The AI Text Generator That's Too Dangerous to Make Public [www.wired.com]
OpenAI's new language bot is scary good at predicting text [thenextweb.com]
OpenAI unveils multitalented AI that writes, translates, and slanders [www.theverge.com]
OpenAI let us try its state-of-the-art NLP text generator [venturebeat.com]
OpenAI’s new language system writes really great fake news [www.vox.com]
AI can write disturbingly believable fake news [www.engadget.com]
Better Language Models and Their Implications [blog.openai.com]
An AI that writes convincing prose risks mass-producing fake news [www.technologyreview.com]
Roses are red, this is sublime: We fed OpenAI's latest chat bot a classic Reg headline [www.theregister.co.uk]
““We have the technology to totally fill Twitter, email, and the web up with reasonable-sounding, context-appropriate prose, which would drown out all other speech and be impossible to filter.” Scary thought from @jeremyphoward https://t.co/9jmBeoHOY8
— Joe Morrison (@mouthofmorrison) February 15, 2019
OpenAI has created a new AI language modeling system and it is *wild*
— James Vincent (@jjvincent) February 14, 2019
It was trained on a dataset of 8 million webpages, and without specialized tooling can not only generate text but translate, summarize, and answer trivia questions...
Details here: https://t.co/z4KqScGbb2
This is wild. OpenAI developed an news-writing AI that beat a bunch of records for reading/writing comprehension *without training*. https://t.co/5Upn76Iia3
— Kevin Roose (@kevinroose) February 14, 2019
In latest AI news: An algorithm re-wrote Kafka's "Metamorphosis" based on just one sentence from the original text & conventional deep-learning language modeling, and it's a riotous potboiler that should give us some faith in humans vs. robots. https://t.co/V0wozsEkMN pic.twitter.com/KSPtJHTYro
— Yuliya Komska (@ykomska) February 14, 2019
How great would it be to have a program capable of summarizing texts/books to a desired reading time?
— Alex Krüger (@Crypto_Macro) February 15, 2019
These guys are working on it: https://t.co/Q97pjvF9le #AI
wow we are making incredible advancements in the production of fake news https://t.co/fK528JryaY pic.twitter.com/xz4SGvWvJv
— Nitasha Tiku (@nitashatiku) February 14, 2019
"Tom Simonite does not keep it simple...And of course, you never really feel like the author has a clear vision of his subject."
— Tom Simonite (@tsimonite) February 14, 2019
I asked @openai's spooky new text generator to write a 1-star review about me and it didn't hold back. The future of spam?https://t.co/UwxIbN1kO2
This AI is one of the few genuinely breathtaking advances in technology I've seen in a long time: https://t.co/LmKLaJc7Or
— alex hern (@alexhern) February 14, 2019
OpenAI's new program rewrote the opening to 1984 and honestly I would read this novel https://t.co/uqoJf2yotK pic.twitter.com/gFptPlfdfC
— Dylan Matthews (@dylanmatt) February 14, 2019
New AI from @OpenAI can generate a news story from only 2 sentences of context.
— MIT CSAIL (@MIT_CSAIL) February 14, 2019
This one's about the discovery of talking unicorns.https://t.co/7i3XJRG6cc #ML #MachineLearning pic.twitter.com/hh8Oszlf0X
I confess that I can't even wrap my head fully around the possible implications of this, but it seems terrifying nonetheless. https://t.co/FaoiGtpK3q
— Shannon Proudfoot (@sproudfoot) February 14, 2019
Get a load of this.
— Quite Frankly (@PoliticalOrgy) February 15, 2019
The AI Text Generator That's Too Dangerous to Make Public | WIRED https://t.co/G62nH8vzmG
The AI Text Generator That's Too Dangerous to Make Public | WIRED https://t.co/89cX4lexKt
— Evan Kirstel at #HIMSS19 (@evankirstel) February 15, 2019
“It could be that someone who has malicious intent would be able to generate high-quality fake news,” Luan says. #AI #artificialintelligence #OpenAI https://t.co/drT9BxqBb0
— Paulina Maludy (@PaulinaMaludy) February 15, 2019
I love @OpenAI and their work in Language Modeling. The language their system generates is gorgeous. Have to admit, I love even more that Narrative Science stories are all grounded in fact. Reading well is good. Being true is even better. @narrativesci https://t.co/xbCJvPlUfj
— Kristian Hammond (@KJ_Hammond) February 15, 2019
This is a helluva thing. Love the fake 1-star review @tsimonite that bot is savage https://t.co/3atelUhkp3
— Shannon Vallor (@ShannonVallor) February 15, 2019
"Tom Simonite does not keep it simple...And of course, you never really feel like the author has a clear vision of his subject."
