Hacker News Reader

Comment by liendolucas | Hacker News Reader

I'm gonna say something that probably some people won't like. I'm impressed by how much money the industry is moving by this so called "AI" movement and literally having zero concrete improvement on one's every day life (and by this I mean common people, not developers/techies).

I haven't seen, nor witnessed anything that you can benefit from all this AI hype. All that has done is opening the door for polluting the web with crap. And we keep on and on with the madness of training, wasting money (and resources) on gpus, generating tb of data. For what? To generate text, images?

If someone can actually point out a real example on the real world how all this madness helps to a better world, much appreciated, because honestly I don't see it.

View on HN

ChatGPT is far and away the best tutor I have ever had. It starts to fail once you get past mid level concepts but its ability to provide links to other sources, explain things, not get offended when you call it out as wrong about something (which human teachers certainly do) and many other aspects make it very helpful. Specifically its useful in expanding your breadth of knowledge on a subject. Less so depth.

Regardless I have learnt things and taken on projects I never would have attempted without it.

> ChatGPT is far and away the best tutor I have ever had.

> Regardless I have learnt things

who cares what we learn is true as long as we learnt something?

They alluded in their past comment that human teachers make errors too which students can identify. So perhaps if students don't implicitly trust human tutors, then so too can they corroborate ChatGPT against reality and solidify their understanding.

I don't know about where you're from but in my country to teach requires several years of qualifications, assessment and continuing observation by an independent regulator

and there are consequences if you're caught teaching rubbish

not to say they are perfect, but there is some level of quality control

meanwhile ChatGPT just bullshits endlessly with no possible way to punish OpenAI for it

I think perhaps you had a bad experience with GPT at some point or didnt ask it to provide sources that you then read through. Either way it is far more reliable than you seem to believe.

It's also getting better, while other resources aren't, and won't.

I've just listened to the latest Numberphile podcast, and one of many surprising anecdotes from Donald Knuth was that as a kid he'd been disillusioned by maths when he "proved" that 1=0 and his teacher couldn't find the mistake.

> meanwhile ChatGPT just bullshits endlessly with no possible way to punish OpenAI for it

Sure there are, and more importantly there's also ways to "punish" the models themselves — that's what the "thumbs down" and "regenerate response" buttons in the chat UI are for.

I'll join you in being downvoted.

Human teachers can be audited without advanced knowledge in maths, computer science, and information theory to discover what went wrong.

Human teachers can be held accountable.

A Human teacher has limited scale thus limited damage over bad information with more safeguards than "Big Corp promises not to be bad"

A human teacher can realize their own mistakes.

Human teachers have complex incentive structures that can modify their behavior.

Many humans (even ones with technical training) tend to see computers as infailible in first principals of math and carry that forward to deep learning which is where the math becomes tough to audit, couple that with LLMs affinity for overcoming the uncanny valley enough to seem like a rational human actor because of the datasets being used and you have "magic". This is before you even add financial incentive.

Not saying these issues in deep learning can't be tackled but the disengenous hype-train and the downvoting because people point out flaws with the new internet ATM is wearing thin.

Idk how you're promoting, but hallucinations are rare for me, even at graduate-level material.

There's a reason GPT-4 scores so highly on advanced exams like the USMLE.

> There's a reason GPT-4 scores so highly on advanced exams like the USMLE.

yes, it's called overfitting

There are entire industries devoted to helping humans overfit on standardized tests. Humans love to overfit, it's often easy and faster to get what you want--which in the case of education, is usually credentials, not knowledge.

> Humans love to overfit, it's often easy and faster to get what you want

The problem is Goodhart's. Tests aren't used to measure your ability to overfit them, but to measure your general knowledge. Obviously the latter is exceptionally difficult to measure so we settle for an easier metric. The problem comes from believing that the measurement well aligns with the thing we set out to measure. The same is for humans as is for machines and it is why it is insane to say something like "look, GPT did well on the BAR exam, it should be able to practice law." Because we explicitly do not want a legal system that follows the letter of the law (rather spirit of the law) and it's clear which one GPT would do. Yes, there's a problem that humans go this way too, but it's something we can solve if we are willing to admit the fuzziness of our metrics.

