Hacker News Reader

Comment by skocznymroczny | Hacker News Reader

Am I the only one who's not very concerned about ChatGPT and "AI" in general? I am young, but still lived through several hype phases. I remember when 3D TVs were going to be mainstream and 2D was considered legacy. I remember when PC was to die soon and to be replaced by smartphones. I remember when VR was to become as common accessory as a game controller. It's 2023, and I still don't have automated self driving car that can get me to work. At work I am still using the boring old keyboard and monitor. I am not using a VR headset to connect myself to shared office space inside of a metaverse. Oh and I don't have to call my hairdresser for an appointment, because my phone will use artificial intelligence to do that for me (remember that? yeah it was 5 years already, where's my magical AI tech).

I played with technologies like Stable Diffusion for a while. They are fun to use for a while, there are too many unsolved issues such as coherent style, stable style transfer for videos an despite my best effort every second image will have a human character with two heads or four arms.

I feel like ChatGPT is similar. It makes for a fun parlor trick, but when it gets things wrong, it gets them very wrong and it doesn't let you easily know that it's wrong. People are already plugging ChatGPT to anything from code to managing investments, it's just a matter of time until it crashes and burns. We are just waiting for the first ChatGPT version of "autonomous car rams pedestrian".

As for OpenAI, it's in their best interest for people to be scared and governments to propose regulations. It further solidifies ChatGPT as a force to be reckoned with, even if it isn't. They're trying to sell it as AGI even though it isn't anywhere near, but actions like this are helping to maintain that image.

View on HN

> I am young, but still lived through several hype phases.

My impression is that this is a tech revolution unlike any that has gone before allowing for absolutely massive concentration of power and wealth. The previous tech revolutions have shown the downsides of such concentration and I'm quite sure that the people that are currently in the driving seat of how, when and where this tech gets used and/or weaponized are exactly the people that I would not want to see in that position.

The problem is not with the fact that the pig is dancing in a crappy way. The problem is that it is dancing at all, this is the thing that makes this a different kind of tech revolution. So far tech was doing what it was told, within certain boundaries. This is the kind of tech that will end up telling you what to do, either directly or through some intermediary, it is a substantial shift in the balance of power and has the potential to lead to even stronger divisions in our populations than what social media has been able to effect as well as to cause massive loss of jobs.

Extrapolating from the last two years of this development into the future over the span of say a decade (a reasonably short span of time for me) means that 'all bets are off' and that we can no longer meaningfully predict the outcome of the moves that are being made today. For me that is a warning to be cautious to get more understanding of how we will deal with the problems that it will generate rather than to jump in head first to see where it will lead.

I'm surprised by your impression that there are people "currently in the driving seat of how, when and where this tech gets used". I think we are in the grip of a wonderful, highly creative dis-order.

The most comparable moment in my lifetime was the late 90s, where the inevitability and transformative power of the Internet (through the web) became mainstream over a couple of years. This time that transition seems to be taking weeks! And yet, it is FAR more broadly accessible than the Web was in the late 90s. In America at least, where 85% of individuals have smart phones, ChatGPT is instantly accessible to a huge portion of the population. And unlike other advances it doesn't require specialized expertise to begin extracting value.

Meanwhile, LLM owners are being compelled by competition to continue offering services for free and release research into the public domain. The engineering advances that power their platforms face a sideways challenge from research like LORA that makes them less relevant. And because the training data and methods are ubiquitous, public institutions of many kinds can potentially build their own LLMs if rents get too high. Outside the runaway-superintelligence scenario, the current accessibility of LLMs is one of the best ways this could have played out.

I'm afraid those attempting to slow research will be unwitting accomplices for people who use the opportunity to reduce competition and consolidate power over the new technology.

I agree with this 100%, while also disagreeing with the "robot god will kill us all" objections to AI which unfortunately tend to get most of the mindshare.

I think it's important to realize that these are two _completely separate_ concerns. Unfortunately, a lot of the people who get the most air time on this topic are not at all worried about authoritarianism or economic collapse compared to a hypothetical singularity.

Personally, I'm worried that if the proponents of LLMs are correct, it will directly lead to authoritarianism and economic collapse.

Who's gonna pay for GPT's in an economic collapse?

You're presupposing the collapse touches the rich.

"A quart of wheat for a denarius, and three quarts of barley for a denarius; but do not harm the oil and the wine!"

So how does this situation actually look:

Bill Gates sitting on a yacht controlling a robot army while we're all starving ?

The wealthy and powerful. Same as now.

I wish this viewpoint were more common. It's frightening to see how rapidly the "AI safety / alignment" discourse is being co-opted into arguments that AI should follow the whims of large centralized corporations. We have no idea what secret instructions OpenAI or the others are giving their "helpful" assistants. I find the notion that AI will spontaneously become a paperclip maximizer much less immediately terrifying than the clear and present danger of it being co-opted by our existing soulless paperclip maximizers, corporations, to devastating effect.