— Tom Simonite (@tsimonite) February 14, 2019
I asked @openai's spooky new text generator to write a 1-star review about me and it didn't hold back. The future of spam?https://t.co/UwxIbN1kO2
The #AI Text Generator That's Too Dangerous to Make Public | via @WIRED https://t.co/urCN3ZW16c
— Rob McCargow (@RobMcCargow) February 14, 2019
wow we are making incredible advancements in the production of fake news https://t.co/fK528JryaY pic.twitter.com/xz4SGvWvJv
— Nitasha Tiku (@nitashatiku) February 14, 2019
The release of an intentionally weak model by @OpenAI is great step in conversation about dual-use AI—but it leaves many open questions.
— Aviv Ovadya (@metaviv) February 15, 2019
Most interestingly—who should still be granted access to the better models?https://t.co/rB3b9DP56Y
OpenAI's new language bot is scary good at predicting text - perhaps too good https://t.co/2yWqHH7Ev7
— TNW (@thenextweb) February 15, 2019
How predictive text features are evolving through AI algorithms. Learning the nuances and subtleties of the human languages. This would prove to be groundbreaking in so many ways, exploitable too, on a large scale. https://t.co/RzWulZcpfw
— sighpad.m™️?❓ (@sighpad) February 15, 2019
It's great to hear from informed and thoughtful reporters like @jjvincent, who are working to put AI advances in context, and consider their wider implications.
— Jeremy Howard (@jeremyphoward) February 15, 2019
His latest piece on @OpenAI's new language model is very nicely done.https://t.co/HduGSutCLL pic.twitter.com/0g7X3Gh9vc
““We have the technology to totally fill Twitter, email, and the web up with reasonable-sounding, context-appropriate prose, which would drown out all other speech and be impossible to filter.” Scary thought from @jeremyphoward https://t.co/9jmBeoHOY8
— Joe Morrison (@mouthofmorrison) February 15, 2019
OpenAI has created a new AI language modeling system and it is *wild*
— James Vincent (@jjvincent) February 14, 2019
It was trained on a dataset of 8 million webpages, and without specialized tooling can not only generate text but translate, summarize, and answer trivia questions...
Details here: https://t.co/z4KqScGbb2
OpenAI’s new multitalented AI writes, translates, and slanders https://t.co/T5sQEoTkOD pic.twitter.com/vgB30hjKvZ
— The Verge (@verge) February 14, 2019
"Roses are red, violets are blue ..." my attempt to AI-generate a poem using @OpenAI's #GPT2 model https://t.co/4kFCRD7LAT pic.twitter.com/ZjkyOkDnMk
— Ignacio López-Francos (@iglpzfr) February 15, 2019
OpenAI’s new multitalented AI writes, translates, and slanders#AI #MachineLearning #futuretech #BigData #IoT #ArtificialIntelligence #deeplearning
— Reinard Mortlock (@MortlockReinard) February 15, 2019
https://t.co/rhZWn7YCN8 via @Verge
#OpenAI state-of-the-art #NLP text generator or sentences#ArtificialIntelligence #ML #MachineLearning #IoT #Blockchain #Robotics #Fintech #BigData #Technology #TechnologyTrends mt: @mikequindazzi #ai #deeplearning #iot #infographics @MikeQuindazzihttps://t.co/cDzkMG4LC4
— John C. Maxwell III (@maxjcm) February 15, 2019
OpenAI let us try its state-of-the-art NLP text generator https://t.co/21kRjdXKVE #AI pic.twitter.com/XFKnzKIphk
— Evan Kirstel at #HIMSS19 (@evankirstel) February 14, 2019
On GPT-2, the most convincing text-generating AI yet: https://t.co/yq0wZ04uZd
— Jacques Rousseau (@JacquesR) February 15, 2019
OpenAI scaled up some work they'd been doing with unsupervised learning for language AIs, and the results are incredible: https://t.co/xA6Fmk5d8A
— Kelsey Piper (@KelseyTuoc) February 14, 2019
So @KelseyTuoc used the new OpenAI tool to help write an article about AI risk and it did a pretty good job — and then added "Artificial intelligence is changing our lives in the fastest and most positive way possible." https://t.co/UC8E0aoSsf
— Dylan Matthews (@dylanmatt) February 14, 2019
I knew that @gdb was much smarter than me when we were both freshmen in college together, but even given that, it's crazy that only a decade later he and his team have taught a computer to do my job https://t.co/UC8E0aoSsf
— Dylan Matthews (@dylanmatt) February 14, 2019
AIs: They're just like us!
— Libby Nelson (@libbyanelson) February 14, 2019
No, really, way too much like us. https://t.co/SOIbE1UPjG pic.twitter.com/EAm4XK4IFu
“some great things are coming our way soon,” says the AI that helped write this article. https://t.co/EQNh9JdAem
— Vox (@voxdotcom) February 14, 2019
AI can write disturbingly believable fake news:
— StarTalk (@StarTalkRadio) February 14, 2019
Elon Musk's @OpenAI is keeping a tight lid on the technology. https://t.co/kntQbdDLsn
OpenAI's new language model can translate, summarize & answer questions without task-specific training: https://t.co/7i3XJRG6cc #ML #MachineLearning pic.twitter.com/Md41dSLRl2
— MIT CSAIL (@MIT_CSAIL) February 14, 2019
New AI from @OpenAI can generate a news story from only 2 sentences of context.
