unique viewpoint. That just doesn't line up with my experience (depending on how strict you about about "unique"). I don't even ask it to debate me, but if I'm discussing a concept and make a claim about my own unique viewpoint it will readily point out things I may not be considering or things I appear to have misunderstood. It's shockingly effective at poking reasonable holes in things. | |
| ▲ | squidmaster 3 hours ago | parent | prev [-] | | [dead] |
|
|
| ▲ | bryanrasmussen 3 hours ago | parent | prev | next [-] |
| >Why must there be a dichotomy: "use AI/don't think for yourself", vs "don't use AI/think for yourself"? because studies show that use of AI in subject areas of which one has expertise consistently decreases that expertise? As these subject areas are generally ones with intellectual requirements "use AI" = "Don't Think for Yourself", which underlines well known common sense aphorisms such as "use it or lose it", where the it is the expertise that one stops to use in favor of AI. You lose your ability to think for yourself in your field of expertise, but you gain the AI. |
| |
| ▲ | tegling 3 hours ago | parent | next [-] | | That's kind of the whole question. You're correct, and we're likely heading towards a complete brain-drain if we don't adjust something. But those studies are descriptive. It's what it looks like now. I'm wondering what we can do in order to both have our cake and eat it. How to keep skill and knowledge mastery in humans AND use AI. And it's not helping that education/internships/junior positions are all pivoting for AI use rather than providing a proper learning environment with relevant friction, a curated corpus and time to mold our minds until we truly grok stuff ourselves. | |
| ▲ | audience_mem 3 hours ago | parent | prev [-] | | do you think it's not possible for a person to be doing the thinking but AI to be doing the gruntwork? (eg coding) | | |
| ▲ | bryanrasmussen an hour ago | parent | next [-] | | if I thought that then it would be up to me to explain why the studies that show this is not the case are wrong, if you think that it is then it is up to you do to so and not to try to put the task on to me. In other words, the data shows it is maybe not possible, you think it is, describe why the data is wrong and you are correct. | |
| ▲ | galactus 2 hours ago | parent | prev [-] | | coding = thinking | | |
| ▲ | zahlman an hour ago | parent [-] | | I don't think this is actually true, or rather it's hiding something. We used to not be able to separate the gruntwork from thinking, because even at the level of typing out an individual line of code there's an instinctive review process. But this kind of "thought" is on a different level from the "thought" which led to starting the typing. Not even all the way up at the architecture level, but more the idea of just extracting the essential complexity of the task from the incidental complexity. When I write code by hand and I feel confident about what I'm doing, the confidence comes from knowing that the extraction is generally already done. But I'll still find myself pausing and hesitating, checking for typos, making sure things correspond to their surrounding context… even after decades. It feels like "thinking" in the moment, but not so much when I reflect back on the nature of the task. AI seemingly is not limited by psychology, so it doesn't have these hangups (and of course it also just "types" far faster). It doesn't get an ego boost from being right the first time, and its "confidence" in the compiler, tools and tests is just not relevant. |
|
|
|
|
| ▲ | lproven 4 hours ago | parent | prev | next [-] |
| > Can't AI be that persuasive and contrarian, at times adversarial, friend? No. It generates text like that in its corpus. If there is a pattern that looks like the interaction you ask for, it will imitate it, but it can't be adversarial because it can't think. So you can always derail it. |
| |
| ▲ | zahlman an hour ago | parent [-] | | Yes, you can always derail it. Just like how, in the pre-AI world (or ignoring AI), you could choose not to examine your beliefs. But now you have a tool that makes it easier, if you choose to try. Given that the LLM has a goal to be sycophantic, by pretending to have different beliefs from your own, you may learn something about how the existing evidence is used to be sycophantic towards others, which is probably the best insight into why they hold those beliefs that you'd ever get. All you really need is the ability to represent those beliefs accurately. (And then, the "guardrails" on LLMs, in combination with biases in the training data, plus various effects in the training process, all unite to give the model a default "perspective" which may well differ from your own.) |
|
|
| ▲ | lelanthran 4 hours ago | parent | prev | next [-] |
| > Why must there be a dichotomy: "use AI/don't think for yourself", vs "don't use AI/think for yourself"? It doesn't have to be any extreme, although your next statement: > I think we'll need to grow in our individual relationship with AI. makes me think that anyone who develops a relationship with a tool is probably on one end of that extreme already. |
| |
| ▲ | tegling 4 hours ago | parent | next [-] | | It was good to get a reality-check. I'll reconsider my approach to AI analogies as I realize anthropomorphizing is understood as problematic. Still, as you can see in one of my other comments, my key issue is that AI interaction interfaces are designed to mimic human interaction. That kind of side-steps our whole mental model of non-human things. I'm at loss here.
