maxdoop

And what is “knowing”?

Everyone repeats the retort you gave, yet I’ve yet to see a clear definition of “knowing”.

I struggle with this take as it’s not different from humans.

Am I intelligent at age 3? Then I learn language. Am I intelligent? Then I progress through education. I become more and more intelligent as I am exposed to more data.

I don’t know about computer science, so I go expose myself to computer science literature. I take courses, and see examples. What emerges is a “more intelligent” version of myself that.

What is intelligence, even?

Continue asking it to provide details and it can. Or , prior to asking it about performance, ask it to respond with as much detail as it can and have it include details you specially want to see.

Comparing GPT-4 to a monkey with a typewriter , and claiming the absolute of “there’s no getting better from this” when we’ve literally seen dramatic progress in just months?

I think you’re missing out on some of the utility this stuff can actually provide .

While I’m aware of my own mediocre programming skills and by no means claim to be an expert, I’ve been a software engineer for 10+ years and have done pretty well for myself. I’d hope that doesn’t qualify for “quasi-technical”, but maybe you are lucky enough to work on novel problems that require advanced problem solving every day.

Regardless, I’d argue that GPT-4 is actually far better at programming assistance, understanding concepts (it’s phenomenal at explaining things when prompted within a context), writing in general, and kick-starting creative pursuits than it is being a Google-replacement (for now, at least).

Have you seen or used GPT-4? What has your experience been? What has it failed at, or rather, what would you wish to see in such a system that might make you to, “huh, ok — that is pretty cool.”

Genuine question as I definitely need more to balance my own thinking —

What has your experience with GPT been? For me, GPT-3 was not really useful as a software dev.

But GPT-4 is miles ahead of that. It’s helped me write code maybe 4-8x faster than usual, and has even allowed me to debug existing issues far, far quicker and more accurately than I’d ever be able to on my own.

Part of the gap very well might be my own mediocrity with development . I wouldn’t argue that folks with far superior skills and novel challenges day-to-day might be unimpressed.

But as an average dev writing pretty boring code (REST APIs and system integration mostly), I’ve been blown away by GPT-4. I am pretty well compensated and have been in the field for 10 years, too; but I am aware of my own shortcomings.

What exactly is a novel idea? Is there any true “novelty” that just suddenly arises in a human brain without having prerequisite experience / knowledge?

But still, what does it mean to know something?

And your point about humans lying about knowledge only to be found inexperienced is quite the opposite of an LLM (albeit there is the hallucination problem, but GPT-4 is a massive improvement there):

These models do have “experience” aka their training data. And I would argue with most every one of your examples of things that GPT doesn’t know.

You can ask it about performance implications, side effects, costs. It’s quite good at all that right now even! Imagine the future just a few years out.

Apologies, I’m maybe low on coffee this morning.

Do you mean things like tone, facial expression, the general “energy” around a conversation, etc?

But what does it mean to “know” something?

And when something gives the increasingly-accurate illusion of knowing, I fail to see how it matters (with regard to impact on society and overall utility).

I’m not saying GPT-4 is this amazingly accurate, near perfect model. But if you extend the timeline a bit, it’ll be able to become more and more accurate across a broader range of domains.

Furthermore, how can we prove a human “knows” something?

I completely agree with you.

Note that I’m not making those claims about sentience and similarity. What I am pushing back on is the confidence with which proclaim humans are “so different”, when I’ve yet to see actual proof of this dissimilarity.

I think it’s (understandably) an emotional response from folks to dismiss ideas around LLM progress because it FEELS like we are thereby lessening what it means to be human. But I’m not at all trying to make that claim; I’m just trying to see how we can explore the question.

I think there’s a huge misunderstanding of what “original ideas” are and what constitutes novelty.

Mark Twain quote on originality:

“ There is no such thing as a new idea. It is impossible. We simply take a lot of old ideas and put them into a sort of mental kaleidoscope. We give them a turn and they make new and curious combinations. We keep on turning and making new combinations indefinitely; but they are the same old pieces of colored glass that have been in use through all the ages.”

I am not sure how humans “come up with new ideas” themselves. It does seem to be that creativity is simply combining information in new ways.

Then this implies that you’d maybe think differently if LLMs could have different inputs, correct?

Which they are currently doing. GPT-4 can take visual input.

I totally agree that humans are far more complex than that, but just extend your timeline further and you’ll start to see how the gap in complexity / input variety will narrow.

Humans require training data as well.

And what do you think of the Mark Twain quote:

“ There is no such thing as a new idea. It is impossible. We simply take a lot of old ideas and put them into a sort of mental kaleidoscope. We give them a turn and they make new and curious combinations. We keep on turning and making new combinations indefinitely; but they are the same old pieces of colored glass that have been in use through all the ages.”

I’d argue ChatGPT can indeed be creative, as it can combine ideas in new ways.

I used “all” instead of “most”, which is indeed inaccurate. A better response would have been “a collection of humanity’s creativity”.

However, you fail to recognize that OpenAI also created Whisper, which is a quote capable speech-to-text transcriber; and this tool easily converts audio and video (those verbal bits you mentioned) into text.

So the pool of creativity which OpenAI can train its models on is far larger than just original text; further, they demoed image modality a couple weeks back which would allow for VISUAL creative works to be parsed as well.

