Hacker News Reader

Comment by nwienert | Hacker News Reader

I mean Google revolutionized search. Apple revolutionized personal computing.

OpenAI revolutionized… rewriting things with slightly different wording?

I’ve seen so many breathless people posting “this would have taken me so long to search” and then I type 3 keywords from their massive prompt they crafted and find it instantly on Google. We’re talking 1000x or more faster. I feel like the same is happening in your comment. How often have I thought “damn I wish I wrote this blog post ever so slightly differently” in my life? Maybe a handful of times? And yes I’m including all generalizations of that question.

But certainly fake girlfriends and summarization will be mid size fields. Image generation has some mid size potential. But these will be spread between many companies.

I really think it has uses no doubt, but is it a revolution? Where? It’s not creative in the valuable sense - media, art, fashion, etc all will adopt it marginally but ultimately it will actually only serve to further the desire for genuine human experience, and cohesive creativity that we see it really falls flat at. It saves some marginal time perhaps if you’re ok sounding like a robot.

Taking into account the downsides it looks like a hype bubble right now to me, and a draw in the long run. There’s just a whole lot of tech people trying to cash in on the hype.

View on HN

You can program GPT in English.

Let me repeat that: You can program GPT in English. ENGLISH!

You're complaining about the first nuclear test bomb being impractical and uninteresting. How will this change the world? That huge monstrosity had to be affixed to the top of a test gantry and took years of effort by a veritable army of the best and brightest to make! No way it could change war, or geopolitics, or anything. No way..

This is the day after Trinity. The bomb has gone off. A lot of physicists are very excited, some are terrified, and the military is salivating. The politicians are confused and scared, and the general public doesn't even know yet.

That doesn't mean the world hasn't changed, forever.

> You can program GPT in English. > Let me repeat that: You can program GPT in English. ENGLISH!

What does that mean to "program GPT"? Do you mean program (software) USING GPT?

I thought we already had COBOL, which is pretty much like English so business people can use it. Same for SQL.

And don't we already have lots of low-code or no-code tools? Why do we need to program with ChatGPT if we already are beyond programming?

Not the person you replied to, but I see it the same way. GPT is an English (and other natural language) compiler.

Not in the sense that you get a computer program out (though you can), but in the sense that it can automate anything without even needing a programming language, compiler, and domain specific UX.

Low code and no-code tools still require thinking like a programmer. You define what you need to do, then implement, then get results. GPT often lets you go directly from spec to results.

If the goal is programming, GPT is nothing special. If the goal is quickly reasoning over very abstract instructions, it’s amazing.

The trick is seeing the new use cases. It really does come back to the GUI revolution: if you want to list files in a directory, the CLI is just as good, maybe better. But GUI makes photoshop possible.

GPT makes it possible to say “summarize the status emails I sent over the past year, with one section per quarter and three bullet points per section”. And the magic is that is the programming.

> What does that mean to "program GPT"? Do you mean program (software) USING GPT?

A sibling comment already explained the second part of the question, but there is something I find more exciting. You can program GPT, as in you can tell it to change its behavior. The innumerable "jail break" prompts are just programs written in English which modify GPT itself. Like macros in lisp I guess. The first time I truly saw this potential was when someone showed me you don't actually have to change the temperature of chatGPT in code, you can just tell it to give low and high temperature answers in the prompt[1]. That's programming the model itself in english.

[1] https://news.ycombinator.com/item?id=34801574

> You can program GPT in English.

> Let me repeat that: You can program GPT in English. ENGLISH!

How?

Let me repeat that: How?

I had a little script that from time to time parses a list of jobs from a specific board, extracts some categories, inserts them into an SQLite and have a frontend that displays them to me in a way I want.

The board has since changed some things which would mean maybe 2 hours of commitment from me to update the script.

How do I program GPT in English. ENGLISH! To do that for me? What are the steps involved? I've been using ChatGPT and GPT-4 for awhile and I can't imagine what the steps are to make this happen without a lot of back and forth. I can't imagine how to program the infrastructure. I can't imagine how the API endpoint is more than a fancy autocomplete. I need help understanding what it means that I can program it in ENGLISH! (I can also program it in my country's language for what it's worth).

> That doesn't mean the world hasn't changed, forever.

I sort of agree with this.

