Berlin!

@Riedl you mean something like this :)? https://arxiv.org/abs/2108.04324

I have a really bad research habit.

1. Decide to do something fun in storytelling AI

2. It’s really hard, so we invent stuff to make it work

3. That something is probably a general solution to a wide range of things but I don’t want to go get a bunch of data on boring tasks

4. But we write a paper that is very narrow about storytelling AI

5. A few years later someone does something similar in a more general domain

6. I shake my fist and yell into the void that I did it first

7. Goto 1

📢 SUBMISSION DEADLINE APPROACHING on July 21 !

The 6th VISxAI workshop is happening this year in October as virtual workshop. Consider submitting your interactive explainables. #visualization #NLProc #XAI

My latest dispatch from the AI boom for the Washington Post looks at a billionaire-backed movement to push concerns about "existential risks" around AI from the fringes of tech culture into the mainstream. Recently tech philanthropists have helped fund about 20 student-led “AI Safety” clubs at elite colleges like Stanford, Harvard, MIT, NYU, Columbia as part of an effort to recruit idealistic talent to focus their time on the speculative risk that AI could kill us all.
https://www.washingtonpost.com/technology/2023/07/05/ai-apocalypse-college-students

Note to any followers considering switching to bluesky, Dorsey has endorsed RFKjr which, for me, is a good reason not to be on his new platform. Having been Jacked once I have no desire to repeat the experience. https://thehill.com/homenews/campaign/4034584-ex-twitter-leader-jack-dorsey-endorses-rfk-jr-for-president/

@elm woohoo — congratulations ! That sounds so cool!

Sweet home Scandinavia! 🇸🇪🇩🇰🇳🇴After 15 wonderful years in the US (6 at Purdue and 9 at UMD), my family and I are moving to Aarhus, Denmark 🇩🇰 this summer. I have accepted a position as professor of Computer Science @csaudk. This 40M DKK (~$6M) will help me jumpstart my research! I’m in Copenhagen now to receive the center award together with 10 other new Villum investigators. https://cs.au.dk/news-events/news/show-news/artikel/new-villum-investigator-at-the-department

In today's Washington Post, @geoffreyfowler tests Turnitin's AI Writing Detector and gets a false positive in only 16 samples of writing.

It was my pleasure to talk to him about Turnitin's accuracy claims, which mirror those of AI proctoring companies.

Paywall-free link: https://wapo.st/430WBpm

🎉Tata - here comes Prompt Tester, my newest
side project to help you iterate through text prompt
ideas by providing rich feedback or teach prompting. It pre-ships with some examples (see video). Enjoy 🌻

Tool: https://prompt-tester.vizhub.ai
Video: https://youtu.be/2xIzy0rJ5Ak

#NLProc #AI #HCAI

@dr_pi thank you for having me - was fun :)

@Donnagleize there has never been an access limitation (paywall) for GLTR. The backend might have crashed before the refactoring- should be limited now. Be aware that both tools might not give you a right answer.

@erikjacobs

@erikjacobs Well, both tools test "only" how unsurprising the text is to a given model. That can be an indicator but if the test model is significant different from the original model, it can easily be fooled. Also, if your text is similar to training data. Try adding the first couple of sentences of the US constitution and you will find that both tools think its AI created :)

Today, 10 years ago, I arrived in the USA to start a new and crazy life adventure - as the only one of my family moving here. But thanks to great mentors** and amazing friends, this decade was truly amazing! Thank you all. 🇺🇸🇩🇪

** @ebertini, Hanspeter Pfister, Mauro Martino

It's still live 🎉- despite a big increase in usage :). Our 3yo tool to help distinguish if text was written by GPT-2 or humans. Does it help against ChatGPT? Most likely only as much as GPTZero - both tools use a similar model for testing. #chatgpt

http://gltr.io/

What if writing felt more like sculpting clay?

Re-built an experiment from 1.5 years ago (!) and wow! text summarization is so much better these days.

Hoping to have something to share with you this week 🤞

If you plan a relaxed stay @Lufthansa lounge in Munich (MUC) you are in for a surprise. Over-Crowded, extremely limited food, and dirty bathrooms. Well below par 👎👎👎

I'm looking forward to it!!
---
RT @SAILhealth
At #SAIL2023 (Puerto Rico, May 9-12), Dr. Narges Razavian (@narges_razavian) will discuss inpatient (COVID-19) and outpatient (Dementia) detection using millions of EHR from @nyulangone.

Submit your abstract before Jan. 20! ⏱
https://sail.health/call_for_papers.html
https://twitter.com/SAILhealth/status/1608487205365743616