Yeah, the LLMs at the three letter agencies communicating directly with their LLM counterparts at FB and Google. And Twitter, once Musk moves on, and that site gets brought back into the fold.

The social issues need to be addressed now.

> AI should follow the whims of large centralized corporations

I'm not arguing that AI should follow the whims of large centralized corporations, but given the cost of training large models such as GPT-4, what's the alternative?

Do we need large language models as a taxpayer-funded public utility? Perhaps a non-profit foundation?

I'm not sure what the solution is here, but I am concerned that right now, large corporations may be the only ones capable of training such models.

> So far tech was doing what it was told, within certain boundaries. This is the kind of tech that will end up telling you what to do, either directly or through some intermediary

The financial crisis in 2008 was caused by investors who made risky bets based on models they built telling them that mortgage-backed securities were a safe investment. Our understanding of the world is guided by social science that seems to involve an incredible amount of p-hacking using statistics packages that make it easy to crunch through big datasets looking for something publication-worthy. It seems like tech already gives people plenty of tools to make poor decisions that hurt everybody.

Indeed it does, and we haven't got that under control yet, not by a very long distance. So I figure better take it easy before adding an even more powerful tool to the mix.

> Extrapolating from the last two years

Therein lies the error. People forget reality is finite and just because it improved now doesn't mean it will continue indefinitely.

  An exponential curve is just a sigmoid curve in disguise.

Most AI I've seen suffer from catastrophic errors in the tail end (famous example of two near identical cat pictures classified as cat and dog respectively).

What are the odds scaling continues to lead to massive AI improvements? No one is saying 100%, you seem to be arguing that they are. If you're willing to put a confidence interval on the odds with evidence we can have an actual conversation about what the best course of action is, but just talking past each other with "it might continue scaling" "no it won't" doesn't seem particularly helpful.

I think the important thing here though is the difficulty in creating an accurate confidence interval that isn't [0-100]. We are truly in uncharted territory.

> No one is saying 100%

Points at AI moratorium. I think people are arguing its inevitable.

Putting error bars on gut feeling. Interesting idea. I'd say in 10-20 years we'll not see anything revolutionary, as in AI smart enough to continue working on improving AI.

So in 10-20 years I don't expect fully self driving cars (Unsupervised, any terrain, better driver than 99.9% humans).

AI might see use in industry, but I doubt they will be unsupervised, unless we start living in Idiocracy and decide highly risque tech is better than average person.

>I'd say in 10-20 years we'll not see anything revolutionary, as in AI smart enough to continue working on improving AI

You do realize we've just at least doubled the amount of cognitive bandwidth on earth right? For every one brain, there are now two. A step-wise change in bandwidth is most definitely going to have some very interesting implications and probably much sooner than 10 years.

> You do realize we've just at least doubled the amount of cognitive bandwidth on earth right?

What you mean? Human population is now past exponential curve and entering the saturation point.

Or do you mean adding ChatGPT? Then it's not doubled. Pretty sure that's centralized.

> Or do you mean adding ChatGPT? Then it's not doubled. Pretty sure that's centralized.

Is it though? Think about it. If I have a medical or legal conversation with it that I derive value from, that's akin to a fully trained professional human who took several decades to grow into an adult and then undergo training just popping into existence in seconds during inference and then just as quickly disappearing. Just a few short months ago, the only way I could have had that conversation was to monopolize the cognitive bandwidth of a real human being who then couldn't utilize it to do other things.

So, when I say doubled, what I mean is that in theory every person who can access a sufficiently powerful LLM can effectively create cognitive bandwidth out of thin air every time they use the model to do inference.

Look what happened every time there has been a stepwise change in bandwidth throughout history. Printing press, telephone, dial up internet, broadband, fiber optic. The system becomes capable of vastly more throughput and latencies are decreased significantly. The same thing is happening here. This is a revolution.

> You do realize we've just at least doubled the amount of cognitive bandwidth on earth right?

I don't realize that. What do you mean exactly?

LLMs can perform cognitive labor. Moreover they can increasingly perform it at a level that matches and/or surpasses humans. Humans can use them to augment their own capabilities, and execute faster on more difficult tasks with a higher rate of success. In addition, cognitive labor can be automated.

When bandwidth increases, latency decreases. We're going to be able to get significantly higher throughput than we were previously able to. This is already happening.

Nope, it's now another voice, a colleague who's work you constantly have to check over because you can't trust it enough to just say, "go ahead, I know you're trust worthy".

This is already happening. ??

Sort of just proves my point, no? It's faster for me to just check its work than to do the work from scratch myself and that work isn't monopolizing the wetware of another human being. Cognitive bandwidth has increased. The system is capable of more throughput than before. In fact, how much more throughput could you get if you employ enough instances of LLMs to saturate the cognitive bandwidth of a human who's sole job is to simply verify the outputs?

