I'm curious whether my #photoshop skills have been rendered completely obsolete by #generativeAI tools, so I am going to challenge #dalle to recreate some of my better contest submissions from 20 years ago.
First prompt: "Computer showing an MC Escher tessellation with lizards climbing out the screen walking on the desk and climbing back in."
Hi #medlibs@medlibs and others interested in AI hype in research methods
Anyone interested in collaborating with me on a letter to the editor saying, actually, no, creating #SystematicReview / #EvidenceSynthesis searches is not in fact a task aligned with the capabilities of #generativeAI ?
Several times I have pointed out to folks that #GenerativeAI systems is almost certainly inappropriate for use in any scenario where factual output is required and they've often countered that they'll "train the model" to fix that. Which won't work, because it's not a quality of data issue.
So-called hallucinations are not an accident, they are the only things these systems can produce. Nothing they generate is more or less true. Every output is a hallucination, but some are obviously so.
There are some techniques that can be used to nudge #GenerativeAI towards outputs that humans will judge to be "more accurate", like RLHF (reinforcement learning from human feedback). But to be clear, these methods don't make the underlying system any more aware of facts than before: they tweak the output such that it matches our expectations a bit more closely. But this approach has flaws when it comes to constraining those outputs to only present facts.
"We need to regulate on principle." — last week, @paulk spoke at the Access Partnership-hosted webinar to discuss the European Union’s #AIAct and the broader regulatory debate on #GenerativeAI and the creative industry 🎬 You can watch the recording below: https://www.youtube.com/watch?v=J_I_1WND_XY
I'm tired of the #AI bros warning us of superintelligence again and again while they absolutely make sure that the exact future they are warning us about appears. They want to create a mote so that only they can do the research and productize the #MachineLearning work without taking into account the actual harm that current systems might face. They don't want to solve those problems, only the imaginary problems. Look over there and not here.
Just saw a presentation by @pamelasamuelson on #GenerativeAI and #Copyright (https://www.youtube.com/watch?v=6sDGIrVO6mo). I agree with the general sentiment that AI output is likely not going to infringe the copyright of the original training data, but if you've any interest in the topic, this is a great summary from one of the leading legal scholars in the field.
I really do wonder whether I’m too Millennial for #Discord.
Like, whenever I revisit one of the servers I’m on that I haven’t looked at for weeks or months, Discord will be like “new mentions!” in the channel list, and have red number badges next to channel names, but clicking them just takes me to the channel and there doesn’t seem to be a way to see these mentions (which are most likely just @everyone). The “inbox” on the top right also doesn’t show mentions older than 7 days. 🤷♂️
Du fragst nach einer aktuellen Newszusammenfassung, der 1. Punkt stimmt noch und dann kommen nur alte News. Auf den Fehler angesprochen wird dir erklaert, dass diese aber doch so wichtig waeren.
Auch das ist ein Pre-Programmes Bias und das macht die ganze Nummer richtig schwierig zu handlen! #Bard#ChatGPT#GenerativeAI
Hey #LLM#GenerativeAI experts out there: How the heck do you debug things when your model returns the “wrong” answer? Where do you even start? If I controlled all prompts and the model yields an unexpected result, what can I even do about it?
Aus einem hochkant Dom-Foto ein horizontales gemacht mit der Photoshop-Beta mit generativer KI. So mittel, sobald Leute dabei sind. Landschaften und natürliche Objekte kann es sehr gut. Sobald es kreativ werden muss, kommt es ins Schwimmen.
Aus einem hochkant-Foto, dass ich vor zwei Wochen in Köln gemacht habe ein sehr breites Foto gemacht. Das ist toll! Die Vorlagen war natürlich auch sehr geil. :-) #Photoshop#GenerativeAI#KI
Sam Altman says #OpenAI "of course" has no plans to leave Europe, reversing a threat to leave the region if it can't comply with the EU's upcoming #AI Act.
🎈 According to former Google researchers, it looks like the Bard chatbot has a glaring keyword manipulation exploit open to any black-hat SEO who wants to try.
➥ @baldur
#AI writing assistants can cause biased thinking in their users according to research conducted by Maurice Jakesch, a doctoral student of information science at Cornell University. #LatentPersuasion
A biased "assistant" was able to influence the arguments made in essays.
As applications like ChatGPT and rival products rise in popularity, experts and stakeholders are split on whether and how Section 230 of the Communication Decency Act — a liability shield for internet companies over third-party content — should apply to the new tools.
Ashley Johnson, a senior policy analyst at the Information Technology and Innovation Foundation said the “most likely scenario” is if generative AI is challenged in court, it “probably won’t be considered covered by Section 230.”
“It would be very difficult to argue it is content that the platform, or service, or whoever is being sued in this case had no hand in creating if it was their AI platform that generated the content,” Johnson said."
#AI#GenerativeAI#OpenSource#LLMs#Chatbots: "I’ve been leading and researching generative AI system releases, including OpenAI’s GPT-2, since these systems first started to become available for widespread use, and I now focus on ethical openness considerations at Hugging Face. Doing this work, I’ve come to think of open source and closed source as the two ends of a gradient of options for releasing generative AI systems, rather than a simple either/or question."
Why computer-generated art and literature is ultimately boring (at least when it's an end in itself), because generative systems, including today's generative AI, boil down to glorified kaleidoscopes:
I wrote this as lecture notes for a university conference in 2020. Since no proceedings of that conference were published, I finally uploaded it to my homepage.
It pretty much summarizes my 2006 dissertation and why I got bored with 'new media', though I could not always escape its discourse. In retrospect, its last paragraph also seems to contain some accurate predictions for ChatGPT, Dall-E & Co.
#GenerativeAI#LargeLanguageModels rely a lot on the human to do the reasoning for it, and even then #BingChat (Creative) has problems with following the guidance. Notice I only specified the use of "unwieldy" and never required it to use "beard" or "weird" yet the #LLM got fixated on that instead.