Honestly don't think it will be long before gpt can read this comment, then politely ask you for the urls of the job board and your git repo and 2 seconds later you will have a pull request to review

You might find this interesting - https://github.com/Torantulino/Auto-GPT

> Auto-GPT is an experimental open-source application showcasing the capabilities of the GPT-4 language model. This program, driven by GPT-4, autonomously develops and manages businesses to increase net worth. As one of the first examples of GPT-4 running fully autonomously, Auto-GPT pushes the boundaries of what is possible with AI.

Perhaps, but I'm talking about right now. I would love to be able to do this as I have 1000 ideas and no time to try them out.

I have some scrapers built in Scrapy, and from my experimentation with GPT4, I bet you could paste in your scraper code, the html source from the website in question (at least the relevant part), and tell GPT4 to update your scraper and you'd get something that's at least 95% correct within 30 seconds.

> make this happen without a lot of back and forth

Perhaps this is the part you're missing. When I've watched people program with ChatGPT it _is_ a lot of back and forth because an enormous amount of context is able to be stored and back referenced. I.e. one wouldn't say "make me a Flappy Bird clone for iOS", they'd start with:

"Give me the code for a starter SpriteKit project". Then

"Now draw a sprite from bird.png and place it in the center of the screen".

"Now make it so the bird sprite will fall as if it's affected by gravity"

I won't bore anyone with how might one go from that all the way to a simple game, but I'm sure you see the idea. There are obviously _huge_ limitations to this approach and professionals will get hit them fast, but the proof is in the pudding: people who can barely code are producing real software through this approach. It's happening.

> Perhaps this is the part you're missing. When I've watched people program with ChatGPT it _is_ a lot of back and forth because an enormous amount of context is able to be stored and back referenced.

I've tried to build a lot of fun stuff with it so far. Haven't been able to properly 'program it in English' for anything non-trivial. Back and forth ended up in loops of not what I wanted. I'm just utterly confused at the difference in experiences I've had with it vs. what some people are preaching.

> There are obviously _huge_ limitations to this approach and professionals will get hit them fast, but the proof is in the pudding: people who can barely code are producing real software through this approach. It's happening.

I've had 4 product people I know try to create products using ChatGPT. All 4 of them basically got stuck on the first steps of whatever they were trying to do. "Where do I have to put this code?", "How do I put it online?", "How do I store user data?", "Where do I get a database from?". Basic questions to any professional, but to them it was impossible to overcome the obstacles from code to deployment.

I don't doubt that it's happening and it will become better in the future; I'm just having a hard time trying to grasp where some people are coming from when my experience as a professional, using it, has been mixed.

i've observed this schism between people who can get LLMs to produce useful output and people who are baffled, I think it's a mixture of two things:

expectations: using to the LLM to break problems into steps, suggest alternatives, using the LLM to help them think through the problem. I think this is the people using it to write emails - myself included, having a loop to dial in the letter allows me to write the letter without the activation energy needed to stare at a blank page

empathy: people who've spent enough time interacting with an LLM get to know how to boss it around. I think some people are able to put themselves in the LLMs shoes and imagine how to steer the attention into a particular semantic subspace where the model has enough context to say something useful.

GPT4 writes boilerplate python and javascript servers for me in one shot because I ask for precisely what I want and tell it what tools to use - I think because I have dialed in my expectation for what it's capable of and I learned how to ask in precise language, I get to be productive with GPT4's code output. Here's a transcript: https://poe.com/lookaroundyou/1512927999932108

Interesting point about empathy. Sorry I'm abusing the comment system to get back to your comment in the future.

The same way people used to write code in the early days: trial and error, and a boatload of cursing!

Let me give you a simple example. I had to deal with a desynced subtitle file recently. I described the exact nature of the desync (in terms like "at point X1 the offset is Y1, and at X2 it is Y2") to GPT-4 and asked it to write me a Python script to fix this. It did require a couple tweaks to run, but when it did, it "just worked".

"Automatic Language-Agnostic Subtitle Synchronization"

Link: https://github.com/kaegi/alass

It's basically magic.

> Let me repeat that: You can program GPT in English. ENGLISH!

For problems that are fine being defined ambiguously. Try to program a database in English, let's see where it goes.

You laugh, but this is why SQL reads kinda-sorta like English. People have tried, and failed.

Meanwhile, if you give Chat GPT your database schema and ask it to write a SQL query for a report, it can do that for you.