If you look at the leap from GPT-3 -> GPT-4 in terms of hallucinations and capabilities etc, and you combine that with advances in cognitive architecture like reflexion and AutoGPT, it's pretty clear the trajectory is one of becoming increasingly competent and trustworthy.

The degree to which you need to check it's work depends on your use case, level of risk tolerance, and methods for verification. I think one of the reasons AI art has absolutely exploded is because there's no consequences for a generation that fails and it can be verified instantly. Compare that to doing your taxes where it's high stakes if you get it wrong, you're far less likely to rely on it. There is a landscape of usefulness with different peaks and valleys.

What are one of the professional use cases where you would just feel comfortable YOLOing some ChatGPT generated code into prod? Publishing a journal without verification etc?

You should also take note of the warnings in the GPT-4 manual, it's a much more convincing liar than GPT-3. Quite explicitly says that.

My fear is that I just get lazy and trust it all the time.

I think one of the reasons AI art has absolutely exploded is because there's no consequences for a generation that fails and it can be verified instantly.

What are you talking about exactly?

What's with the assumption that anyone needs to YOLO anything? Your coworkers don't let you YOLO your code to prod, and you don't let them YOLO their code to prod. Trust but verify, right?

My point with the AI art comment is that not every output of these models is something that needs to go to production! There's a continuum of how much something matters if it's wrong, and it depends on who is consuming the output and what it is they need to do with it, and the degree to which other stakeholders are involved.

Not at all. They're saying the current probability is high enough to warrant a cease research agreement because the risks outweigh the rewards.

Your definition of what would be revolutionary is likely the last thing humans will achieve, there is a lot of revolutionary things to happen between here and there.

I'm not sure what you are using as a definition of AI but I would say it is already being used massively in industry, and a lot of harm can be done even if it isn't autonomous.

Why is that an error? This is very new tech, if anything the rate of change is accelerating and nowhere near slowing down. That it will flatten at some point goes without saying (you'd hope!), but for the time being it looks like we're on the increasing first derivative part of the sigmoid, not on the flattening part, and from two years to a decade takes you roughly from '2 kg luggable' to 'iphone' and that was a sea change.

> Why is that an error?

First. It seems like these AI models depend on the underlying hardware acceleration doubling, which is not really the case anymore.

Second. All AI's I've seen suffer from the same works fine until it just flips the fuck out behavior (and starts hallucinating). You wouldn't tollerate a programmer that worked fine except he would ocassionally come to work high enough on bath salts to start claiming the sky is red and aliens have inflitrated the Wall of China. AI's that don't suffer from this aren't general purpose.

Third. I'm not convinced in we'll make AI whose job will be to make smarter AI which will make smarter AI argument. A smart enough AI could just rewire its reward mechanism to get reward without work (or collude with other AIs meant to monitor it to just do nothing).

I bought my GPU back in 2021. I had the same computing power as I have now, but back then it could only generate crappy pictures. AI image generation has improved massively in a few months without increasing hw requirements.

Why do you believe there will be no significant improvements beyond SOTA in the coming years / decades?

That's an incredibly strong stance...

I’d love to hear your assessment of the improvements from gpt-3.5 to gpt-4. Do you not think it is a large jump?

I think the fact that LLMs are basically fixed models (plus some RLHF and a tiny context window) means they won't likely foom in their current form, but we're trying to change that. Meanwhile, a whole bunch of ML researchers are saying "hey, our field is getting dangerous and we don't know how to keep it safe".

I'm verrrry skeptical of governmental regulation here, but I'm also not willing to dismiss the experts shouting "our work is might quickly lead to disaster!" AI is advancing very rapidly.

Yes, people were wrong about 3D TVs, but they were also wrong about the growth of solar power in the other direction, repeatedly under-estimating its rate of improvement/adoption year after year[0]. It'd consider that a much better comparison: it's not a single "3D will replace 2D", but "solar power will rapidly iterate to be better and cheaper faster than expected". Well, AI is rapidly iterating to be better and cheaper faster than expected. (To be fair, only some AI; LLMs and image generation.)

> People are already plugging ChatGPT to anything from code to managing investments, it's just a matter of time until it crashes and burns

It's not X-risk, but it's worth asking whether just ChatGPT crashes and burns, or also the market it's being plugged into.

> As for OpenAI, it's in their best interest for people to be scared and governments to propose regulations. It further solidifies ChatGPT as a force to be reckoned with, even if it isn't. They're trying to sell it as AGI even though it isn't anywhere near, but actions like this are helping to maintain that image.

No one (or very few) think ChatGPT is an AGI, and anyone who expresses high confidence about how close it is, regardless of whether they say "very close" or "very far" is over-confident. There's widespread disagreement among the people best positioned to know. We don't know, and that's bad.

