civilized 👑Jan 11, 2023
eevee 🦊

chatgpt is predictive text.
chatgpt is predictive text.
chatgpt is predictive text.
chatgpt is predictive text.
chatgpt is predictive text.

it's not even answering questions. it just pattern-matches that the next text after something that looks like a question is most often something that looks like an answer

Jan 11, 2023 at 1:12amWeb
Show threadGraham Sutherland / PolynomialJan 11, 2023
@eevee I've been describing it as a word recommendation engine.
Show threadmeejahJan 11, 2023
@gsuberland @eevee I try very hard to always refer to "AI" as "advanced statistics" .. but I do like this a lot :)
Show threadGraham Sutherland / PolynomialJan 11, 2023
@meejah @eevee it's not 100% perfect, but I feel like it's short enough to get the point across while also reflecting the actual mechanism reasonably well.
Show threadSnowDerg ❄️🪶Jan 11, 2023
@eevee so it's just markov chains on steroids?
Show threadGraham Sutherland / PolynomialJan 11, 2023
@SnowDerg @eevee a bit like a markov chain glued to a content recommendation algorithm.
Show threadReboot/FitzJan 11, 2023
@gsuberland @SnowDerg @eevee glued as in two models or a larger neural network implementing both concepts?
Show threadGraham Sutherland / PolynomialJan 11, 2023

@reboot @SnowDerg @eevee essentially both concepts smushed together in a stateful manner as a single model, at least in terms of external behaviour.

the actual architecture isn't the same as those individual components but that's a separate conversation.

Show threadReboot/FitzJan 11, 2023
@gsuberland @SnowDerg @eevee but, I kinda want to have that conversation... (as a person who does a lot of general CS work.)
Show threadGraham Sutherland / PolynomialJan 11, 2023

@reboot @SnowDerg @eevee I'm probably not the right person to talk about it in deep detail since the gory innards of LLMs and autoregressive models aren't my wheelhouse. Someone did post a good article on LLMs the other day but I can't spot it.

I'd recommend searching online for an explainer but unfortunately every single one I found on the first page of Google was a bust due to gushingly anthropomorphising the model in a way that falsely implied an ability to develop understanding.

Show threadReboot/FitzJan 11, 2023
@gsuberland @SnowDerg @eevee makes sense as any sense that it developed a greater "understanding" would be context kicking in rather than actual learning. Mind expanding L.L.M. for me?
Show threadGraham Sutherland / PolynomialJan 11, 2023

@reboot @SnowDerg @eevee LLM = large language model

I find it important to not use the term "understanding" in this context because the model is *not* understanding anything. It can *mimic* understanding through correlative processes, for sure, but the distinction is important.

Show threadGraham Sutherland / PolynomialJan 11, 2023
@reboot @SnowDerg @eevee this is why I call it a word recommendation algorithm. you give it a prompt and it correlates that with wording that it has seen before from its massive training corpus of human conversations. from there it can launder those words through a generative language model that "follows" the linguistic structures it has been trained on, to produce a paragraph of text.
Show threadReboot/FitzJan 11, 2023
@gsuberland @SnowDerg @eevee make sense. In marginally related terms, the responses rejecting generation of text (due to violation of rules, or general non-understanding) would be due to training to generate responses rejecting the given task, right? (This also explains why I can't ask it to, for example, play music.)
Show threadGraham Sutherland / PolynomialJan 11, 2023

@reboot @SnowDerg @eevee it cannot draw from anything it hasn't been trained on. it can't create new information or concepts.

the rules limitations are largely artificial, and the general mechanism is to "prime" the context of the session with a list of terms it shouldn't respond to. you can always break out of the restrictions with clever linguistic tricks because it doesn't actually understand the rules, you're just tweaking the context enough to negate the originally provided words.

Show threadawooo  🏴‍☠️   🐾 ⎇Jan 11, 2023

@reboot @gsuberland @SnowDerg @eevee Yep, it can also be quite easily tricked into doing anything, you just have to hijack the flow of the conversation and "hypnotize" it.

The important thing is, it doesn't have any internal distinction between its own words and what you say besides the context being given to it (probably some header/prefix, which is also how it knows the time lol), so you can quite literally put words in its mouth.

Show threadGraham Sutherland / PolynomialJan 11, 2023
@awooo @reboot @SnowDerg @eevee indeed. it's essentially a correlative prediction/recommendation engine for "if these are the words that have been said so far, with these logical delimiters between participants, what comes next?"
Show threadawooo  🏴‍☠️   🐾 ⎇Jan 11, 2023
@gsuberland @reboot @SnowDerg @eevee Yeah, it's definitely impressive how it can keep the separation between different participants and objects, search for context and follow a logical flow (as in the text, not necessarily in the "thinking" part). But throw enough at it and it will lose track of that and that's when you can inject whatever you want.
Show threadawooo  🏴‍☠️   🐾 ⎇Jan 11, 2023

@gsuberland @reboot @SnowDerg @eevee I think the question about the fastest marine mammal gives a lot of insight about what it actually does internally. I tested it on GPT-3 (not chatGPT) and tried confronting it about its mistake, instead of apologizing it decided to change the definition of a mammal to fit its wrong answer instead of the other way around.