If anyone wants to see the output of GPT4 when asked to define the tables and some sample queries for a hackernews clone in sqlite:

https://poe.com/lookaroundyou/1512927999932134

This is a very simple case that doesn’t reflect the complexity of a real project. Like so many attempts before to produce code, using a little effort, it degrades when the complexity level increases even slightly. Once there are more tables, ones that have names which cannot be easily translated from English, it breaks down quickly. These types of tools work ok for brand new projects, but work on existing projects will prove to be exponentially harder or more difficult than it is worth.

Nonetheless, it could prove useful for looking up algorithms, patterns, and generating boilerplate code. However, an important issue is will it generate similar code if queried at a later time? Not likely, which will make it less useful or result in an inconsistent codebase. Maybe you can request a version of the code generation? In-house code generators will generate consistent code, so it will be interesting to see how it is used in real projects.

Here's a more extreme example, using SQL as an API to give model access to game world state to reason about it.

https://gist.github.com/int19h/4f5b98bcb9fab124d308efc19e530...

Note that in this case it isn't even asked to write specific queries for specific tasks - it's just given one high-level task and the schema to work with.

You're right, though, that the effectiveness of this approach depends very much on schema design and things like descriptive table/column names etc (and even then sometimes you have to make it more explicit than a human would need). You really need to design the schema around the capabilities of the model for best results, which makes it that much harder to integrate with legacy stuff. Similarly, not all representations of data work equivalently well - originally, I gave the model direct access to the typed object graph, and it handles that much worse than SQL. So if your legacy software has a data model that is not easy to map to relational, too bad.

On the other hand, GPT-4 is already vastly better at this kind of task than GPT-3.5, so I think we can't assume that this will remain a limitation with larger models.

> You really need to design the schema around the capabilities of the model for best results, which makes it that much harder to integrate with legacy stuff.

This may end up being a feature of some high level frameworks … “compatible with ChatGPT” or “designed to work with xxx LLM”.

It will be very amusing if, eventually, our jobs as software engineers will be crafting bespoke AIs to maximize the efficiency of their use by an LLM.

The Ford Nucleon was a 1957 concept car that featured a compact nuclear reactor. Look at how well that prediction aged. It's apt that you mention the Trinity test, since 1950s inflated expectations of the applicability of nuclear everything are exactly where we are now.

Perhaps I could interest you in some Radium Water? It's new and trendy and good for your health.

Wikipedia has a fascinating article:

https://en.wikipedia.org/wiki/Radioactive_quackery

You can't program GTP in anything if you can't program.

If your prompt is garbage then the output will be garbage and if you don't know how to program you won't even realize the output was garbage.

It's not the language part of programming language that is hard. It's the programming part because it means you have to have a good understanding of what you want. Just like a human programmer won't read your mind an AI programmer won't read your mind either.

But I can already foresee bosses dismissing employees that raise issues (performance, maintainability, scalability, etc., etc.) by saying "Look, the AI can do it. So if it can do it you can do it too.". I foresee this because I have already seen it.

> You can't program GTP in anything if you can't program.

That's why it makes this so interesting - this type of automation impacts our jobs directly. Of course, I'm not sure who would use this in a corporate codebase without legal concerns.

Work in a financial place, we have a contract with open AI so they don't hoover our user input. Normal URL is blocked.

> Let me repeat that: You can program GPT in English. ENGLISH!

The very existence of "prompt engineering", numerous discussions about how to prompt ChatGPT in order to get the result you want, etc. imply that while it may be in English, it still requires similar care and attention to do properly as a programming language does.

Which makes me wonder what the advantage of using English is. A formal language seems like it would be more productive and accurate.

For one, GPT-4 requires far less prompt engineering and generally interprets intent better.

The advantage of using English (natural language that is), the humans around you tend to speak it. I don't naturally speak powershell. Instead I want a script that searches for particular filenames, under a particular size, between a particular date in a directory path I specify. I told GPT I wanted that and in a few seconds it dumped out what I needed. It wrote the script in a formal language, which is then interpreted by the machine in an even more formal manner. Let the code deal with accuracy, and lets let language models argue back and forth with humans on intent.

> The advantage of using English (natural language that is), the humans around you tend to speak it.