[0] https://www.vox.com/2015/10/12/9510879/iea-underestimate-ren...

You don't even have to look at other "fads" like 3D televisions and VR to be skeptical of the claim that recent advancements in ML* represent a major technological revolution that will change society like the Internet or the Printing Press did.

Just look at ML's own history. A few years ago we had the "deepfake" scare. People were terrified that now videos are going to surface of politicians saying and doing things that they did not and we would have no way to tell if it was AI generated or not. But we had already been dealing with this for decades with doctored images a la Photoshop. Videos could already be edited by human beings to make it look like someone was communicating a message that they were not.

What we have now is a tool that is able to generate text-based content that is indistinguishable from that written by an 8th grader. Even with inevitable improvements that get us to 12th grade level, so what?

People will use ChatGPT to write emails and code for them. Personally I don't see the appeal since I am a very creative person and don't want to outsource that work to a human let alone software, but who cares? Currently I can tell when people that I know well are using ChatGPT to "refine" their writing to me, but the content is what matters anyway and I don't know anyone who says they don't "massage" the ChatGPT output for correctness first.

Certain people will find creative uses like generating SEO content for websites etc. That's a problem for the search engines. Basically the Internet is about to get noisier .. but it was already so noisy that the older I get the less attention I'm starting to pay to it, in general, anyway.

Then again, I am limited by my own imagination. Maybe people will come up with truly "disruptive" ways to use LLMs ... but meh. Same shit different day IMO.

* and let's be honest here, ML is a MUCH more accurate term for what we have than AI ... though it's not as good of a marketing term since ML doesn't put "The Matrix" or "The Terminator" in the minds of the lay public like "AI" does.

This is not a fad. This is like the advent of writing.

Plato opined that people would lose their facilities for memorization and the passing of oral histories would falter and die. He was correct about that. We don't have the same techniques and same facilities to memorize the Iliad verse for verse, epithet for epithet as the ancients did.

He was incorrect about that being as ruinous as he imagined, but it was as dramatic an impact on the human race as agriculture was to hunter-gatherer society.

I think we're at one of these societal delineations; before agriculture and after, before writing and after, before generative AI and after.

I know it sounds silly, but despite all the positives, the invention of writing was indeed ruinous; after all, without writing we couldn't have invented AI which is going to be truly ruinous...

(Fast-forward N years) I know it sounds silly, but despite all the positives, the invention of AI was truly ruinous; without AI we couldn't have invented X which is going to be truly ruinous...

> We don't have the same techniques and same facilities to memorize the Iliad verse for verse, epithet for epithet as the ancients did.

it's rhymes and metric, and I'm pretty sure in ancient time people got repetitions wrong too, which is why we have stories with variations everywhere in the world. What we miss is dedication, time and a facility with the language.

There's plenty of Italian people who have memorized Dante's Divine Comedy, for example.

Other fads have the disadvantage of being easily identifiable and avoidable. But AI chips away at what it means to be human. Now imagine if every comment in this thread is not a real person, but made by a LLM. Get any Truman show vibes?

> Get any Truman show vibes?

No.

It's an interesting thought experiment but it changes nothing. Not for me, anyway. Commenting on these threads is an activity that I do for my own entertainment and "mental flexing." If it turns out that I'm not talking to real people then it doesn't make much of a difference because I don't actually perceive these messages as anything other than pixels on a screen anyway.

I hope that doesn't sound "cold" but I come from a generation that was born before the Internet existed. I was about 10 years-old when we got our first modem-equipped computer and that was still early for most people (1992). Having those experiences early on with "early world-wide-web" meant that even though you knew you were talking to "real people" ... there was no real time chat, or video streaming or voice-over-ip etc. ... and so everyone's messages to each other were always pure text all of the time. And no one dared ever give their real name or identity online. You had no idea who you were talking to. So the web forced you to think of communication as just text between anonymous entities. I never got over that mental model, personally. Maybe a little bit with close friends and family on Facebook. But I'm not much of a social media user. When it comes to twitter and forums ... you all might as well be AI bots anyway. Makes no difference to me!

EDIT (addendum):

It's interesting, the more I think about your comment and my response the more I realize that it is THE INTERNET, still, that has fundamentally changed the nature of what it means to be human and to relate to others.

If you see interactions between people online as "real", meaningful human interactions, no different than relating to people in the physical / offline world, then it must be somewhat disturbing to think that you might be caught up in a "meaningful" relationship with someone who is "faking it." But that reminds me of romance scams.

For 18 years I ran a high traffic adult website and I would have men email me from time to time to share their stories about scammers luring them into relationships using doctored video, images and false identities etc. These men got completely wrapped up in the fantasy that they were being sold and it cost many of them their life savings before they finally realized it was a lie. I felt awful for them and quite sympathetic but at the same time wondered how lonely I would personally have to be to lose my skepticism of what I was being told if an attractive young woman were to express an interest in me out of nowhere.

