The rapid spread of artificial intelligence has people wondering: who’s most likely to embrace AI in their daily lives? Many assume it’s the tech-savvy – those who understand how AI works – who are most eager to adopt it.
Surprisingly, our new research (published in the Journal of Marketing) finds the opposite. People with less knowledge about AI are actually more open to using the technology. We call this difference in adoption propensity the “lower literacy-higher receptivity” link.
I think this is true for a lot of things. iPhones, Nike, Spam
… Trump.
“Surprisingly”? This should be a surprise to no one who is paying any kind of attention to any online communities where techy people post.
Hey, buy my new CoinCoin! No, don’t research what it is, just buy it!
Most people do not pay attention to them.
I’m tech savvy and I use AI daily.
Probably not the AI you think of. As it’s not LLM or image generation.
But I have a security system self hosted using frigate, which uses AI models for image recognition.
I am a system admin and one of our appliances is a HPE Alletra. The AI in it is awesome and it never tries to interact with me. This is what I want. Just do your fucking job AI, I don’t want you to pretend to be a person.
They will come for you first /s
Even using LLMs isn’t an issue, it’s just another tool. I’ve been messing around with local stuff and while you certainly have to use it knowing it’s limitations it can help for certain things, even if just helping parse data or rephrasing things.
The issue with neural nets is that while it theoretically can do “anything”, it can’t actually do everything.
And it’s the same with a lot of tools like this. People not understanding the limitations or flaws and corporations wanting to use it to replace workers.
There’s also the tech bros who feel that creative works can be generated completely by AI because like AI they don’t understand art or storytelling.
But we also have others who don’t understand what AI is and how broad it is, thinking it’s only LLMs and other neural nets that are just used to produce garbage.
Image recognition has gotten crazy good
People susceptible to marketing gimmicks more likely to want marketing gimmick.
« Ignorance is bliss »
- Cypher
How exactly is this a surprise to anyone when the same applied to crypto and NFTs already? AI and blockchain technologies are useful to experts in tiny niches so far but that’s not the usual tech savvy user. For the end user it’s just a toy with little use cases.
AI is much more broadly applicable than Blockchain could ever be, although somehow it’s still being pushed more than it should be.
i think we give silicon valley too much linguistic power. there should really be more pushback on them rebranding LLMs as AI. it’s just a bunch of marketing nonsense that we’re letting them get away with.
(i know that LLMs are studied in the field of computer science that’s known as artificial intelligence, but i really don’t think that subtlety is properly communicated to the general public.)
here should really be more pushback on them rebranding LLMs as AI.
Those would be AI though wouldn’t they?
The pushback I would like to see is the rush of companies to rebrand ordinary computer programs as “AI”.
I actually think in this case it’s the opposite-- your expectations of the term “AI” aren’t accurate to the actual research and industry usage. Now, if we want to talk about what people have been trying to pass off as “AGI”…
i think that’s fair point. language does work both ways, and i am certainly not in the majority with this opinion. but what bothers me is that it feels like they’re changing the definition of the word and piggybacking off of its old meaning. i know this kind of thing isn’t all that uncommon, but it still rubs me the wrong way.
I mean, we’ve been calling pathfinding + aimbot “AI” in games for years. The terminology certainly does feel different nowadays though…
there should really be more pushback on them rebranding LLMs as AI.
That’s because the target of the language is the know-nothing speculative investor class. The distinction doesn’t matter to us because we’re not being sold a service, we’re being packaged as a product.
The increasingly-impossible-to-opt-out-of nature of LLMs/AIs illustrates as much. We’re getting force-fed a “free” service that’s fundamentally worse than what came before it, because its an extractive service.
What form of AI are we talking about? Because most of them exposed to the people are glorified toys with shady business models. While tools like AlphaFold are pretty useful.
I suspect it’s truly more of a dunning-Kruger situation. When you know nothing You’re down to use it for everything. When you start to understand the problems, limits and the morality of it, you start to back off some. And as you approach the ability to host it yourself and do actual work with it, you fully welcome the useful bits in your workflow.