This is true, but of limited utility. English is so bad at this sort of thing that even native-speaking humans are constantly misunderstanding each other. Especially when it comes to describing things and giving instructions.

That's why we have more formal languages (even ignoring programming languages) for when we need to speak with precision.

That's the other nice thing about ChatGPT - if you say it something and it misunderstands, you can correct it by saying, "no, actually, what I meant is ...". Which, again, is how people generally do that kind of thing outside of programming. The advantage is that you're still remaining on a much higher level of abstraction.

As far as formal languages... GPT doesn't know Lojban well, presumably because of its very small presence in the training data (and dearth of material in general). But it would be interesting to see how training on that specifically would turn out.

Every text interface will eventually include a half baked implementation of a programming language.

> Which, again, is how people generally do that kind of thing outside of programming.

Yes, and with people, that's insufficient if you really need confidence of understanding.

There's a reason that lawyers speak legalese, doctors speak medicalese, etc. These are highly structured languages to minimize confusion.

Even in less technical interactions, when you need to be sure that you understand what someone else is saying, you are taught to rephrase what they said and tell it back to them for confirmation. And there's still a large margin of error even then.

This is why, whenever I have an important conversation at work, I always send an email to the person telling them what I understood from our exchange. In part to check if I understood correctly, but also so that I have a record of the exchange to cover my ass if things go sideways because we didn't understand each other, but thought we did.

> and then I type 3 keywords from their massive prompt they crafted and find it instantly on Google.

Seems I and you have different Googles and you still have the one I had pre 2010.

For over a decade now, Google has been including things I never asked about to the point where it would sometimes be easier to find it using Marginalia.

Some say it is just because internet has changed and there is less ham and more spam, but the last few months I have been using Kagi and it proves it is possible to create a better search experience.

And, if Google works for you, fine. Maybe you search other topics, use other keywords or are in another bucket wrt experiments, but from my perspective Google is now the same as its predecessors.

I actually agree Google has gone downhill. Yet for the 8 or so examples I’ve tested where I saw hyped GPT results, every single one google answers, usually in the top snippet, always in the first result.

For politics shopping and some other topics it can be terrible, but I don’t think GPT is good at those either.

I’m actually happy to be proven wrong here. If you have some examples let’s test it out. If it’s a true step function improvement I’d expect it to be easy to source examples.

Haven't used Google in a while but let me try.

I think this is a classic case of us overestimating the immediate impact and underestimating the long term impact.

Right now, they are definitely useful time savers, but they need a lot of handholding. Eventually, someone will figure out how to get hundreds of LLMs supervising teams of millions of LLMs to do some really wild stuff that is currently completely impossible.

You could spin up a giant staff the way we do servers now. There has to be a world changing application of that.

>Eventually, someone will figure out how to get hundreds of LLMs supervising teams of millions of LLMs to do some really wild stuff that is currently completely impossible.

This is an intuitive direction. In fact, it’s so intuitive that it’s a little bit odd that nobody seems to have made proper progress with LLM swarm computation.

In particular, it's odd that the greatest software developer in the world (ChatGPT) hasn't made progress with LLM swarm computation.

How is "LLM swarm computation" different that single bigger LLM?

It's easy to distribute across many computers which communicate with high latency

LLMs are already running distributed on swarms of computers. A swarm of swarms is just a bigger swarm.

So again, what is the actual difference you are imagining?

Or is it just that distributed X is fashionable?

Significantly higher latency than you have within a single datacenter. Think "my GPU working with your GPU".

There are already LLMs hosted across the internet (Folding@Home style) instead of in a single data center.

Just because the swarm infrastructure hosting an LLM has higher latency across certain paths does not make it a swarm of LLMs.

> There are already LLMs hosted across the internet (Folding@Home style)

Interesting, I haven't heard of that. Can you name examples?

I read about Petals (1) some time ago here on HN. There are surely others too, but I don't remember the names.

1. https://github.com/bigscience-workshop/petals

Rather large parts of your brain are more generalized, but in particular places we have more specialized areas. Now, you looking at it would consider it all the same brain most likely, but if you're looking at it in systems thinking view, it's a small separate brain with a slightly different task than the rest of the brain.

If 80% of the processors in a cluster are running 'general LLM' and 20% are running 'math LLM' are they the same cluster? Could you host the cluster in a different data center? What if you want to test different math LLM modules out with the general intelligence?