ML will undoubtedly be used by scammers and ne'er-do-wells as a way to do their misdeeds more efficiently. But I think that the remedy is education. I miss the days when people were a bit weary of what they did, said or uploaded to the interwebs. I don't see the problem with trusting what you read online a little bit less.

> It matters a lot to me because the whole point of commenting here (or anywhere) is to talk to other humans, not just to talk to myself.

That's fair, but one of my points was that even prior to ChatGPT the ability existed for you to be "sucked into" a relationship with another under false pretenses. LLMs might make it easier to put this sort of thing on "autopilot", but if what you seek is a guarantee of talking to other humans then I don't see how, in a post-LLM-world, that can't be done. I have no doubt that online forums and communities will come up with methods to "prove" that people are "real" (though I fear this will hurt anonymity online a bit more), but also try going out and meeting people in the real world more.

It's funny, I've been an ultra tech savvy computer nerd since I was a little kid. I owe a lot to the Internet. I was working from home for 20 years before the pandemic, running my own business. Grocery delivery services have been a godsend for me, because I find grocery shopping to be one of the most stressful activities in life. But as I enter middle age I'm becoming less and less enthusiastic about tech and "online existence" in general. The number of things that I would miss if the Internet just went away entirely gets fewer and fewer every year. Working remotely and grocery delivery services are probably the only two things that I couldn't live without. Everything else ... meh. Maybe I'm just getting burned out on tech and hype trains ... but "talking to real people" is something I start to value doing offline more and more when social interaction is what I seek.

> one of my points was that even prior to ChatGPT the ability existed for you to be "sucked into" a relationship with another under false pretenses

That's true, of course, but it's still interacting with a real human being. An adverse interaction, but at least a human.

> I don't see how, in a post-LLM-world, that can't be done.

I don't see how it can be done without losing much of the value of online interactions.

> also try going out and meeting people in the real world more.

I go out and meet real people plenty, thank you very much. But we're talking about online interactions here. There should be room for people online, too.

> That's true, of course, but it's still interacting with a real human being. An adverse interaction, but at least a human.

Actually, not entirely. Some of the stories that really made me raise an eyebrow were people who claimed that they were video-chatting with "the girl." An important piece of context is that these men reached out to me because they found pictures of the woman they believed they were in a relationship with on my website. They wanted to know if the woman was employed by me or if we could verify certain details about her to try and make sense of what they had gone through.

Of course there were people driving this interaction. But a video chat? Obviously it was faked. What I think that AI advancement is going to allow these scammers to do in the future is possibly have extremely convincing voice chats, because when I probed about these video chat claims often times the scammers would have excuses about the microphone not working etc. so they were clearly just feeding pre-recorded video.

Anyway I've gotten the sense by your reply that you are under the impression that we are having some sort of debate or argument. I'm just making conversation and sharing my point of view and experiences. In my opinion I'm not sure the Internet "should" be anything in particular.

> Anyway I've gotten the sense by your reply that you are under the impression that we are having some sort of debate or argument. I'm just making conversation and sharing my point of view and experiences. In my opinion I'm not sure the Internet "should" be anything in particular.

Oh, no, I didn't think that at all. I'm sorry that I gave that impression. I'm just doing the same as you, sharing worldviews. I'm not trying to convince anyone of anything. Just having interesting conversation.

> If it turns out that I'm not talking to real people then it doesn't make much of a difference because I don't actually perceive these messages as anything other than pixels on a screen anyway.

It matters a lot to me because the whole point of commenting here (or anywhere) is to talk to other humans, not just to talk to myself.

> ..whole point of commenting here (or anywhere) is to talk to other humans..

honestly, if i can't tell the difference between an AI and a human here then why does the difference matter? If every comment on this story was AI generated except for mine I still received the same insight, enjoyment, and hit of dopamine. I don't think I really care between communicating with an AI or human if i can't tell the difference.

I understand that point of view. I simply don't share it. If I can't tell the difference between AI and a human being in my conversations, that would undermine my trust in some extremely important things. I'd withdraw from such fora entirely as a result, because there's no way for me to know if those conversations are real or just me engaging in mental masturbation.

Some would say that if you can’t possibly tell the difference, then both are equally real or unreal, in the sense that it doesn’t matter if the neural net is biological or electronic.

Right, that's why I say I understand what chasd00 is saying. I happen to disagree -- I think it matters quite a lot.

Even ignoring philosophical arguments, it matters to me on a personal level because I consider people to be important, and want to interact with them. If I'm talking to a machine that I can't tell isn't a human, then I'm accomplishing nothing of importance and am just wasting my time.

How do you know other humans aren’t machines?