> hallucinations are rare for me, even at graduate-level material.

How do you know? If you're using it as a tutor you're using it to get knowledge that you don't know. Sure, it doesn't get things that are obviously (to you) wrong, but what about subtle things? At the graduate level subtly dominates and many concepts are counter intuitive or non-obvious in the first place (otherwise they wouldn't be expert knowledge).

I know because I check often with what I already know, or official material/explanations written by humans.

Gpt helps me connect dots and uncover useful details I didn't previously have, and that official explanations didn't include. And just about every time I've double-checked, it's right, even when it didn't make sense to me (counter-intuitive/non-obvious due to my mistakes in reading comprehension or previously erroneous concepts in my mind).

Okay, with this added nuance I'm not worried. But I think we should be explicit about this because it is difficult to distinguish if people are doing this or not and it is quite reasonable to believe that they are not.

I have no issues with anyone using GPT as a __noisy__ tutor. But we just must be clear it is easy to get fooled, difficult to validate, and easy to cause GPT to change its answer. There's a high rate of people using it as a non-noisy tutor, so we can't be implicit.

FWIW, I use GPT in a very similar way, and frequently. Especially with the decline of Google and especially for high level topics. It functions very well as a fuzzy search and often even when wrong (which is frequent for me when asking high level math questions, think post PDEs) there's still words or phrases that help me significantly narrow down a Google search.

Which any Google Search engineer should not be reading this as "replace Google Search with GPT/Bard"!!!!! They work in tandem!!! I CANNOT STRESS THIS ENOUGH and do not want to see the hell that that creates.

The level of cautiousness many feel about the risks of AI hallucinations feels a bit excessive in a world where many people can easily obtain chainsaws without the need to fulfill any significant prerequisites

This is a cynical take on what I think is a meaningful benefit that the above commenter is talking about. If you use ChatGPT to learn concepts and approach it with the caveat that you should make sure you understand what you are trying to learn (something that helps to verify that what is being presented to you by the system is "accurate"), it's a fantastic resource. That isn't to say it's true for learning all things, but I'm generally impressed and optimistic for ChatGPT to be as valuable a teaching/learning resource as any other I might find on the web.

Human teachers aren't omniscient, but often have absolute authority over their pupils. This flaw isn't unique to AI.

I have more confidence in a teacher, especially in niche subjects. One of the big reasons why is that I can ask a teacher a question, they will infer what I am trying to ask, and not easily flip flop. Because a teacher has confidence and GPT doesn't. A teacher will even find their own mistakes without prompting (though of course, everyone needs it sometimes. Just a teacher won't say that a kilogram of feathers and a pound of bricks weigh the same and then explain how 1 kg is 2.54 lbs).

Sure, no system is perfect, but not all imperfections are equal, and I'm tired of pretending like they are.

> Sure, no system is perfect, but not all imperfections are equal, and I'm tired of pretending like they are.

I think this is the actual issue, rather than the other specifics you raise — my anecdote of human mistakes includes having a 44kHz/16-bit audio recording of a whole university lecture where the lecturer insisted that nobody can record mocap joint positions for a mere dozen joints at a mere 24Hz/8-bit for more than a few minutes because that is too much data to store, and it was their confidence that kept them from updating.

But, back to the point.

Every instance of any specific AI release (GPT or LLM or whatever else) are identical; errors are therefore necessarily correlated. Humans get an advantage from our imperfections being un-equal, we benefit from the wisdom of the crowd (with conditions) and second opinions are meaningfully independent (again with conditions), but not so for any specific AI.

(Possible twist: Mixture of Experts).

That's not true even before mixture of experts because the systems are stochastic. They are only identical if we were to always initialize with the same starting seed and rng_state.