I think I would consider them split when the different modules are interchangeable so there is de facto an interface.

In the case of the brain, while certain functional regions are highly specialized I would not consider them "a small separate brain". Functional regions are not sub-organs.

The same reason why you don't let Mr Musk do all the work. He can't.

One LLM is limited, one obvious limitation is its context window. Using a swarm of LLMs that each do a little task can alleviate that.

We do it too and it's called delegation.

Edit: BTW, "swarm" is meaningless with LLMs. It can be the same instance, but prompted differently each time.

> The same reason why you don't let Mr Musk do all the work. He can't.

Better to limit his incompetence to one position.

I beg to differ. Imagine him taking down Twitter, Facebook, Instagram, and all the others in one fell swoop!

Context window is a limitation, but have we actually hit the ceiling wrt scaling that? For GPT, you need O(N^2) VRAM to handle larger context sizes, but that is a "I need more hardware" problem ultimately; as I understand, the reason why they don't go higher is because of economic viability of it, not because it couldn't be done in principle. And there are many interesting hardware developments in the pipeline now that the engineers know exactly what kind of compute they can narrowly optimize for.

So, perhaps, there aren't swarms yet just because there are easier ways to scale for now?

I am sure the context window can go up, maybe into the MB range. But I still see delegation as a necessary part of the solution.

For the same reason one genius human does not suddenly need less support staff, they actually need more.

Edit: and why it isn’t here yet is because it’s new and hard.

I've read about people doing it, I haven't read about people achieving anything particularly interesting with it.

It's early days. There will be a GPT 5 I'm sure, maybe that one will be better at teamwork.

This sounds like that old economics joke that says it's impossible to find $20 on the ground, because if it had been there, someone would have already picked it up.

I”m not in ML, so excuse this maybe naive question:

> get hundreds of LLMs supervising teams of millions of LLMs

What does this mean or what can you do with this setup… do you mean running LLMs in parallel?

Yes, that's called 'ensembling'. There is a lot of work being done on this kind of solution. One way in which it could work is that you can use multiple models that have been fine tuned for various problems and then use the answer that returns the highest confidence.

You can also have adverserial generation where models given different expertises and attitude can go back and forth criticizing each others work

Sounds like the 'dead internet' is just around the corner!

Ask the LLM to perform a complex task by splitting it into sub tasks to be performed by other LLM instances, then integrate the results...

Is this something like langchain is working towards?

AutoGPT, BabyAGI

We call that “companies”. We just need to apply what we learned in business school to a different set of workers, slightly deficient workers.

> We just need to apply what we learned in business school

Please don't. You've already ruined enough industries. Let the MBAs do finance and Wall Street and leave them out of the chain of command in organizations that make things.

Take shelter under my protective wings, O, sweet summer children.

Every time you go to the store and find that the store is still in business and there is food on the shelf, it is because someone went to business school and knows how to optimize demand estimation, pricing, and logistics.

Yes, some MBAs fuck things up. Just like some CS grads fuck things up. But advocating against the study of business is just as naive as advocating against the study of computer science just because there are some bad CS grads.

> Every time you go to the store and find that the store is still in business and there is food on the shelf, it is because someone went to business school

Are you contending that business were not successful before Wharton started pumping out MBAs?

> But advocating against the study of business is just as naive as advocating against the study of computer science

I didn't say 'don't study business', I said 'stick to finance'. MBAs tend to end up destroying innovation and productivity for short term growth and stats.

Jack Welch showed what a successfully motivated 'business oriented' leader can do to an innovative and productive legacy organization when given complete control over it. The MBAs happen to just do it on a smaller scale.

Stores existed before the MBA, but MBAs could be why food prices are up 30% since last year.

> advocating against the study of business is just as naive as advocating against the study of computer science just because there are some bad CS grads.

Criticizing garbage MBA programs is not criticizing the study of business. Business schools don't study business. They're a place where people make a lot of money selling theories about business that are useless at best and it many places, quite harmful. Learning about business by going to business school is like learning to kiss by reading books about kissing.

That is a great analogy.

I would say that just as every person is unique so is every company unique. And just as there is plenty of pseudoscience plaguing psychology so are MBAs full of pseudoscience. Two fields that are far too obsessed with generalising their advice. Which is not to say that there aren't any useful ideas in these fields. But the vitriolic reaction above is warranted.