The gap between AI “acceptance / exploration” and “AI dismissal” continues to widen.

Right now, the top post on HN is about how ChatGPT is “a glorified text prediction program.”

Right under that post is this post.

“It’s a glorified word predictor” is becoming increasingly maddening to read.

Do tell— how can you prove humans are any different?

The most common “proofs” I’ve seen:

“Humans are more complex”. Ok, so you’re implying we add more complexity (maybe more modalities?); if more complexity is added, will you continue to say “LLMs are just word predictors”?

“Humans are actually reasoning. LLMs are not.” Again, how would you measure such a thing?

“LLMs are confidently wrong .” How is this relevant ? And are humans not confidently wrong as well?

“LLMs are good at single functions, but they can’t understand a system.” This is simply a matter of increasing the context limit, is it not? And was there not a leaked OpenAI document showing a future offering of 64k tokens?

All that aside, I’m forever amazed how a seemingly forward-looking group of people is continually dismissive of a tool that came out LITERALLY 4 MONTHS AGO, with its latest iteration less than TWO WEEKS ago. For people familiar with stuff like Moore’s law, it’s absolutely wild to see how people act like LLM progress is forever tied to its current , apparently static, state.

Can you explain your proof of that?

For a supposed smart group of individuals who are good at leaning into the latest innovations, I feel most HN comments as of late are completely , ignorantly dismissive of LLMs simply because LLMs aren’t “amazingly perfect” in their current state .

Yet… we are are talking about a tool that came out literally FOUR MONTHS AGO. And the huge advancement on that came out TWO WEEKS AGO.

Yet everyone here continues to proclaim, “it’s not even that good , honestly.” As though no progress will ever be made from this current moment in time.

I feel like I’m taking crazy pills.

How do you know it won’t? I don’t mean to sound like a doomer , but I never see evidence for these claims around “AI not ever being creative like a human!”

These AI’s are literally trained on all of human creativity… i struggle to see how there’s any indication AI would fail to be as creative as a human in even the near-ish future.

HN is ridiculous at times.

“Meh, most amazing technology since the internet? Lame. Tools that came out 8 hours ago still can’t do everything I imagine .”

Is OpenAI just extremely prepared for their releases, or are they using their own tech to be extremely efficient? I'm imagining what their own programmers do each day, given direct access to the current most powerful models.

Our of curiosity, what is GPT-4 getting wrong so often? It’s prettily wild to my own , admittedly easily impressed, mind.

The author assumes that asking GPT to “take longer and reconsider” is something that an LLM actually does.

As in, temporally.

The author claims that “even after taking longer to reconsider” an answer, ChatGPT was still wrong.

The author appears to misunderstand how LLMs work.

I agree 100%.

I don't understand why we can't look at the potential, or even current, capabilities of these LLMs and have a real conversation about how it might impact things.

Yet so many folks here just confidently dismiss it.

"It doesn't even think!" -- OK, define thinking?

"It doesn't create novel ideas!" OK -- what do most devs do every day?

"It is wrong sometimes!" OK -- is it wrong more or less often than an average dev?

I find it equally worrying how so many techies continue to dismiss LLMs as “bullshit generators “.

I struggle to see how even the current GPT-4 is any worse than your average human.

How this is dismissed because it’s not 100% perfect (might at add, “yet”) is beyond me.

As with every thread on this:

Define “knowing”, and then say why it even matters if an LLM gives a great illusion of knowing.

I have fun on these HN chats responding to comments like yours .

It’s just fancy auto complete to you? You honestly can’t see the capability it has and extend it the future?

What’s that saying about “it’s hard to get someone to understand something when their salary depends on their not understanding it”.

It seems so many of you guys are extremely lucky to be working on novel problems require elegant new solutions each day. That must be the case, otherwise I don’t understand these comments shrugging off GPT-n capabilities around coding.

“Psh, it’s just doing stuff it saw from its training data. It’s not thinking. It can’t make anything new.”

In my 11 years as a professional software engineer (that is, being paid by companies to write software), I don’t think I’ve once had come up with a truly original solution to any problem.

It’s CRUD; or it’s an API mapping some input data to a desired output; or it’s configuring some infra and then integrating different systems. It’s debugging given some exception message within a given context; or it’s taking some flow diagram and converting it to working code.

These are all things I do most days (and get paid quite well to do it).

And GPT-4 is able to do that all quite well. Even likely the flow diagrams, given it’s multi-modal abilities (sure, the image analysis might be subpar right now but what about in a few years?)

I’m not acutely worried by any means, as much of the output from the current LLMs is dependent on the quality of prompts you give it. And my prompts really only work well because I have deeper knowledge of what I need, what language to use, and how to describe my problem.

But good god the scoffing (maybe it’s hopium?) is getting ridiculous.

I don’t think you’ve tried GPT-4.

If you have, I don’t think you are like majority of devs (maybe not on HN, but in real life).

You sound lucky to have true, novel problems to solve each day. I’m with many here commenting that this is quite powerful stuff, especially when my day-to-day is writing simple CRUD apps, or transforming data from one format to another within an API, or configuring some new bit of infra or CI/CD.

I’d love to be challenged in some new way, and have access to truly fascinating problems that require novel solutions. But most enterprises aren’t really like that nor do that need that from majority of engineers.

So really, human input to get the ball rolling. AI takes it all the from there.

Wild.