You should readcwhat you just wrote

> If it turns out that I'm not talking to real people then it doesn't make much of a difference because I don't actually perceive these messages as anything other than pixels on a screen anyway.

Sorry, but no normal person can say that. Suppose I would try to bull you, it wouldn't matter? It wouldn't make a difference whether a person would have typed it or not?

This comment is mind blowing

If AI can chip away at your meaning of being human I am afraid your definition and understanding before is in need of improvement. Destroying a faulty understanding should be celebrated, not feared.

> AI chips away at what it means to be human

Can you name a broad technological revolution that didn't?

You start by talking about ML yet your point only touches on LLM's. There is plenty of harm they can provide by automating propoganda and clearly generative models will/can create things we couldn't via e.g. photoshop and most importantly, with a massively lower barrier to entry.

ML is a paradigm shift in how we program computers and we're only talking about surface level details of 1 or 2 use cases here.

E.g. generative models have already proven very effective at conceiving new nerve agents and toxins, that is not a barrier to entry we want to dramatically lower.

> Then again, I am limited by my own imagination.

On that note, it seems clear these models will be disruptive in areas where we previously needed human imagination, and wrongness of outputs can be tolerated or iterated past.

I'd like to see a transformer trained on a massive dataset of MIDI-like quantized piano performances, so it can be my virtual personal classical/jazz/whatever pianist, or play a real mechanized piano at a club or something. Directed auto-composers (music) in general are most likely being worked on.

South Park probably wasn't the first to use some ML to assist in plot development for their show.

A nice ML model to do directed building architecture, ("give the kitchen more natural light" or directed interior design ("more playful furniture") would be very useful.

I've got a pile of ideas, really, but minimal experience and no means (network, resources) to execute. Now that I think about it, ChatGPT could probably synthesize many more decent ideas for ML applications, if so directed.

>Just look at ML's own history. A few years ago we had the "deepfake" scare. People were terrified that now videos are going to surface of politicians saying and doing things that they did not and we would have no way to tell if it was AI generated or not. But we had already been dealing with this for decades with doctored images a la Photoshop. Videos could already be edited by human beings to make it look like someone was communicating a message that they were not

The difference is bandwidth. We're making it about 1000x easier and cheaper to do.

Sure, but how can anyone know that rate of progress will continue?

In some sense that doesn't even really matter. We're already at an inflexion point now where the effort to reward ratio of certain activities that used to be possible, but difficult has tipped into possible and not difficult. Once you reach that point, further progress is kind of irrelevant. The confetti has left the cannon. We're not going to go back to a world where it was as difficult as it used to be, so even if no further progress were to be made, we are still going to learn in due course what the implications of the current level of progress are.

Sam Harris has a straightforward, yet somewhat compelling argument re: your actual question here, and that is "if it's at all possible for us to improve our technology, then we are going to" and iirc he notes that it doesn't necessarily matter how fast that happens, just the fact that it's possible means we're likely to do it.

> What we have now is a tool that is able to generate text-based content that is indistinguishable from that written by an 8th grader.

To be fair, this 8th grader passed the bar exam..

So many people conveniently disregard facts like this. It's much easier to write it off as "impressive auto-complete", "a writing assistant", "simply regurgitating the training data", etc.

It's an alien intelligence that we barely understand. It has many limitations but also possesses insane capabilities that have not been fully explored.

What happens when you take gpt-5, give it 100x more context / "memory", the ability to think to itself in-between tokens, chain many of them together in such a way that they have more agent-like behavior, along with ten other enhancements we haven't thought of? No one knows...

The biggest limitation of GPT capabilities is our imagination.

Well, lawyers pass the bar exam and they're not human either (ba-dum dum!)

In all seriousness, I know of a few lawyers who would tell you that's not as impressive as non-lawyers think it is.

And the reality is, it did not technically "pass" the bar exam. That's media spin and hype. It doesn't have personhood, it's not a student, it's not being evaluated under the same strict set of conditions. It was an engineering exercise done specially crafted conditions and that makes all the difference in the world.

I'm a magician and this reminds me of ESP tests in the 70s where frauds like Uri Gellar fooled scientists (at NASA no less) into believing they had psychic powers. The scientists were fooled in large part because it's what they wanted to believe, and the conditions were favourable to the fraudster doing parlour tricks.

The most interesting part about the results are that it "passed" the essay portion, otherwise we would expect any computer software to be answer questions correctly that have a single correct answer. But who is evaluating those essays? Are they top lawyers who are giving the essays extremely close scrutiny or are they overworked university professors who have a hundred to read and grade and just want to go home to their families?

And what is the objective criteria for "passing" those essay questions? Often times the content, in a formal education setting, is not as relevant as the formatting and making sure that certain key points are touched upon. Does it need to be an essay that is full of factually-verifiable data points or is it an opinion piece? Is the point to show that you can argue a particular point of view? I mean when it comes to something open-ended, why wouldn't any LLM be able to "pass" it? It's the subjective evaluation of the person grading the essay that gets to decide on its grade. And at the end of the day it's just words that must conform to certain rules. Of course computers should be "good" at that sort of thing. The only thing that's been historically very challenging has been natural language processing. That's ChatGPT's contribution to advancing the field of ML.

