mrtranscendence

That doesn't work for library projects, though.

> Your local plumber is going to want a funny action movie trailer slash plumbing advertisement to advertise their services. They wouldn't have even been in the market before.

And why would your local plumber hire someone to produce this funny action trailer (which I'm not convinced would actually help them from an advertising perspective), when they can simply have an AI produce that action funny action trailer without hiring anyone? Assuming models improve sufficiently that will become trivially possible.

> Independent filmmakers will be making their own Miyazaki and Spielberg epics that cater to the most niche of audiences - no more mass market Marvel that has to satisfy everybody, you're going to see fictional fantasy biopic reimaginings of Grace Hopper fighting the vampire Nazis.

Well, first of all, if the audience is "the most niche of audiences", then I'm not sure how that's going to lead to a sustainable career. And again -- if I want to see my niche historical fantasy interests come to life in a movie about Grace Hopper fighting vampire Nazis, why will I need a filmmaker to create this for me when I can simply prompt an AI myself? "Give me a fun action movie that incorporates famous computer scientists fighting Nazis. Make it 1.5 hours long, and give it a comedic tone."

I think you're fundamentally overvaluing what humans will be able to provide in an era where creating content is very cheap and very easy.

> "We're profitable on inference. If we didn't pay for training, we'd be a very profitable company."

That's OpenAI (though I'd be curious if that statement holds for subscriptions as opposed to API use). What about the downstream companies that use OpenAI models? I'm not sure the picture is as rosy for them.

Microsoft claims that they have an AI setup that outperforms human doctors on diagnosis tasks: https://microsoft.ai/new/the-path-to-medical-superintelligen...

"MAI-DxO boosted the diagnostic performance of every model we tested. The best performing setup was MAI-DxO paired with OpenAI’s o3, which correctly solved 85.5% of the NEJM benchmark cases. For comparison, we also evaluated 21 practicing physicians from the US and UK, each with 5-20 years of clinical experience. On the same tasks, these experts achieved a mean accuracy of 20% across completed cases."

Of course, AI "doctors" can't do physical examinations and the best performing models cost thousands to run per case. This is also a test of diagnosis, not of treatment.

Waar heb je het over? "Welgelegen Buitenrust Nooitgedacht Rustenburg" is volkomen cromulent Engels.

For what it's worth, I do use AI for language learning, though I'm not sure it's the best idea. Primarily for helping translate German news articles into English and making vocabulary flashcards; it's usually clear when the AI has lost the plot and I can correct the translation by hand. Of course, if issues were more subtle then I probably wouldn't catch them ...

> the iPhone was only "in charge" for a brief year or two, and then Android ate its lunch in terms of marketshare.

This is true worldwide, but there are significant regions where iOS quite handily beats Android (such as the US, Japan, and even some parts of Europe).

Oh man, I played the shit out of the Lone Wolf books when I was a kid in the eighties and nineties. My brother had the first couple of them, and I still have his old books somewhere with his notes. I should find them for nostalgia's sake (my brother having long since passed).

> Character stats, names, details about players - those are inputs, and structured ones at that.

Some details about players are structured and can be easily stored and referenced. Some aren't. Consider a character who, through emergent gameplay, develops a slight bias against kobolds; who's going to pick up on that and store it in a database (and at what point)? What if a player extemporaneously gives a monologue about their grief at losing a parent? Will the entire story be stored? Will it be processed into structured chunks to be referenced later? Will the LLM just shove "lost a father" into a database?

Given current limitations I don't see how you design a system that won't forget important details, particularly across many sessions.

You sound like a fun guy.

You seem to readily agree that my use case is inappropriate for LLMs, but not ToucanLoucan?

We're not allowed to record meetings at work, with some exceptions such as trainings. I'm not at all sure why, but I believe it has to do with merger plans. We're not working on anything particularly shady, to my knowledge.

Notes are often taken but not always. Depends on the people in the meeting.

I suspect you're overstating the degree to which an LLM might be unsuitable for some types of work. For example, I'm a data scientist who works primarily in the field of sales forecasting. I've found that LLMs are quite poor at this task, frequently providing answers that are inappropriate, misleading, or simply not a good fit for the data we're working with. In general I've found very limited use in engaging LLMs in discussion about my work.

I don't think I'm calling myself a super special snowflake here. These models are just ... bad at sales forecasting.

LLMs aren't entirely useless for me. I'll use ChatGPT to generate code to make plots. That's helpful.

If I'm not conscious, it's rather difficult to explain what I'm experiencing right now.

Prove what point? There have clearly been crooked or underhanded companies that achieved success. Microsoft in its early heyday, for example. The fact that they paid a price for it doesn't obviate the fact that they still managed to become one of the biggest companies in history by market cap despite their bad behavior. Heck, what about Donald Trump? Hardly anyone in business has their crookedness as extensively documented as Trump and he has decent odds of being a two-term US President.

