Post AU8PCXTVHChnCSD8oi by mia@hcommons.social
(DIR) More posts by mia@hcommons.social
(DIR) Post #AU7uPjptX3UgisUxAu by simon@fedi.simonwillison.net
2023-03-30T01:58:59Z
0 likes, 1 repeats
What's the right name to use for systems like ChatGPT, Bing, Bard and Claude?If you think "something else" please reply in a comment.And I know "bullshit artists" might be a tempting description, but I'm looking for a term that's useful in mainstream discourse.
(DIR) Post #AU7ut8OHJlb9H8b8j2 by TedUnderwood@sigmoid.social
2023-03-30T02:06:11Z
0 likes, 0 repeats
@simon I can see a case for "chatbots," because after RLHF they're not exactly language models in the original sense of that term. But "chatbots" is not quite right for things that can be invoked via API. So I'm sticking with LLMs. It has the virtue at least of making people wonder what a model is.
(DIR) Post #AU7v8TT7ETXEzXAvnU by irwin@saturation.social
2023-03-30T02:03:48Z
0 likes, 0 repeats
@simon SALAMI is good and on point but I can’t bring myself to use it in serious discussion.
(DIR) Post #AU7vLKGEmHnUCOHOsK by kissane@mstdn.social
2023-03-30T02:05:42Z
0 likes, 0 repeats
@simon I think I’m at LLMs for a nerd audience, AI chatbot for general
(DIR) Post #AU7vqwiLrIsqRAVWoS by danlyke@researchbuzz.masto.host
2023-03-30T02:06:54Z
0 likes, 0 repeats
@simon I've been using LLM, but I see reasons to adopt the SALAMI (Systematic Approaches to Learning Algorithms and Machine Inferences) acronym that's being used by various people exploring the ethics of the space.
(DIR) Post #AU7w1Oeson8i03ruN6 by SnoopJ@hachyderm.io
2023-03-30T02:08:07Z
0 likes, 0 repeats
@simon I side with LLMs in these options, but would settle for "chatbots" without the side of """AI""" which I consider to be harmful in its credulity
(DIR) Post #AU7wPtcbLisSyrU2Cm by w8emv@hachyderm.io
2023-03-30T02:12:33Z
0 likes, 0 repeats
@simon “trained weasels”, see the post from @tess https://mastodon.social/@tess/110105460869464011
(DIR) Post #AU7waXPeGIg1HoV0e8 by paul_ipv6@infosec.exchange
2023-03-30T02:18:40Z
0 likes, 0 repeats
@simon "the great and powerful oz"pay no attention to the lame ML behind the curtain.
(DIR) Post #AU7wn2Q8niYuvlgqVk by simon@fedi.simonwillison.net
2023-03-30T02:22:27Z
0 likes, 0 repeats
@SnoopJ Yeah, I have real problems with "AI" - but "chatbots" on its own doesn't feel quite enough, since there have been plenty of chatbots in the past that are nothing like the LLM-powered ones
(DIR) Post #AU7xT4hchszCAdubZ2 by SnoopJ@hachyderm.io
2023-03-30T02:26:22Z
0 likes, 0 repeats
@simon it'd be great if I could get away with "semi-random text generator" but I know that I cannot
(DIR) Post #AU7xqVRZRyUblcZnlo by evanescent@mamot.fr
2023-03-30T02:32:52Z
0 likes, 0 repeats
@simon I think “LLM Chatbot” is about right and avoids getting distracted by the “is or isn’t it an AI” discussion.
(DIR) Post #AU7y1FufX7UgmjJRaa by FeralRobots@mastodon.social
2023-03-30T02:36:32Z
0 likes, 0 repeats
@simon The horse has already left the barn but as long as we're spitballing, they should never be called AIs.They *are* LLMs, so it's accurate to call them that - & it provides an opportunity to clarify how they work & why there's no 'I' there.