This is honestly my exact experience. Albeit I’m far from an expert, but it’s great with document templates and code snippets.
At the state of AI today, it helps noobs to get to average level but not help average to get a pro
The real question in my opinion is how does a pro truly benefit from it other than being a different type of a search engine
Yea, if you are a pro in something it most of the time only tells you what you already know (I sometimes use it as a sort of sanity check, by writing prompts that I think I know the output that comes)
I only found it useful doing trivial chores such as converting between data structures, maybe create a test for a function, parsing and some regex. Anything deeper than that was full of errors or the it offered was suboptimal at best. It also fails a lot of times in fetching the relevant docs/sources for the discussion. I gave up trying after so many times it basically told me " go search for yourself"
I often use it as my Python Slave because I am lazy
Like i write in bad fast human Language what my Script needs to do and then iterate from there giving it errors/ bug reports back (and fix some stuff that am I not too lazy for myself)
Scrripts that I needed were in complexity like, API calls, serial communication or converting PO to CSV and back (pls don’t ask 😅 it is for Work and I can not tell more)
But I guess, that because my skill is not too high, I‘m sure, if I was more skilled, I might be faster just writing it directly as code 💁🏻
But for code that needs to be built (like C), I mostly use it to make it explain me what existing code does, if I am not 100% sure after a short read. Have tried some generated code there as well, but then I get nothing but build errors 😆 at least, it, most of the time, can tell what the build error is trying to say.
Ah, and currently, I use my free chatGPT to make it teach me how to make music using only open source tools 😄
I very much agree with your conclusions and general approach.
LLMs are great for certain tasks that are programming related and it does it very well. I, too, often find myself needing scripts that as long as they did what they were suppose to, I really didn’t care how.
Another thing I’ve noticed(which is probably related to amounts of training data) is that it can help better with simple Python tasks as opposed to how it handles simple rust tasks.
But you mentioned one of my main issues with. Ice been programming for 15 years or so, and still learning. All the available llms did crucial errors about fundamental tabd complex topics and got the answer so very wrong but also sounding very convincing. Couple it with lack of proper linking to the sources of the response, you might see why having it explain code might cause your learn wrongly. Although it is also possible to say this about randoms internet tutorials. I always try to remind myself that it’s a tool that produces output that always needs to be verified.
I often make in a new chat with a prompt including assumptions based on the info from output of previous chat. Most of the time, it then makes a good job factchecking itself and for example tells many things not matching with what it told in previous chats. Then you know that it has not enough training data in that regard and failed to get relevant infos from it’s web search.
More than once above happened to me on copilot (from enterprise ms365) and then chatGPT limited free promts saved me 😂
That tracks for sure. The most enthusiastic guys at work also happen to be the ones who put in the least actual work. Sure, it has some uses… but the things it gets wrong are significant enough that no sane individual should rely on anything that AI is involved with making/running. The intelligence part just isn’t there yet. People are effectively getting wowed by a glorified ELIZA chat bot.
the things it gets wrong are significant enough that no sane individual should rely on anything that AI is involved with making/running
The fundamental use-cases for AI are almost never customer oriented, either. You don’t see these tools deployed to reduce wait times or improve authentication or approve access, because the people who deploy them don’t actually trust them to do positive scope client interactions. What you see them doing is robo-calls, front-line customer service, claims denials, and (in the bleakest use cases) military targeting operations. Instances where efficiencies of scale accrue to the operator and an error/problems rebounds to the target of the service rather than the vendor.
People are effectively getting wowed by a glorified ELIZA chat bot.
An ELIZA chatbot that double-processes your credit card and then keeps denying you a refund when you manually catch and report it.
Its like mice and traps. The stupid mice get the most bendy necks as the trap slams at high speed on them.
I am skeptical that the people they put in the “understands AI” bucket have even a bit of a clue.