While they do indeed have a random number generator, I think the main strengths and weaknesses of these models (and humans) are in the structures rather than at the token temperature stage — I expect a human to answer any given question slightly differently at different times, but asking the same doctor for a second opinion as the first is considered the setup for a joke rather than a sensible idea.

This is the assumption of a flat and smooth latent manifold. If either of those conditions are not met, then seeds can have a wild effect on the model output. There's only bounds on the latent structure being locally smooth or locally flat btw. This is the underpinning of such works as the Lottery Hypothesis but there's a whole rabbit hole to dig down here and it is why Alpha Go did so poorly in round 2. This is an important part of the discussion of generalizability btw. I do not think your analogy works out and it is inconsistent with your prior comment, because you're right that anyone asking GPT is akin to anyone asking the same doctor. The second doctor is a second model (or potentially a second GPT checkpoint depending on how we want to consider divergence).

Very good. Now apply that healthy skepticism to the contexts where you should have been applying it in the past but haven't been, from Stack Overflow to Wikipedia to your high school and college teachers to the evening news.

Chatgpt will not get offended when you call it out. Even when it was right and you are wrong. You call it out and now your incorrect notion is gospel as far as chatgpt is concerned. How can that possibly teach you anything?

This is another false belief that a lot of people have. ChatGPT sometimes wont correct false statements but just as often it will. This is also why we read the source and dont believe anything anyone says without proof.

Professors and chatbots are both wrong all the time and its up to the student to extract useful knowledge.

That's the hardest thing for me is that it nearly always accepts your corrections. It will often change answers even if you just ask it if it is sure or suggest it might be wrong. It's weird to say, but GPT has no confidence. Weird because it spouts things out without concern for their validity, and we might call that "confidently incorrect." But it isn't confidence because GPT won't stick to its guns like a human would. So even some of our common notions are completely divorce from how the machine actually works.

My major concern when using GPT and especially with hearing how common it is for people to use it as a tutor, is that being a novice you explicitly want a confidently correct tutor. There are certainly good aspects of GPT but you're here to gain knowledge. Being a novice explicitly means you are going to have difficulties recognizing incorrect information so you put yourself in a concerning position of becoming confidently incorrect.

I believe there is more to it than just problematic or “almost correct” code examples after 7 retries with ELI5 explanations.

But part of me believes this is another “self-driving car” thing where in 10 years we’ll still be in exact same place.

Today I saw an interview from a Greek YT economist who was asked about BTC. His answer was “it’s been 10+ years that has bene around. We know by now that is not new or groundbreaking in any substantial way” and I agree. There are many tech that “if this and that” but we never seem to reach “that” point.

But self driving has improved. Maybe not as fast as people would like but its clearly better than it was 5 years ago.

You can't tell people you're "24 month from full self driving in every driving condition" every 12 months for a decade an expect people to continue drinking the koolaid

It's the same for "AI", it's happening in cycle since the 60s, nobody learns a damn thing, it's borderline a pump and dump scheme at that point

You're making a terrible mistake.

I can come back to this discussion in 10 years and we still won't have self driving cars.

I wouldn't even be surprised if personal cars are banned entirely before we reach self driving cars

We're not talking about self-driving cars. That's not a technical problem, it's a human problem.

No we can't. Instead we can expect people to start dismissing it, despite the fact that it keeps steadily improving, so when it finally gets good enough it takes everybody by surprise.

> But part of me believes this is another “self-driving car” thing where in 10 years we’ll still be in exact same place.

I want us to be still having these discussions in 2034, because that would mean I'm still capable of gainful employment for the next ten years.

But a mere 3 years ago, what ChatGPT is doing now — even with the mistakes — was SciFi level of natural language processing.

Not sure if this is related to a better world, but I know that some of my family members use it to write nice, "thoughtful" words to family and friends. Like for birthdays and stuff. There's something in me which strongly goes against this kind of use, but there's your example.