It’s a hype bubble for hundreds of years and saying that doesn’t make chatgpt worth any less. I have definitely been surprised by this and gotta say I’m expecting AGI a lot faster now. Even if literally all it did was predict what the average internet user would write in a certain context, that’s huge, cuz when you integrate all the little advantages of all the weird things one person knows another doesn’t, the collective knowledge is worth more than the sum of the parts. A tool which can tap into the sum total of human knowledge 24/7 and more rapidly than I can propose more questions for it, mainly I’m just excited to play with larger context size models so I can include more code and get big picture ideas about groups of stuff that are too numerous for my feeble meat brain to reason about. 7-9 things in working memory has always been the thing that would make humans inferior to AI in the long run. Even if it’s not that insanely smart (but realize: intelligence is a probabilistic concept and computers are great at multiplying probabilities precisely) if the thing can fit more stuff in memory than us and type faster than us and it doesn’t get tired or overwhelmed and give up (imagine your capability in a world where you had no tiredness and unlimited self discipline) in time it’s inevitable the transformers put us all to shame, and the more complicated the topic, the bigger of a shaming it’ll be, since the more complicated topics have exponentially more relations to reason about. Who’s gonna trust a human doctor to diagnose their stuff if the human brain holds 9 things and the AI holds thousands?

"Who’s gonna trust a human doctor to diagnose their stuff if the human brain holds 9 things and the AI holds thousands?"

The human brain can hold much more than 9 things and even though AI will be used in medicine broadly very soon, I really want the final diagnosis done by a human.

Once true AGI arrieves, I might change my opinion, but that might take a while.

> AI will be used in medicine broadly very soon

We have been hearing this since forever.

Revolutions do happen but not the way we expect. My anecdotical experience: no one in my team of about 30 people developing SW uses ChatGPT or similar in their day to day. This may change, or not.

AI is being used in medicine already. For example, in diagnostics. Most new diagnostics devices (e.g. CT scan, cardiograms) include AI systems that suggest an interpretation and point towards possible problems that a doctor might occasionally miss.

Granted, currently deployed systems are mostly awful, way behind the state of the art, and therefore mostly useless. Maybe it's because designing medical devices and getting them approved takes so long. Maybe it's because the manufacturers put AI in there for marketing purposes only, while assuming nobody will use the suggestiona anyway. In any case, I strongly expect the trend to continue and these systems to become very useful quite soon.

9 things is considered a standard for working memory (kind of like processor registers), for people with ADHD it's even less - 3-5.

Try writing a number from one piece of paper to another. If it's more than 7-9 numbers, you won't do it in one shot, unless you spend extra time memorizing it.

The equivalent for computers would be L1 cache on the CPU which is tiny.

More like cpu registers I would say :)

That can be increased quite a bit with practice. But it's also not important. It's just the cache memory -- it isn't the limit of what can be learned and recalled.

It is a limit on what you can reason about without a piece of paper.

I’m proficient at math, but my working memory is around 6, so I cannot add two three digit numbers to each other in my head (unless I see numbers to be added in front of me).

OK, but I and nearly all of my friends can, so we have duelling anecdotes here.

> Who’s gonna trust a human doctor to diagnose their stuff if the human brain holds 9 things and the AI holds thousands?

I will. As another commenter says, the brain isn't limited to 9 things at all. There's no way that I'll trust the diagnosis of a machine that won't understand me.

If a doctor uses AI to help with research, that would be OK. Just so long as the doctor is actually the one doing the diagnosis and prescribing the treatment.

The difference between your search query and theirs is clearly the level of expertise. Chatgpt has a great use case when you get started on a new subject; even with a very cluncky description it can point you into the core concepts of any field. Instead of reading 10 papers with are somewhat related but not what you are looking for, you can spend 3 minutes writing clumsy prompts and that's about it :)

Exactly. Does anybody remember in the early internet days people laughing at their parents for googling things like "please help with my back pain my doctor sucks".

Well, who's laughing now?

To add to this, using ChatGPT feels great in the moment, because it seems to work so well. For example, asking it for an itinerary while traveling gives you something that looks great.

However, once you actually start using it and see that the "ten minute walk" is actually an hour walk, or that a full third of the attractions it has shepherded you to are permanently closed, you realize that building that itinerary yourself from scratch using Google or TripAdvisor would take you less time than manually double checking everything ChatGPT says.