What about the guy who repaired my TV once, where it worked for literally a single day, and then he 100% ghosted me? What was I supposed to do, try to get him canceled online? Seems like being a little shady didn't manage to do him any harm.

It's not clear to me whether it's usually worth it to be underhanded, but it happens frequently enough that I'm not sure the cost is all that high.

Act as high and mighty as you please, I won't mind. But bear in mind that:

* most people will find it surprising that showing a photo from the internet to ChatGPT is as unsafe as opening a random, untrusted exe.

* many people don't even understand that it's unsafe to open random, untrusted exes.

Are you seriously suggesting that we should leave all these people to the wolves, because they're less knowledgeable about security vulnerabilities than you?

Wait, how does rewriting a prompt until it gives you the output you expect help the LLM learn? Are you suggesting better prompting gets fed back into the training process in some helpful way? This feels confused.

People aren't infallible, but in my experience they're much less likely to give me incorrect factual information than LLMs. Sometimes lawyers are wrong, of course, but they are wrong less frequently and less randomly. I've typically been able to get away with not verifying every single thing someone else tells me, but I don't think I'd be that lucky relying on ChatGPT for everything.

Edit: and it's a good thing, too, because I'd never be able to afford getting second legal opinions and I don't have time to verify everything my lawyer tells me.

X says "I would like RCTs on this subject", and you reply that most RCTs are worthless. Great, what do you want, a cookie? Presumably they want good RCTs, not low-quality ones.

True, you might Xerox copies but you wouldn't Kleenex your stuffy nose (though I've heard of someone bandaiding a cut before). Like Kleenex I actually think AI is slightly too awkward to say to become a common verb.

I did something similar to this using RAG except for Vampire rather than D&D. It wasn't overwhelmingly difficult, but I found that the system was quite sensitive to how I chunked up the books. Just letting an automated system prepare the PDFs for me gave very poor results all around. I had to ensure that individual chunks had logical start/end positions, that tables weren't cut off, and so on.

I wouldn't fine-tune, that's too much cost/effort.

Navigation between cells with j/k doesn't seem to work, unfortunately. I'm used to not having to move my hands to use the mouse when editing a Jupyter notebook.

I generally tend to think that complaints about dependency management in Python are way overblown -- it's usually not a problem for me. But lately I've been trying out some off-PyPI projects to investigate time series foundation models, and it would have made my life so much easier if the implementers of these libraries hadn't decided to pin to extremely specific versions of ten or twenty different dependencies. No, your library does not need exactly NumPy 1.22.3. You're just throwing unnecessary obstacles in the way of people using it.

That pinning wouldn't really be a problem ordinarily, where I control the environment, but I'm running this particular code in a managed environment that I can't easily modify.

You're missing the point. What's not realistic is to tell everyone that they should abstain from any type of entertainment that requires power (TV shows, movies, video games, etc) and should only read books and play board games instead. I don't care what kind of renaissance board games are undergoing, most people still only play the mass market classics, and then only rarely.

I don't know how much energy Netflix uses serving a movie, but playing a video game on my PC for two hours where I'm located might generate a kg of CO2. That's about as much as I'll breathe in a day. Relative to other sources of atmospheric CO2 I'm not that concerned.

> If advertisements were opt-in, I expect a single digit percentage of people would ever elect to see them.

I mean, you'd see the same thing if paying for your groceries were opt-in. Is that also a net bad for the world? Ads do enable the costless (or cost-reduced) provision of services that people would otherwise have to pay for.

I used to be an Informatica ETL dev, briefly. I laugh at the description of it as a low-code solution, as we used it as a simple frontend for staggering amounts of hand-coded SQL. I'm not sure I ever used Informatica as it was intended.

It's open source if you're using language like a normal human being. If you're a bit of a pedant and wish everyone to adhere to definitions imposed from on high regardless of real-world usage, it's absolutely not open source.

Sometimes, when I'm undertaking the arduous work of assigning probabilities to everything I could possibly say next in a conversation, I wish that I weren't merely a stochastic autoregressive next-token generator. Them's the breaks, though.

I'll be honest, I'm less concerned by any movement to make us "lobotomites" -- a movement which I haven't witnessed at all -- than I am by people who really want to be able to keep saying slurs.

That's a dangerous path to go down. I've encountered many biases that I don't feel are especially reflective of reality. These range from dumb (women are poorer drivers than men) to extremely harmful (black persons are stupid and lazy).

I for one would not be prepared to defend the persistent bias against black persons and immigrants as having a basis in reality. YMMV.

You forgot to preface that with "Uhm ackshully..."