They also use it to learn about people and places, a bit like using Wikipedia, which they never used.

Regarding non-dev uses, I use it to simplify and "neutralize" a text I'm writing to others, like getting a second opinion on the text. For example for the homeowners association.

For example now I passed `What does this mean? "Regarding non-dev uses, I use it to simplify and "neutralize" a text I'm writing to others, like getting a second opinion on the text. For example for the homeowners association."` and read the answer to see if it coveys my message properly. The answer was helpful: "[...] This could be particularly useful in ensuring that the message is clear, diplomatic, and devoid of unintended connotations, which is important in formal or communal communications like those in a homeowners association context."

> like for birthdays and stuff

I would never look them in the eyes again if I found out someone did this to me.

> I know that some of my family members use it to write nice, "thoughtful" words to family and friends.

Which makes those expressions have about as much value as the ones in greeting cards: almost none.

The foundamental question is: is making people more productive "making world better"?

For example, ChatGPT has already brought huge benefits to all the non-native speakers who work for/with companies in English speaking countries. Does it count as "making world better"? Or it makes lives of white-collar workers in these countries worse, since it brought more competition in?

You are on the right track. The truth is whether its making the world better or not is unknowable. To know it you have to know the future. Which is why people who strive to make value judgement always do so by predicting the future. But even if you get it right, the future has future still. Whats immediately bad might be good for the big picture, vice versa. This is doubly true for complex things like AI.

All I'm saying is that with thing like AI its entirely dishonest to say its making the world better with a straight face. It is what it is. Don't fool yourself by thinking you're doing it for 'the greater good'.

A couple of examples:

Using RAG on a set of docs for a project it's possible to get good answers to specific questions (and ask follow ups, etc) - really speeds up understanding, like having someone who knows their stuff answering questions for you all the time.

Making meeting notes, summarizing discussions - frees the note taker and produces a nice record of what happened.

And even that application is leading to this abuse which kills the benefit:

Person A: ChatGPT, expand these bullet points into a ten page presentation.

(sends to person B)

Person B: ChatGPT, scan this ten page presentation and give me the main points.

Yeah but someday, the ChatGPT service will fail, people will send the bullet points directly because that's what they're accustomed to writing now, recipients will be pleasantly surprised to see the summarizer working automatically, and countless hours of wasted human effort will be saved with no further computation required.

To me, that is like printing out an document and scanning it later to send to someone. It happened a lot in the early days of email.

I don't know about anyone else, but for me the value in taking notes is the actual process of taking the notes, not the end product of the notes themselves.

Having AI (or, indeed, a human colleague) do it for me is a bit pointless, as it eliminates 90% of the value.

You might want to specify that you're talking about generative AI specifically, as opposed to LLMs in general. I abhor stuff like what's in the OP, but don't really have a problem with using it as, for instance, a research tool, with human experts actually confirming its output: https://www.newscientist.com/article/2411374-ai-comes-up-wit...

Kind of reminds me of this screenshot someone took of the fastest growing subreddits back in 2023, and all of them were AI-related, but here's the kicker, I think 8 out of 10 of them had to do with AI waifus and porn.

To be fair that's just human nature, which AI can't really change. Even if we achieve AGI, a lot of the most asked questions will probably be money/sex related.

I'm impressed by how much money the industry is moving by this so called "AI" movement and literally having zero concrete improvement on one's every day life

One small thing: Amazon.com appears to be using an LLM to provide a summary of all of a product's reviews. But aside from that, I haven't seen anything useful out of "AI" yet.

> If someone can actually point out a real example on the real world how all this madness helps to a better world, much appreciated, because honestly I don't see it.

Anecdote time.

My sister, who doesn't care much about computers, found ChatGPT useful in her job, as it gave her a template to build from that included things she'd not thought of initially. I've just listened to the latest Numberphile podcast, and was surprised to hear Donald Knuth saying similar about a handful of the responses it gave him when he was playing with it.