(DIR) Post #AU7yDClP4JN6UZg0WW by laedit@piaille.fr
2023-03-30T02:37:09Z
0 likes, 0 repeats
@simon artificial learning chatbots
(DIR) Post #AU7zOSx9JhXnqI6WQ4 by DalzAsylum@framapiaf.org
2023-03-30T02:54:53Z
0 likes, 0 repeats
@simon @affordance coined a perfect term to me. Generative artifacts.
(DIR) Post #AU80JFQyFivAQ782bI by simon@fedi.simonwillison.net
2023-03-30T03:03:25Z
0 likes, 0 repeats
LLMs seems to be winning by a lot - clearly it's a less obscure acronym than I thought (though a Mastodon audience likely skews these results)It's not why I asked the question, but this did inspire me to add a new "llms" tag to my blog, https://simonwillison.net/tags/llms/ - then back-filled the tag to 114 posts
(DIR) Post #AU80gUvyn0wT9ZieUi by benschwarz@front-end.social
2023-03-30T03:08:32Z
0 likes, 0 repeats
@simon no non nerd is calling it LLM. AI chatbot for sure
(DIR) Post #AU81ND6UxSkeJ10RWa by drwho@hackers.town
2023-03-30T03:15:18Z
0 likes, 0 repeats
@simon Constructs.
(DIR) Post #AU81khX5zR9lL8eF7o by simon@fedi.simonwillison.net
2023-03-30T03:20:12Z
0 likes, 0 repeats
@evanescent Yeah that definitely feels most correct to me - especially as they're not pure LLMs, they're specifically the ones that have been fine-tuned to work with chat dialog
(DIR) Post #AU82NBvjQCaAcUWbmi by JesseSkinner@toot.cafe
2023-03-30T03:26:32Z
0 likes, 0 repeats
@simon I voted LLM, but whenever I talk to my friends and family about this, they all call them AIs.Though this lumps in things like Midjourney and even some of the "make me look old" kinds of things.
(DIR) Post #AU830AXAXw4daEccYy by dltj@code4lib.social
2023-03-30T03:31:29Z
0 likes, 0 repeats
@simon LLM is a sub-field of AI. I’ve found that there is actually a sub-discipline in between: LLMs are a form of Machine Learning, which itself is a subset of the Artificial Intelligence field. Visual Language Models are also a sub-field. See https://arstechnica.com/information-technology/2023/03/embodied-ai-googles-palm-e-allows-robot-control-with-natural-commands/
(DIR) Post #AU83Nh3VnWGN6ky2Oe by peladocollado@hachyderm.io
2023-03-30T03:31:50Z
0 likes, 0 repeats
@simon yeah, I think selection bias is strong here. If I asked my wife about LLMs, she’d look at me funny. If I asked about AIs or AI chatbots, she’d know instantly what I meant
(DIR) Post #AU84ONePWulvAW7NQG by kornel@mastodon.social
2023-03-30T03:49:55Z
0 likes, 0 repeats
@simon LLM is probably best for now, but I bet it's going to be short-lived. Next breakthrough will add another technique to the mix, the acronym will change, and LLM will be demoted to being just a stepping stone, like deep learning, word2vec.Applications of LLMs seem wide enough that when they become mainstream, it will just be called "AI".
(DIR) Post #AU88qsMPLTRSL0XYuG by benjamin@framapiaf.org
2023-03-30T04:40:54Z
0 likes, 0 repeats
@simon i like bullshit generator.
(DIR) Post #AU8EtNy9OYxKftm3qC by petermhoward@hachyderm.io
2023-03-30T05:48:36Z
0 likes, 0 repeats
@simon i've been going with “AI LLMs” to acknowledge the broader bundling and the misconception but clarifying that it's a particular implementation of AI... I'm wary of using “chatbot” cause I think that label is already throwing up a lot of misconceptions
(DIR) Post #AU8FEQne4AxdXDtb96 by com@mastodon.social
2023-03-30T05:49:05Z
0 likes, 0 repeats
@simon All AI chatbots are LLMs, but not all LLMs are chatbots. A chatbot is a LLM + prompt training and interface.