It's also quite surprising that people still think ChatGPT is capable of logic. Even for a complete layperson, all it takes is asking it to draw someone's family tree as an ASCII chart to see that text prediction only goes so far and there's not enough of a relational concept in there to comprise knowledge. There are many examples of asking it to solve famous puzzles with minor variations where it fails spectacularly.

The marketing behind ChatGPT is genius, but there is only so far you can go before the honeymoon is over and people start to really question what you brought to the table. Aside from that, ChatGPT isn't unique in what it can do, and others (including open source) are catching up fast.

That being said, I'd still use it for something like language learning (and other types of learning), where follow up queries (such as why you'd use one word instead of another, or how to rephrase something to be more polite) unlock a significant amount of value. It can also be useful to write trivial code, though I doubt a serious professional would do this (for several reasons, such as privacy and liability). Ultimately, ChatGPT fits squarely under "tool" and not under "intelligence".

It seems that as of right now, the killer app of ChatGPT is the boost in views you get by putting it in the title of your YouTube video.

As for googling, here are some examples of queries you can try and see how it works:

- summary of all the carbon neutral concrete methods, especially ones that can be done in a small industrial workshop as a prototype

- I have allergies in Thailand, mid-february. What may it be related to?

- list all the companies from Japanese stock exchange that have high debt rate

Those are top of my head, but really anything that is either a super-specific niche, or requires merging a few niches together, Google won't help you with.

For all of those, you HAVE to be okay with a complete garbage answer. Are you?

A machine generating confident bullshit will be the perfect companion for con-men to partially automate their workflow.

Humans are really gullible for the appearance of confidence. And humans are also very prone to wishful thinking.

How do you deal with it straight up lying? My problem with this whole system is, if I’m asking those questions it’s because I don’t understand the field well enough to answer it myself, which means I can’t pick up on if ChatGPT is lying…

Fair, but not completely true. The Thailand examples gives a detailed reasoning. You can use those building blocks to check. If it says Thailand is a cold country and uses that in its argumentation, it's shaky. You don't have to be an expert climatologist to make this judgement.

It's not just one clean answer and we're done. In my experience it is helpful in breaking the problem down into stuff you can Google.

> In my experience it is helpful in breaking the problem down into stuff you can Google.

Yeah I can see that being useful. I’ve also seen a lot of non-technical people straight up accept whatever comes out of it, so that’s a little worrying. It’s true of Google searches too, of course, but at least a google search gives N results someone can check rather than 1.

They straight up accept until they discover first mistakes :)

Fact check with google.

With the example questions I provided, it would take many hours to do research on the subject. GPT provided initial answers instantly, and then fact checking was easy.

That’s what we did with gpt-3. With plugins you can have gpt fact-check itself.

Also, if you have a system for dedicated knowledge, you can use embeddings - with embeddings gpt has very little room for hallucinations, and it can provide detailed references.

> Taking into account the downsides it looks like a hype bubble right now to me, and a draw in the long run. There’s just a whole lot of tech people trying to cash in on the hype.

Techies will realize that they are just giving ideas to O̶p̶e̶n̶AI.com, Microsoft Word, Google Docs and Notion. It is just the same AI bros re-selling their hallucinating snake oil chatbot that are under a new narrative for AI.

There is a reason why the only safe serious use-case of LLMs is summarization of existing text, since everything else it does is untrustworthy and is complete bullshit.

Their so-called 'revolution' is a grift.

Notion going all-in on the "AI" stuff is annoying/concerning to me. Mostly just that I live and die by a personal Notion wiki to keep my life organized, and if they eventually tank their service by investing too many resources into features that don't take off and I have to find a new tool to offload my brain into, I'm gonna be pissed...

I went with Logseq and for the first time in a number of years (actually since OneNote 2016, the last self hosted version) I am actually happy with my tooling again.

It doesn't cover everything OneNote 2016 did, but it does a lot more in other areas and it is progressing nicely.

This looks great! Do they have a (decent) mobile app? Being able to jot stuff down whenever, wherever is a make-or-break feature for me...

The app is already usable, at least on iOS, but for now sync is a bit rough around the edges, i.e., I need to verify it is synced or it will overwrite and I have to fix it using the page history which thankfully exist.