So I'm not that that shaken by a chat-bot being able to bullshit it's way through the bar exam since bullshitting is the base job qualification for being a lawyer anyway :P (couldn't help bookending with another lawyer joke .. sorry).

Thanks for this. HN is great to burst my bubble a bit sometimes.

You're not alone. Although I am concerned about the people who wield these latest bits of tech.

All of these service providers jockeying for first-mover advantage in order to close on a monopoly in the space is asking for trouble.

The abilities it gives to scammers to increase the plausibility of their social engineering is going to be problematic. Scams are already quite sophisticated these days. How are we going to keep up?

Those sorts of things. ChatGPT itself is a giant, multi-layered spreadsheet and some code. "It," is not smart, alive, intelligent, or "doing" anything. Speculation of what it could be is muddying the waters as people get stressed out about what all of these charlatans are proselytizing.

have you lost a game of chess to it yet?

This could also be this era's equivalent of:

- "There is no reason anyone would want a computer in their home."

- "Television won’t be able to hold on to any market [..] People will soon get tired of staring at a plywood box every night."

"Forget all that. Judged against where AI was 20-25 years ago, when I was a student, a dog is now holding meaningful conversations in English. And people are complaining that the dog isn’t a very eloquent orator, that it often makes grammatical errors and has to start again, that it took heroic effort to train it, and that it’s unclear how much the dog really understands."

https://scottaaronson.blog/?p=6288

You should be concerned. You need to reframe this technological shift.

At some point these AI's will no longer be tools. They will be an alien intelligence that is beyond our intelligence in the way that we are beyond an amoeba.

Perhaps we should tread carefully in this regard, especially considering that the technologies that are public (GPT-4 specifically) have already displayed a multitude of emergent capabilities beyond what their creators intended or even thought possible.

My concern, even if it doesn't pan out, is the disruption as everyone tries to jump on the bandwagon.

I saw this at the end of last decade with 'low code' tools; lots of Directors/CIOs trying to make a name for themselves, via buying into a lot of snake-oil sales [0]

[0] - I left a job right as they were jumping on this bandwagon. My last day was when the folks finished 'training' and actually were trying to do something useful. They all looked horrified and the most innocent, honest engineer spoke up "I don't think this is going to be any easier."

I am maybe not so young and have lived through various hype cycles as well, plus I'm sceptical wrt AGI/sentience via LLMs as well as being a "they should call it ClosedAI" moaner to boot.

So I think I'm pretty hype-averse and a natural scoffer in this instance, but the reality is I've been stunned by the capabilities, and think we are in a transformative cultural moment.

I know "this time it's different" is part of the hype cycle meme, but you know what - this time...

Right, we're just seeing the early phase of the Gartner hype cycle play out in an unusually public and aggressive manner. At this time we're still racing up the initial slope towards the peak of inflated expectations.

https://www.gartner.com/en/research/methodologies/gartner-hy...

Eventually people will realize that LLMs are useful tools, but not magic. GPT99 isn't going to be Skynet. At that point disillusionment will set it, VC funding will dry up, and the media will move on to the next hit thing. In 10 years I expect that LLMs will be used mostly for mundane tasks like coding, primary education, customer service, copywriting, etc. And there is tremendous value in those areas! Fortunes will be made.

The thing about predicting doom is that you're wrong often--until one day you're not.

Most hype cycles involve people at the top making wild claims that fail to deliver. I don't know anyone outside the industry who ever thought 3D TVs were worth anything, and barely anyone who thought VR was worth anything. Google pitched AI making appointments, but that never made it off a stage. Hype? Only for some definition of hype.

Smartphones have changed the world, but it was primarily Apple who pushed the "post-PC" narrative, and that was to promote the iPad, one of their least successful product lines. (To be clear: it's still a HUGE business and success, but it didn't usher in the post-PC world Steve Jobs claimed it would.)

One you left out is cryptocurrency, and that's the only one I can think of where the hype came from more than just the people at the top, mostly because everyone down the chain thought they were also people at the top by virtue of buying in. Financial scams are always hype by their nature.

I'm older than some, younger than others, but in more than 30 years as a professional developer, I think this is as close to a "silver bullet" as I've ever seen. Like GUIs and IDEs, I think LLMs are tools that will make some things much easier, while making other things slightly harder, and will generally make developers as a class more productive.

There's no question that developers using a nice IDE to write high-level code on a large monitor are able to product more code more quickly than someone writing assembler on a dumb terminal, I hope. The shift from monochrome ASCII to GUIs helps, the shift from a text editor to an auto-completing, stub-writing IDE helps, and similarly, I think having an LLM offer up a first pass approximation for any given problem definition helps.