(DIR) Post #AU8FReVwO61pplv2OW by osma@mas.to
2023-03-30T05:52:51Z
0 likes, 0 repeats
@simon Years ago I decided that every time someone else might say "autonomous vehicle", I would instead say "automated vehicle" - because they're not really autonomous.Chatbot is a perfectly adequate label for chatbots, but you stuck the "AI" there so I could not choose it. LLMs of course do have other uses beyond chatbots.
(DIR) Post #AU8NW0VtKuicr4U2Fc by kylejohnston@mastodon.online
2023-03-30T07:24:57Z
0 likes, 0 repeats
@simon LLMs is definitely more accurate, but AI chatbot feels clearer for a mainstream audience
(DIR) Post #AU8OpAzIiWAanSwZhw by sleepyfox@hachyderm.io
2023-03-30T07:38:21Z
0 likes, 0 repeats
@simon "box of trained weasels"
(DIR) Post #AU8PCXTVHChnCSD8oi by mia@hcommons.social
2023-03-30T07:45:50Z
0 likes, 0 repeats
@TedUnderwood @simon and 'language model' doesn't imply understanding, which is important
(DIR) Post #AU8Pw52tJ9OCLhTlPE by JohnScottBridle@sigmoid.social
2023-03-30T07:52:14Z
0 likes, 0 repeats
@simon Language fabulators
(DIR) Post #AU8RAK7c9Srt7D2mXY by ethulhu@chaos.social
2023-03-30T08:05:37Z
0 likes, 0 repeats
@simon predictive text? jumbo autocorrect?anything that doesn’t include the term “AI”: a part of why people are getting snookered by them is because there’s so much marketing and hype and science fiction stories that tells us all that “AI” means “person”, and the term is poison now.
(DIR) Post #AU8W7TaR0LwHN7zBnE by xaphania@meemu.org
2023-03-30T08:58:53Z
0 likes, 0 repeats
@simon text generation engines (or text generation bots, or models, or whatever. the important part is they generate text and that's all they really do, currently at least).this would be opposed to a hypothetical knowledge engine, which actual has some kind of internal conceptual model of knowledge/information/facts that it's spewing out.
(DIR) Post #AU8fKwgPRjCiVbbPYO by albynton@framapiaf.org
2023-03-30T10:44:40Z
0 likes, 0 repeats
@simonI heard them being referred to as "smooth talk generators"
(DIR) Post #AU8oCAAuoT79Kol54C by foundhistory@hcommons.social
2023-03-30T10:19:13Z
0 likes, 0 repeats
@simon @mia @TedUnderwood I have very mixed feelings about this. LLM is more accurate. But it’s also easy (for ordinary people, decision makers, regulators, politicians) to ignore. That is, it’s boring.
(DIR) Post #AU8oCAiEoZll0ALh0S by mia@hcommons.social
2023-03-30T11:27:47Z
0 likes, 0 repeats
@foundhistory @simon @TedUnderwood I think it's already cutting through - I've had more conversations about AI with the widest range of people since chatgpt launched than in all the years previously. And things like the Pope's parka fake must mean more people encountered questions about generative AI
(DIR) Post #AU8oCBPUDmMdAIaEPw by foundhistory@hcommons.social
2023-03-30T11:56:28Z
0 likes, 0 repeats
@TedUnderwood @mia @simon That’s great. I’m glad it’s getting attention from people who can do things to help ensure at least some democratic control of it. Frankly, as long as that happens, I couldn’t care less which words we use!
(DIR) Post #AU8oCBzI4f0IxLKpE0 by sharonmleon@hcommons.social
2023-03-30T12:13:38Z
0 likes, 0 repeats
@foundhistory @TedUnderwood @mia @simon I want very much not to call it intelligence. I want people to know about it, consider its transformative power and its potential dangers, but we’ve got to avoid the anthropomorphizing.