— MIT CSAIL (@MIT_CSAIL) February 14, 2019
This one's about the discovery of talking unicorns.https://t.co/7i3XJRG6cc #ML #MachineLearning pic.twitter.com/hh8Oszlf0X
We've trained an unsupervised language model that can generate coherent paragraphs and perform rudimentary reading comprehension, machine translation, question answering, and summarization — all without task-specific training: https://t.co/sY30aQM7hU pic.twitter.com/360bGgoea3
— OpenAI (@OpenAI) February 14, 2019
An OpenAI employee printed out this AI-written sample and posted it by the recycling bin: https://t.co/PT8CMSU2AR pic.twitter.com/PuXEzxL7Xd
— Greg Brockman (@gdb) February 14, 2019
Don't really know what to make of this, and would appreciate guidance from ppl who know NLP better. My *initial* response is "um ..." and second response "holy shit," but maybe I need to be more skeptical. https://t.co/j6uslrc6Zu
— Ted Underwood (@Ted_Underwood) February 14, 2019
. @elonmusk が支援する OpenAI が最新の言語モデル「GPT-2」を発表。しかし、悪意のある人たちが人工知能によるテキストジェネレータを使って欺瞞的な言葉や偏見に満ちた文章、そして不適切な情報を大量に生成することを恐れ、一部だけの情報のみを公開することに ?。https://t.co/uhsHmgs3DQ
— Sangmin @ChoimiraiSchool (@gijigae) February 15, 2019
良質の40GBのテキストを使って学習したTransfomerベースで10億パラメータからなる言語モデルはZero-shotで様々な専門ドメインの言語モデルのSoTAを更新し、条件づけ生成で様々な質問応答や翻訳も専門モデルほどでないが解ける。生成した物語はそこそこおもしろい https://t.co/cnq5WXIExR
— Daisuke Okanohara (@hillbig) February 15, 2019
8백만 개의 웹페이지를 기초로 학습한 AI가 자연어로 글을 생성하는 연구 결과에 관한 글https://t.co/OrXxQoRHIB pic.twitter.com/GJNZWFMTyl
— H. Kim (@metavital) February 15, 2019
Me three weeks ago: A billion trainable parameters is just bonkers! Why would you ever be training models that big?@OpenAI today: "GPT-2 is a large transformer-based language model with 1.5 billion parameters" (link: https://t.co/VMWP3OCRDj)
— Rachael Tatman (@rctatman) February 14, 2019
??? https://t.co/XHSUDH0O00
딥 페이크의 시대에 다시 한 번 경종을 울리는 일각수 개드립...이 아니고 자연어 글 생성 예제. 인간이 한두 문장을 던지고 기계가 나머지를 생성한 예제가 https://t.co/mx6LGSMyGH 에 올라와 있다. 영어를 안다면 예제를 읽고 크게 놀랄 이들이 많을 것이다.https://t.co/jcKvTFnsFo
— Kang Seonghoon (@senokay) February 15, 2019
A step forward in AI text-generation that also spells trouble. #artificialintelligence https://t.co/JmHwvdKeIa
— Zaphod (@Zaphod_stl) February 15, 2019
An AI that writes convincing prose risks mass-producing fake news - via @techreview https://t.co/tSgKThZ6jr
— 窪田 好男 (@YoshioKubota) February 15, 2019
企業や政府の意思決定でどの程度AIが用いられるかは興味深いところで、もし用いられるなら、ライバルのAIを解析したら次の一手が読めるかもしれないと思う人が出てくるかも、、、
An AI that writes convincing prose risks mass-producing fake news - via @techreview https://t.co/tSgKThZ6jr
— 窪田 好男 (@YoshioKubota) February 15, 2019
大学の卒論や修論もこの方法で作れてしまうよ考えられるので、プロセスを重視した論文指導が求められるし、そうなると少人数教育にならざるをえないと思われます、、、
Shoutout to @katyanna_q who fed the system a curveball, which I always like to see. As you might expect by now after seeing AlphaStar, OpenAI 5 etc. etc., if you drag the system away from its training data and into weirder territory, it begins to wobble. https://t.co/0d3I41df1r pic.twitter.com/gUUXFgiQ3z
— mike cook ??? (@mtrc) February 14, 2019
I also wrote about OpenAI's massive language model trained from millions of links scraped from Reddit. It's really interesting because it shows what just how many things models can learn just by analysing relationships in text ?https://t.co/JR31vMhMeG
— Katyanna Quach (@katyanna_q) February 14, 2019
New AI fake text generator GTP2 may be too dangerous to release, say creators...https://t.co/9En6lqFNsb
— Kopjes Kattenoppas (@Kattenkopjes) February 14, 2019
GTP2: "Zuckerberg? He looks like an idiot to me." :)https://t.co/Icpr1Nwwai#AI #ML #DL #NLP #deepfake #text #generator #OpenAI #ethics #innovation #breakthrough pic.twitter.com/WhfihcO67K