I play around with the GPT models regularly, even before ChatGPT came out, and it broadens my abilities by producing code in languages I don't have much professional experience with, but which I can verify.

I sometimes use it for translations, and have occasionally tried using it as a foreign language conversation partner (it's OK at that, one limitation is that it keeps going back to English, though this is also a problem I face in real life with actual humans).

There are lots of groundbreaking applications. But I’ll give you one that by itself justifies the technology.

Early research shows that AI is able to detect cancer 4 years earlier than a doctor traditionally would.

First, I love how “AI” is just a meaningless term here. The topic was generative text and images, and now you’ve brought up image classification, a different problem. But hey, computers!

So let’s bring it back to generative systems. IBM Watson was a complete failure, generating nonsensical diagnoses and treatments. https://slate.com/technology/2022/01/ibm-watson-health-failu...

> I'm gonna say something that probably some people won't like. I'm impressed by how much money the industry is moving by this so called "AI" movement and literally having zero concrete improvement on one's every day life (and by this I mean common people, not developers/techies).

> I haven't seen, nor witnessed anything that you can benefit from all this AI hype. All that has done is opening the door for polluting the web with crap. And we keep on and on with the madness of training, wasting money (and resources) on gpus, generating tb of data. For what? To generate text, images?

> If someone can actually point out a real example on the real world how all this madness helps to a better world, much appreciated, because honestly I don't see it.

The comment I was replying to did not say LLM or Generative AI. It didn’t specify LLM and simply asked for examples where AI is making good.

Instead of saying, “is there a source?”, or taking 2 seconds to Google “4 year cancer detection with AI” people are lighting up the comments about failed AI from IBM from ten years ago.

Here is the source:

https://www.unilad.com/technology/ai-detected-breast-cancer-...

Also re Watson. Top Google article for the search terms “Watson medical ai” leads to this article, which contains the best response to your criticism:

> I don’t think that the failure of Watson means that artificial intelligence isn’t ready to make significant improvements and changes in health care. I think it means the way that they approached it is a cautionary tale that lays out how not to do it.

https://slate.com/technology/2022/01/ibm-watson-health-failu...

Whether LLMs deserve all this hype is debatable, but whether there is merit in AI technologies being used today is not. There are some pretty good use cases in medical, pharma and biotech, agricultural, etc. Few of them are based on LLMs but that wasn’t what the commented was actually asking. They were asking if “the AI hype” had any real applications, and it does.

LLMs are just the manifestation of the Gartner hype cycle anyways. And we haven’t gotten anywhere near the terminus for them either, with almost every vendor and open source project planning to at least double model size in the next year as hardware and GPU capabilities expand. I think we have yet to see what LLMs can really do and will be much more pleased with them in 2-3 years as they mature.

They have been talking about AI being able to do better than doctors at cancer detection for over a decade now: https://www.wired.co.uk/article/ibm-watson-medical-doctor

But it’s not Watson that delivered.

https://digialps.com/the-first-ai-medical-device-that-can-de...

That’s like pointing to GM’s 1990s electric car while explaining that electric cars haven’t delivered.

Yeah I totally was not aware of that. We are specifically talking about being better than doctors. I said that word for word.

> AI is able to detect cancer 4 years earlier than a doctor traditionally would.

Which "AI" ?

Which "cancer" ?

What's the false positive/negative rate ?

Regardless of all that, if it detected it 4 years earlier it means it happened at least 4 years ago and have nothing to do with the LLM hype gp's criticising.

> There are lots of groundbreaking applications.

> proceeds to give absolutely no example nor source

Yeah sure.

I expected people already knew about this or would use Google.

Here’s a link:

https://www.unilad.com/technology/ai-detected-breast-cancer-...

I don't think making the world better is their goal, so I'm not surprised you aren't seeing those results. Their goal is to either make more money, cut costs, or both.

ok I'm biased because I work in exactly this sector: automated meeting minutes make life easier and are powered by LLMs.

