Tabea Wilke | @tabea@sfba.social (@TabWilke) header
Tabea Wilke | @tabea@sfba.social (@TabWilke) profile picture

Scaling privacy & Security. Threat model camp. Founder of Twincler. @acg_usa and @AJCGlobal young leaders alumna. She/her 🇩🇪🇺🇸 | Ⓜ️ @tabea@sfba.social

Joined January 2013
Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)Photo by Tabea Wilke | @tabea@sfba.social (@TabWilke)
Here is a list of other places where you can find me on the web 🌎: Mastodon: sfba.social/@tabea LinkedIn: linkedin.com/in/tabeawilke Instagram: instagram.com/tabeawilke Medium: tabea-wilke.medium.com/ Website: tabeawilke.com
1
1
If you're going to Stuttgart, be sure to wear some flowers in your hair. #cdupt26
86
Happening tomorrow (Wednesday)! Get access to new teaching materials, and hear from Trust and Safety Teaching Consortium members @MarianaOlaizol1, @InbalGoldberger, @Ina_Kamenova, @alexleavitt, @devikamm, @michaelrswenson, @amarashar, @TabWilke, and more!
Have you thought about teaching online safety, either to colleagues or students? Guess what: Over the past few months the Trust and Safety Teaching Consortium - a loosely-organized coalition of academic, industry and non-profit experts - has been creating teaching materials.
1
7
2
13
10,058
Have you thought about teaching online safety, either to colleagues or students? Guess what: Over the past few months the Trust and Safety Teaching Consortium - a loosely-organized coalition of academic, industry and non-profit experts - has been creating teaching materials.
1
1
3
6
13,521
It's spring in Berlin when you mind your business (ice cream) and a stranger gives you a flower.
3
2
302
“The other day” - Minimum two days ago, maximum six months ago “Just the other day” - Any day in the past fortnight
21
168
8
1,821
245,231
"Must not incite subversion of national sovereignty or the overturn of the socialist system, incite separatism, undermine national unity (...), as well as content that might disrupt the economic or social order." Overview by @ChinaLawTransl8. chinalawtranslate.com/en/ove…
1
1
150
"Legal, effective, ethical, safe, and otherwise trustworthy." reuters.com/technology/us-be…
1
1
1
140
Tabea Wilke | @[email protected] retweeted
This is what the world looks like to a self-driving car.
3
7
1
26
53,595
It was a pleasure meeting with the brave Ukrainian drone operators. We talked about the challenges they face on a daily basis, but they also wanted to talk about Wookiees & farm boys & droids, oh my! Show your support here: u24.gov.ua #SlavaUkraini 🇺🇦
Children of the Ukrainian military adore #StarWars. So, they asked @MarkHamill what his favorite film is from the saga. Listen to Mark’s answer to the Ukrainian defenders during their meeting. u24.gov.ua/dronation
2,814
2,040
1,279
19,913
7,890,015
Given another horrifying video is doing the rounds on social media its a good time to share this about looking after you're mental health while looking at that sort of imagery. The best advice is don't do it unless it's your job and there's a purpose to it bellingcat.com/resources/202…
76
793
118
2,992
851,608
Ambitious people need each other. Ambition breeds a particular kind of frustration, and the more ambitious you are, the larger it looms. Without like-minded peers, ambitious people become suffocated by life — chained by tall poppy syndrome and the dogmas of complacency. As social creatures, humans need to feel validated by their peers in order to go all in on who they are. We become like the people around us, so it’s hard to sustain ambition in a complacent environment. This is why actors move to Hollywood, musicians move to Nashville, and basically every podcaster/biohacker/Internet-writer type lives inside a twelve-mile radius in Austin, Texas. Moving in search of ambitious peers isn’t a new idea. It’s why Ramanujan, one of history’s greatest mathematicians, went to Cambridge. At home in India, even though he showed a divine aptitude for math, he flunked out of school and hid under a cot because his parents disapproved of his obsession with math. Though he taught himself number theory by working through problems in a borrowed textbook on his own, he knew that his genius was ultimately constrained by a lack of ambitious peers. And so, he wrote to Cambridge mathematician G.H. Hardy to ask for help getting his work published. Once they met, everything changed for Ramanujan. See, ambitious people don’t just need peers. They need mentors. They need people who will encourage them to pursue hard and meaningful projects. In college, I felt insane for being driven and obsessive. The anti-dedication environment drove me nuts. My ambition was only validated when I arrived in New York City and met people who exuded the kind of heart and hustle for which I’d always been called crazy. The people I interviewed on my North Star Podcast became some of my peers. @paulg, the founder of Y-Combinator and the low-key father of online writing says: “Ambitious people are rare, so if everyone is mixed together randomly, as they tend to be early in people’s lives, then the ambitious ones won’t have many ambitious peers. When you take people like this and put them together with other ambitious people, they bloom like dying plants given water.” Fight to find your kindred spirits, even if it means moving across the world like Ramanujan or staying in on Saturdays to write on the Internet. The more ambitious you are, the more consciously you’ll need to cultivate your social circle. Ask yourself: Who do I need to surround myself with?
80
352
69
1,588
420,090
This is a baby GPT with two tokens 0/1 and context length of 3, viewing it as a finite state markov chain. It was trained on the sequence "111101111011110" for 50 iterations. The parameters and the architecture of the Transformer modifies the probabilities on the arrows. E.g. we can see that: - state 101 deterministically transitions to 011 in the training data, so the probability of that transition becomes higher (79%). Not near 100% because we only did 50 steps of optimization. - state 111 goes to 111 and 110 with 50% probability each, which the model almost learns (45%, 55%). - states like 000 are never encountered during training, but have relatively sharp transition probabilities, e.g. 73% of going to 001. This is a consequence of inductive biases in the Transformer. One might imagine wanting this to be 50%, except in a real deployment almost every input sequence is unique, not present in the training data verbatim. Not really sure where I was going with this :D, I think it's interesting to train/study tiny GPTs because it becomes tractable to visualize and get an intuitive sense of the entire dynamical system. Play with here: colab.research.google.com/dr…
206
1,052
183
8,180
1,952,380
Retweet if you were part of one of those tiny first 4 bars. [source: buff.ly/3jWsfiC]
1,381
2,821
891
6,200
3,131,181
LLMs Are Better Than Human Data Annotators -GPT-3 was helpful but not better than humans (e.g. arxiv.org/abs/2108.13487) -GPT-3.5 is about on par w/ humans (e.g. arxiv.org/abs/2303.16854 w/ self-explanations) -GPT-4 is better than $25/hr humans (e.g. arxiv.org/abs/2304.03279)
17
307
44
1,448
418,366
🤖 Ich habe eine Datenschutz-Checkliste erstellt, die ich allen empfehle, die KI-Dienste direkt oder über Schnittstellen für berufliche Zwecke nutzen wollen: "KI & Datenschutz - Checkliste für den Einsatz künstlicher Intelligenz" datenschutz-generator.de/ki-… #ki #chatgpt #dsgvo
18
1
39
6,143
Because OpenAI just takes any data you enter - and you have no way of removing it - suddenly things happen like Samsung's confidential documents are now part of ChatGPT, and you can probably query them, even when you don't work at Samsung. The EU does not want this to happen.
10
82
32
466
84,862
We accidentally invented computers that can lie to us and we can't figure out how to make them stop
77
204
62
1,110
269,883
This is from @StanfordHAI's #AIIndex2023. The carbon emissions to run AI models is still insane. hai.stanford.edu/news/2023-s…
1
111