(DIR) Post #AU8oCCbZmJd2s5FOts by TedUnderwood@sigmoid.social
2023-03-30T12:25:54Z
0 likes, 0 repeats
@sharonmleon @foundhistory @mia @simon The thing is, though, that it is intelligence? I mean if people prefer to call it “ability to interpret pictures, read captions, make inferences and explain jokes” that’s okay, but I don’t know what distinction we’re preserving by doing that
(DIR) Post #AU8oCDN4vhctFPTKwS by foundhistory@hcommons.social
2023-03-30T10:19:20Z
0 likes, 0 repeats
@simon @TedUnderwood @mia I understand not wanting to amplify the tech industry’s hype, but we also don’t want to downplay the likely consequences of this technology. There has to be a way to tell people that something will be transformative (quite probably for worse) and command their necessary attention without being a cheerleader for it.
(DIR) Post #AU8ohGQ9dot7gOiTUe by TedUnderwood@sigmoid.social
2023-03-30T12:31:32Z
0 likes, 0 repeats
@sharonmleon @foundhistory @mia @simon It’s not a sentient agent with personhood & desires. But in my view that’s the thing we need to work on: the assumption that intelligence can only come in the form of a sentient agent.
(DIR) Post #AU8pOZ7tyCfS9JpJJY by sharonmleon@hcommons.social
2023-03-30T12:39:20Z
0 likes, 0 repeats
@TedUnderwood @foundhistory @mia @simon I guess the distinction comes for me in the inability of the model to ask its own questions. That’s what I see as being at the heart of intelligence. But also I’m not a philosopher.
(DIR) Post #AU8pWiYylhOBjXNZ8C by jamessmithies@hachyderm.io
2023-03-30T12:40:49Z
0 likes, 0 repeats
@TedUnderwood @sharonmleon @foundhistory @mia @simon Katherine Hayles’ Unthought is good on this imho, exploring ‘non-conscious cognition’ in an expansive way that reflects solid scientific and technical research but draws a clear line between that concept and everyday interpretations of (solely human or even animal) ‘intelligence’.
(DIR) Post #AU8pvRZL5bYiGaL01g by sharonmleon@hcommons.social
2023-03-30T12:42:34Z
0 likes, 0 repeats
@jamessmithies @TedUnderwood @foundhistory @mia @simon Thanks for that. She is, indeed, one of the smartest people around.
(DIR) Post #AU8pvSDkfLswHvFH16 by TedUnderwood@sigmoid.social
2023-03-30T12:45:17Z
0 likes, 0 repeats
@sharonmleon @jamessmithies @foundhistory @mia @simon She has a piece on language models, and the sense in which they “think,” coming out in the next issue of NLH. (Special issue on “Culture, Theory, Data.”)
(DIR) Post #AU8uBmUNTceBbB3x9k by millionnl@geekdom.social
2023-03-30T13:29:20Z
0 likes, 0 repeats
@simon AA - Advanced Autocomplete
(DIR) Post #AU8wSk2a25ZCpvHAFE by TedUnderwood@sigmoid.social
2023-03-30T13:58:32Z
0 likes, 0 repeats
@electricarchaeo @benmschmidt @sharonmleon @foundhistory @mia It's just a little calculator for words. Of course, a calculator for words is an awesome thing.
(DIR) Post #AU8ybsymMsVF4dzBtw by elotroalex@hcommons.social
2023-03-30T14:22:34Z
0 likes, 0 repeats
@TedUnderwood @electricarchaeo @benmschmidt @sharonmleon @foundhistory @mia dude, if we could make "word calculator" stick that would be awesome! and way more accurate than the choices above.
(DIR) Post #AU9bY8gVvjVQpyeAmO by bbolli@swiss.social
2023-03-30T21:36:36Z
0 likes, 0 repeats
@simon Stochastic parrots