Johannes Fähndrich

scientist, researcher for Strong AI, Pragmatics and Artificial Meaning

[Wilfried Honekamp @WHonekamp]

Tagungungsband der 7. Fachtagung Polizei-Informatik 2022 veröffentlicht.
@Rittelmeier @TGRuediger @BugSteffen @Datenverlust

[Wilfried Honekamp @WHonekamp]

Das Programm für die Fachtagung #Polizei-Informatik 2021 steht fest. Die Tagung wird am 20.04.2021 online von 09:00 bis 16:00 durchgeführt. Mehr unter https://t.co/qrMiQP1VQc @Rittelmeier @TGRuediger @BugSteffen @Datenverlust

[Denis Emelin 🤖 @emelin_denis]

Our #emnlp2020 paper (with @iatitov and @RicoSennrich) finds that NMT models overly rely on lexical biases for word sense disambiguation, rather than deep language understanding, which can be exploited for model-independent adversarial attacks: https://t.co/y4lYQeJ1MV.

[Dr. Roman Yampolskiy @romanyam]

It is Hard to tell AI what to do in an ambiguous language. https://t.co/rzlF6aNbNc

[Johannes Fähndrich @Datenverlust]

One step closer to usable electric mobility:

https://t.co/yEwLtAQG7W

[Johannes Fähndrich @Datenverlust]

https://t.co/QJIU3CnZcO

[the government man @me_irl]

My game walkthrough text generation model (now trained on over 100MB of game walkthroughs and FAQs) is making up fake Doom cheat codes none of which seem very useful

[Sebastian Ruder @seb_ruder]

Transfer learning is increasingly going multilingual with language-specific BERT models:

- 🇩🇪 German BERT https://t.co/M6Mo1drsgo
- 🇫🇷 CamemBERT https://t.co/flzuE9zNJN, FlauBERT https://t.co/xXlSeytUjj
- 🇮🇹 AlBERTo https://t.co/1nhchd2PG1
- 🇳🇱 RobBERT https://t.co/wR2Bln1Fyx

[Sebastian Ruder @seb_ruder]

This is a nice diagram by Zhengyan Zhang and @BakserWang that shows how many recent pretrained language models are connected. The GitHub repo contains a full list of relevant papers: https://t.co/uQNRGqMJAA

[Alexander Verbeek 🌍 @Alex_Verbeek]

🌎

And while those trolls sink to ever lower deplorable levels, shouting about child abuse, autism, or depression, I just post another impressive speech of @GretaThunberg.

Urging Congress to treat ‘this crisis like the existential emergency it is’

[Andrew M. Webb @AndrewM_Webb]

#ThermodynamicsThursday Simulating a system of gases, and using compressed image filesize as an ersatz measure of its thermodynamic entropy

[Greg Brockman @gdb]

A GPT-2 written essay was submitted to the Economist's youth essay contest.

One judge, who did not know the essay was written by an AI, gave this review: "It is strongly worded and backs up claims with evidence, but the idea is not incredibly original."

https://t.co/RbKrQvN8C0

[Graham Neubig @gneubig]

"Language models as knowledge bases?" they asked: https://t.co/O7BCcy0V6r
"A cat has four kidneys", replied GPT-2.

[MIT Technology Review @techreview]

This algorithm can tell if you got an AI to do your homework for you. https://t.co/rB8Uh4KDxk

[ruchowdh.bsky.social @ruchowdh]

h/t @lizjosullivan

Swiss researchers are able to de-identify individuals in 84% of the confidential legal cases in their study in less than an hour.

https://t.co/fHf5Q08vBT

[Johannes Fähndrich @Datenverlust]

Nice paper on Multi-Hop Reading Comprehention: https://t.co/PeSbFFkL2e
#NLP #NLU

[Sarah-Jayne Gratton @grattongirl]

Artificial Intelligence serves social development - https://t.co/jgFfI9HPn0

https://t.co/klcAw5CrLT

[Dean Anthony Gratton @grattonboy]

Top Artificial Intelligence Investments of July-August 2019 - Analytics Insight

https://t.co/LQ6mrlOCcU

[Johannes Fähndrich @Datenverlust]

Nice collection of games for RL: https://t.co/GCKwW8BfIT

[ruchowdh.bsky.social @ruchowdh]

And this is how the hyper-elite tech giants sell the rest of us on their mediocrity while retaining their wealth.

Now humans behind “AI” isn’t a bug, it’s a feature. This article literally frames people as fuel.