US employees spend 18 hours / week in meetings so that is a noticeable improvement IMHO.

I'm pretty sure everyone made a version of that. Makes me think, if it would be useful to open source that hodgepodge of scripts

yes I've definitely thought quite a few times about open-sourcing our company code (AI note-taker for remote meetings). But then the hard thing is Kubernetes on top...

And as always not sure our "closed-source", "employee only" code would be useful to anyone not working with us :/

Is that real time? Does it have to be?

Google, Microsoft and Zoom will all have it integrated soon enough and many companies will pay extra for it because of “security and privacy” guarantees the tech companies will make.

Slack already has, but it's not stellar to be honest

Yes sure. But that’s not anywhere near ground breaking. Also the math here is akin to those who believe that if they didn’t smoke they’d be driving a Ferrari. In reality owning a Ferrari is way more expensive than your monthly cigarettes and comes with belles and whistles (luxury taxes, maintenance, etc).

AI could change our lives completely over the course of a decade but if each change is just a little improvement that builds on the last one, the whole time people could grumble that "that's not anywhere near ground breaking."

The way I imagine it (sometimes spot on, sometimes hilariously wrong) large companies could eventually have really good company representatives that answer phone/email/chat immediately and are deeply familiar with the company and the customer. Many complaints can be converted into polished product/feature proposals and bug reports. Eventually they should be able to have smaller constraints, draw up plans, fill agendas and make decisions within their budget. A bit like algorithms are trusted to trade stocks.

Seems funny, have a bot from acme inc, give it a list of job descriptions, assign a budget. acme then figures out how to extract as much money as possible.

In the end, humanoid robots running the company, doing meetings and interviews all over the world simultaneously.

Walking around the factory, hey Jim, how is the wife? Oh, things are slow today, I'm going to give you the rest of the day off. Not real empathy but simply the most profitable manipulation.

I'm not impressed that much either (yet) but this is a little unfair, ChatGPT is barely a year old and while there is a lot of hype and talk it's also having concrete effects on the real world -- it was a key sticking point for the Writer's Guild and SAG-AFTRA during their strikes last year, and why they lasted so long.

> If someone can actually point out a real example on the real world how all this madness helps to a better world, much appreciated, because honestly I don't see it.

US government killbots will bring peace in our time!

It's just taking a little longer than we had hoped to perfect the "friend or foe" module

Small increase in economic productivity. Like the invention of the word processor or spreadsheet. It's hard to quantify but it trickles into the economy in the form of lower prices for the same goods (or prices that aren't as high as they otherwise would have been).

>it trickles into the economy

Are we absolutely sure about that?

I didn't say trickles down. My claim is that it leads to GDP growth. How it gets allocated is another question.

Sobering read in an age of alcohol intoxication ("AI").

There is a project that tracks/curates "AI" in the context of bad use or real world (mis-)application [1].

[1] https://incidentdatabase.ai/

What you’re noticing is a sort of active “commodity fetish” with this AI stuff.

Just like NFTs only bigger. Just like 80’s era diet pills. Or 90’s era whatevers.

It’s a facet of capitalism to drive consumption and move to profit as quickly as possible. I think we are still in the “gauge and gamble” phase, but eventually in a season or to we will see it hit and bilk the general public.

We all know “AI” has been a “thing” since ML or Soundex or PID loops, or small models, etc.

It’s just “finally” hit a downhill allowing some momentum to build.

> I haven't seen, nor witnessed anything that you can benefit from all this AI hype.

AI translators are opening up the whole world to communication. I can talk to and do business with people even if we speak no common language. I can parse information from any language now.

Incredible that you hackers are down voting this, I answered his question exactly. AI translators are very useful tools for international business and communication. Some languages were impossible for machine translating up until a few years ago. This has a real impact on the world for non-tech people. It's not a toy.

Reply on news.ycombinator.com