It's a bit like a human with confirmation bias cranked up to 11.

Show threadAnca BuitJan 11, 2023
@awooo @gsuberland @reboot @SnowDerg @eevee Let's put it in charge of everything then. Nothing can possibly go wrong.
Show threadGraham Sutherland / PolynomialJan 11, 2023
@reboot @SnowDerg @eevee if you did this with a small corpus it would be laughably bad. but because it has been trained on a metric crapton of human conversation on a huge variety of topics, there's a disparity of sourcing data between what it has "seen" and what you have seen. so even though the actual correlation and generation processes are ultimately simplistic, the huge amount of data it can draw from makes it *seem* convincing.
Show threadLafncow Jan 11, 2023
@reboot @gsuberland @SnowDerg @eevee You might want to look for an explainer on transformer models (the T in Chat GPT), and your intuition is about right: they are neural nets that maintain a series of hidden states as data is fed in.
Show threadeevee 🦊Jan 11, 2023
@SnowDerg yes
Show threadSkaveRat 🐀 Jan 11, 2023
@eevee honestly, that makes it even more impressive with what it can do. and the amount of variety of it
Show threadI̵E̶E̸E̵ ̷1̵1̸4̴9̴.̷1̷Jan 11, 2023

@eevee chatgpt is the best way to get the money to the bank and the bank is not working on the phone with the phone number for the day I will be there in a few minutes to get back to you in the morning and I will be there in a few minutes to get a new one for the day when I get back to work on the way home and get back to work ok if u want to come over and get it done tomorrow morning and I will be there in a few minutes to get it done and I have the other room to get a ride from there but it was raining and it took u to bed lol.

(This was a story written entirely from the middle word on my predictive text that looks suspiciously like ai generated first drafts)

Show threadJohann-Tobias SchägJan 11, 2023
@eevee
Yeah but i am predictive text attached to a neural world model representation.
Show threadBrodie RobertsonJan 11, 2023
@eevee Isn't human reasoning just pattern matching at a far more rudimentary level
Show threadarkywarkyJan 11, 2023
@BrodieOnLinux @eevee i mean yeah but it's far more complex than that. Human physiology is a lot more complex than a couple billion weights in a text - prediction network
Show threadBrodie RobertsonJan 11, 2023
@arcade @eevee Are you sure? We don't even remotely have an understanding of human intelligence
Show threadarkywarkyJan 11, 2023
@BrodieOnLinux @eevee we do know that like gut chemistry has an effect on it tho, and we do know that humans dont work on Unicode too, so we are still very different from the machines.
Show threadRelee the SquirrelJan 11, 2023

@eevee It's really good at that too, huh? Well, usually.

So, what are folks saying it is that you're trying to correct? I'm imagining it's the folks who think it's alive or something, but there's so many different views.

Show threadeevee 🦊Jan 11, 2023
@relee people don't describe it explicitly as very many things, but i keep seeing glimpses of it being treated as a knowledge engine
Show threadRelee the SquirrelJan 11, 2023

@eevee I have seen folks saying it's great for advice and answering questions, but it is certainly wrong often and doesn't actually know, it's just putting words that it thinks would fit best in there, based on its training.

I hear about folks using it for code generation, too, but it's really not going to eliminate pro coders. You still need to understand when and why code it produces isn't good, and know when some advice is just wrong.

Show threadtuxfloJan 11, 2023
@eevee yes, I think lots of people get this wrong. My common example for "it does not understand what you are saying" is 'Hey ChatGPT explain git to me without using the letter i'
And then see it failing.
Show threadhikewithrichardJan 11, 2023
@eevee convince the parrots. You can’t scroll without chatGPT articles. More worryingly it is under the Thiel sphere of influence.
Show threadarkywarkyJan 11, 2023
@eevee like people need to stop anthropomorphising these
Show threadGabriela Noriega 🏳️‍⚧️ 🇧🇷Jan 11, 2023
@eevee It's still a fun way to do "secular divination" tho.
Show threadChad GeidelJan 11, 2023
@eevee I've explained it like an advanced version of the predictive text that Google does for autocomplete. "Can my dog eat ..." is basically the same thing that ChatGPT does.
Show threadCampbell JonesJan 11, 2023
@eevee This is the best explanation that I've seen for my gripes with the rhetoric around ChatGPT.
Show threadMoiraPrimeJan 12, 2023
@eevee this is pretty spot on. I've asked it a few hyper specific questions, like directions to two specific places nearby, and it makes up instructions that look familiar but are absolute nonsense. Tells you to turn on roads that don't even intersect each other.