Concerned? I'm not concerned, I'm excited! This isn't marketing hype coming from someone on a stage, it's grassroots hype coming from nobodies like me who are finding it actually-helpful as a tool.

> I think having an LLM offer up a first pass approximation for any given problem definition helps.

This is, strictly scoped, true. But the future is people with capital deciding that the computer output is Good Enough because it costs 0.1% as much for maybe 60-70% the value. People who write code are probably not sufficiently upstack to escape the oncoming Enshittification of Everything that this portends, either in terms of output or in terms of economic precarity.

> There's no question that developers using a nice IDE to write high-level code on a large monitor are able to product more code more quickly than someone writing assembler on a dumb terminal, I hope.

This is true. It's also true that the code they produce is of lower quality. In practice, for the most part, this doesn't matter because the industry has decided that it's more economical to make up for poor quality code with more performant hardware.

I feel the same way. It's cool and shiny, but I am just not impressed with a form filled in or a pull request description message, which I like writing anyway. As far as image manipulation, I like making my own images and find pleasure in the actual process. I also can't find any gains in cost effectiveness, because an artist will get paid either way, if they make an image by hand or generate one.

The hype is also a little sickening. If we take a look at nuclear power as an analog modern tech development, we still don't know how to use it efficiently, or even safely, but it hasn't ended anything. It's just too much hype and apocalyptic nonsense from people.

The problem friend isn't that it will actually replace people, but that it will be used to justify firings and economic upheaval for worst results and productivity that only exists in excel. That is my concern, none of this "it will replace all humans" bullshit. It will absolutely be used to thin out labor just as automation already is, and the world is already worse because of it. Everyone but managers are laughing their way all the way to the bank.

> “remember that? yeah it was 5 years already”

I get the impression many HN commenters haven’t even been adults for 5 years so no, they really don’t remember it :) for example articles get posted here and upvoted with the author boasting about their insights from being in the software industry for 3 years!

> author boasting about their insights from being in the software industry for 3 years!

Nothing quite like the not-fully-earned confidence of one's 20s. :)

I mostly agree. Weirdly, coding is actually one of the better things to use it for because it’s trivial to get immediate feedback on how good it was? Does it compile? Was that package it loads hallucinated? Does it pass any tests? I’m sure people could do dumb things, but you inherently have the tools to check if it’s dumb. Other uses aren’t like this. Asking gpt to design a bridge or a chemical plant is a long way off because the result of a mistake are harder to check and more costly. You still need experts for anything that’s not easy to cross check against reality.

>I remember when PC was to die soon and to be replaced by smartphones

Hmm, That's kinda happening slowly I think, but probably more by tablets. Probably most people here on HN have a laptop/desktop at home, but I wonder how many households see it as a necessity anymore.

The issues you mention with Stable Diffusion are teething issues. Here's an example of a current technique to get temporally stable animations with Stable Diffusion - https://www.youtube.com/watch?v=zDvpJIp0rl0

Corridor digital show off a similar technique here - https://www.youtube.com/watch?v=_9LX9HSQkWo

GPT4 with its new plugin architecture is rapidly becoming more capable. I think you're absolutely correct in the tendency of all LLMs to hallucinate is a major issue that we don't know how to correct even in principle. But there are techniques to moderate its impact on responses - triangulating using multiple models, citing sources etc.

You could build an assistant like say this, today - https://twitter.com/rowancheung/status/1641273318463664130 to make your appointments and answer the phone today, using technology like Uberduck for voice generation, GPT for conversation analysis and DeepSpeech, Kaldi or Whisper for speech to text.

I agree with the criticism that we're not likely to develop AGI this way. We're not going to have a wilful agent with it's own goals trying to escape the bonds of its server to destroy humanity any time soon. But that's not necessary for this technology to be incredibly dangerous. A mindless philosophical zombie with the capacity to convince people it isn't could lead to an extinction event too.

Given that we don't understand precisely how these models process information, how to imbue them with goals or motivation (or measure those), how to accurately assess how well they've been trained - or any of a dozen other major alignment issues - there's a strong argument that they are in fact incredibly dangerous.

Literally the only reason GPT4 can't tell you how to build a bomb from common household ingredients, or generate a list of extractive industry CEOs to assassinate is it's been censored to prevent it answering those questions. Open source versions of these models will soon run locally with no such limitations. Precisely the way Stable Diffusion can be used to create nudes, violence etc. But with infinitely worse potential consequences. Say, detailing precisely what to purchase and assemble to cultivate and disseminate a bio or chemical weapon.

Robert Miles has some excellent videos detailing some of the dangerous of (non AGI) AI - https://www.youtube.com/@RobertMilesAI

Reply on news.ycombinator.com