Last, just want to clarify that by "relationship" I meant as in "the way in which two or more _things_ are connected", not as in human relationships. | |
| ▲ | jackling an hour ago | parent | prev | next [-] | | Aren't you reading the comment a bit too literally? It's the same as saying "relationship with food". It's about how you interact with it as an entity versus some personal relationship. | |
| ▲ | DonHopkins 3 hours ago | parent | prev [-] | | >makes me think that anyone who develops a relationship with a tool is probably on one end of that extreme already. I have no regrets naming my cat Emacs. |
|
|
| ▲ | noir_lord 5 hours ago | parent | prev | next [-] |
| It’s a matrix of numbers not a friend. Strangely I’m not friends with my toaster either. |
| |
| ▲ | tegling 5 hours ago | parent | next [-] | | Completely agree it's not human and anthropomorphizing might not be constructing. But I think we need to figure out _some_ kind of analogy that doesn't have us leave or brains in the drawer when interacting with it. Given ample sci-fi lore about all-powerful "Godlike" AI machines, playing the "machine" analogy might be equally problematic. If people think "well, obviously it's not a toaster, it's skynet", or some such. | | |
| ▲ | delis-thumbs-7e 3 hours ago | parent | next [-] | | We do create relationships ships with tools. I used to do plenty of manual labour in the past - railway and building construction, AV installations etc. and I have a degree in sculpture - and having a trustworthy Makita or even better a Hilti -driver as your companion felt like an assuring companion in the cold cruel world. Like one tool that is a bit special and has become almost a part of your arm. “There’s many like it, but this one is mine”. And of course probably everybody here has created a some sort of a loving friendship with their computer. Personally, I have a huge crush on Neovim at the moment. I don’t understand how I ever got anything done before I started using it. I use Kimi K3 on OpenWebUI running on my desktop, which I can access via Tailscale on all my devices. Personally I really dislike using LLM’s to generate code - luckily I don’t work as dev - a mental block I need to solve at some point, but in researching and learning it is wonderful, when you use a system prompt to force it to search for info instead of hallucinating from training memory. It is just so much quicker than researching it with web searches yourself. To me it is a really sharp good knife that helps me to do what I want to do. Not a friend in a human sense, but a companion. Cowboy has a revolver, a samurai his sword… a hacker an nvim and llm, perhaps? | |
| ▲ | jurgenburgen 5 hours ago | parent | prev | next [-] | | It’s an algorithm, a shiny hammer. | | |
| ▲ | hardbass 4 hours ago | parent | next [-] | | So are humans. Everything is an algorithm oer current physics knowledge. We can yave discussions on ai consciousness on various points but absurd statements like assuming humans or anything else are anything beyond molecules and algortihms is an absurdity of such a level I am not sure how someone can come up with it. | | |
| ▲ | ThunderSizzle 2 hours ago | parent [-] | | Humans have souls. There's a spiritual element we're not fully able to measure and see right now. Perhaps we might in the future. The synapsis energy being visualized is probably the closest we are at right now, but there is a lot we don't know. | | |
| ▲ | hardbass 2 hours ago | parent [-] | | All well and good, but why is this belief in souls given any seriousness outside of fiction until that day? I will at least say your idea of a soul is computable and yes it makes it a better idea of souls than most religious versions of souls but also therefore, sorry to say, nothing that can't be implemented by other systems. |
|
| |
| ▲ | tegling 5 hours ago | parent | prev [-] | | It's a motor-powered quadro-head hammer with unlimited nails (as long as you pour x% of world gdp into the hammer industry). It builds tree houses, actual houses, societies and food. Mostly by itself. It can also shoot you in the face. And make everyone fight over it. If used in production, great things can happen. Or some city gets destroyed. But, darn, isn't it SHINY! |
| |
| ▲ | ethnrob 5 hours ago | parent | prev [-] | | Yes but that's an easy thing to say for people that already "think for themselves", this could be (already is, 4o) a dangerous precedent to set. |
| |
| ▲ | tigerlily 4 hours ago | parent | prev | next [-] | | I'm a rubber duck being chewed by a dog on the internet, and my best friend is a matrix of numbers masquerading as a toaster. | |
| ▲ | drivebyhooting 4 hours ago | parent | prev | next [-] | | And stranger yet, I’m not friends either any humans either.
I’m not sure what happened after 15 years of working and being inundated with family responsibilities. | |
| ▲ | hardbass 2 hours ago | parent | prev | next [-] | | Of course, and we are just a collection of dumb inert atoms. | |
| ▲ | exe34 4 hours ago | parent | prev [-] | | Carbon chauvinism at its finest. Your "friends" are just meat bags with electro chemical neural networks. | | |
| ▲ | hardbass 2 hours ago | parent | next [-] | | Its frankly already absurd enough that religious beliefs are treated diplomatically, these days a lot of non religious people are also showing beliefs in souls. Why aren't statements about magic beyond atoms or algorithms in humans or any other physical system not laughed out the room by default is truly strange. | |
| ▲ | zahlman an hour ago | parent | prev [-] | | > Carbon chauvinism The fact that you ascribe meaning to this term and expect to be able to use it to win arguments already tells me that you are way outside the Overton window. |
|
|
|
| ▲ | pandemic_region 5 hours ago | parent | prev [-] |
| I would label AI as friend in the same way I would label ads as informative. |
| |
| ▲ | tegling 5 hours ago | parent [-] | | Agree it's not a friend. Hence, the quote marks on "friend" the second time. I'm just fumbling for analogies that advises for certain caution, highlight the value of one's own knowledge of actual facts, and acknowledges the interaction interface we currently have for AI: which is to mimic interaction with other humans. Any ideas how to capture those things in a mental image that we can relate to? |
|