English speakers tend to miss a huge use case: Translation.

A 'use-case' done worse with LLMs especially with reliability. Translation is already done without a hallucinating LLM and can be done offline.

Summarization of existing text is the *only* safe and serious use-case for LLMs.

How is summarization "safe"? The summary might be wrong just as well.

The use-case is anything where occasional bullshit output is an acceptable downside to speeding up. More reliable outputs will enable more use-cases.

And what is a business that is fine with occasional (or frequent) bullshit output? Fake-news and spam.

Every business is fine with some frequency of bullshit output at some level. The question is how often exactly it happens and how much harm the bullshit can cause.

My point was that spam is the perfect use case for this tech. Of course there are other possible use cases, but spam and fake news content creation are the perfect fit. AI will enable one to easily clone the writing style of any publication and insert whatever bullshit content and keep up with the publishing cycle with almost zero workforce.

Want a flat-earther version of New York Times (The New York Flat Times)? Done. Want a just slightly insidiously fascist version of NPR? Done. Want a pro-Nato version of RussiaToday (WestRussiaToday)? Done.

And we already know people share stuff without checking for veracity and reliability first.

Machine translation is already not reliable even without LLMs, so it's not weak point of LLM translation.

GPT-4 translates much better than anything else out there, esp. when it comes to idioms and manner of speech.

DeepL.com? You dont need a general purpose LLM to do translation.

GPT-4 is a much better translator than Deepl especially for far apart languages.

I do wonder where LLM translator would take us to, considering that Japanese version of Bing Image Creator[0] is still proudly displaying a complete nonsense…

0: https://www.bing.com/create

    作成 *芸術* 開始日
    AI を使用した単語

> Taking into account the downsides it looks like a hype bubble right now to me.

100% agree and so glad to see someone else say it. I feel like people are losing their minds every time we go through the same hamster wheel.

To hear first hand, in the article above, the effect this is having on ML engineers breaks my heart.

Llm are a tool that understand intent. It makes super easy to compose api into agent and give them a task https://python.langchain.com/en/latest/modules/agents/agents...

That is just the introduction, showcasing what level of sophistication you get with just Google and Wikipedia as tools

Now imagine task rabbit or fiver as tools. Ai can make things happen in the real worlds.

These llm have limited attention but infinite focus. You can parallelize them, you can have one direct a fleet of other llm, you can have llm checking input and outputs for correctness from the other models and feedback that information to the controlling model so that it can improve the promp to the other as it tries to reach it's goal

And the goal can be far fetching (manufacture fake artsy trinket and import them from China to distribute etsy) or nefarious (produce subtle propaganda in a moltitude of wordpress website, register accounts on Wikipedia, reddit, create a sophisticate network of citations)

> OpenAI revolutionized… rewriting things with slightly different wording?

Yes, if you try hard enough, you can try to cast transformational shifts as trifling.

- e.g. “Barteen, Shockley , and Brittain made a smaller version of the vacuum tube.” (transistors)

- “Scientists discovered that light could carry information, like electrical wires do.” (fiber-optics)

The effects (including the harder to measure cultural shifts) matter more than some uncharitable characterization.

Also, the “it is not X” thinking is the result of present fixation. Such argumentation is, at best, quite narrow. Perhaps applicable in specific defined markets and situations but hardly a good mindset for making sense of how the world is changing. Hence the cliché, “The Stone Age didn’t end because we ran out of stone.”

The psychological undertones in the comment above are probably “people, stop exaggerating”. From one overreaction to another, it seems.

The popularity of ChatGPT revolutionized time. For learning, for many kinds of busywork (it's redefining what is and isn't "busywork), for planning. And most important: we don't know what we have yet because it's still being built. It's a tool. It's not the "capabilities" it's what people get out of it.

You mention blogging from the standpoint of writing it all yourself, and then using a tool to tweak it. That's not the revolutionary part. It's collaborating with the tool to write the post.

I need feauteres from n article description and chargpt can just extract it without any effort.

I don't know any other library I could just use with this task with nearly the same quality besides some regex soup.

You act like writing itself doesn't take time and energy. It has sped up my grant writing 6 fold. Any long-ish form writing that I need to do now happens at warp speed

if we only focus on this part. this function represents that most of content creator don't need to convert or reproduce their content to fit another group of customers.

Reply on news.ycombinator.com