CEO and co-founder of Tisane Labs. NLP for content moderation / lawtech / forensics
Singapore
Joined January 2015
- Tweets464
- Following190
- Followers79
- Likes4.6K
Vadim Berman retweeted
8. This is a 3200-year old attendance sheet found in Deir el-Medina, Egypt.
Reasons for workers being absent include "brewing beer", "embalming brother", "drinking with Khonsu" and "bitten by a scorpion".
Vadim Berman retweeted
PUTIN'S TROLLS: RU is increasingly turning to social media to sew disinformation and attack Ukraine's supporters. Cambridge University has put together a 5-minute game to show you the techniques trolls use disseminate fake news on any topic. goviralgame.com/en/play
Vadim Berman retweeted
The real-world is the real benchmark:
"... we should be looking at whether AI is generating real-world value instead of mindlessly running after fantastical ideas like AGI ... the proof is in the pudding."
ALT Screenshot from: https://futurism.com/microsoft-ceo-ai-generating-no-value
Vadim Berman retweeted
Here are some giveaway ChatGPT responses in research papers I collected, because I woke up too early this morning and thought it would be funny.
Vadim Berman retweeted
7 Signs Today’s Automatic Content Moderation is Broken
medium.com/tisanelabs/7-sign…
#contentmoderation #tisane #trustandsafety
Vadim Berman retweeted
7 Signs Today’s Automatic Content Moderation is Broken
medium.com/tisanelabs/7-sign…
Vadim Berman retweeted
This Turing Test nitter.cf/blueskykites/status/17…
Vadim Berman retweeted
Walkable urbanism is so underrated. Woke up, walked my daughters two blocks to pick up donuts & coffee, then walked to the the park one block in the other direction. Now sitting five blocks away watching them play by the lake. Zero cars involved. Everyone should have this luxury.
Vadim Berman retweeted
Linguists estimate that it takes about 1,000 years for a language to change enough that people at the beginning and end of the 1,000-year period can't at all understand each other. This holds with English—if you went back to the year 1024 AD and walked around England, you wouldn't understand what the hell anyone was saying because they'd be speaking Old English, which is really a very old version of German.
What's weird is that it also holds for the future. If you time machined to 3024 AD and (if somehow humans and countries are still around) walked around New York or London, you probably wouldn't be able to understand anything people said.
This process happens slowly enough that in any one lifetime, we don't really see it happening. But if you look closely, you see it. Think how confused a 1990 person would be if you asked them to help you find that lit meme about crypto podcasts everyone was posting on their insta stories. Or think about the terms and phrases your grandparents use that have gone out of style and no young person would ever use. Or even the slightly different ways old people and young people pronounce words. Now multiply this little bucket of differences by 20 and you have a totally different language.
Vadim Berman retweeted
The IMO is the hardest high school Math test. A lesser known sibling, the IOL (International Olympiad for Linguistics), starts tomorrow!
Students are asked to translate lesser-known languages purely using logic. 5 problems, 6hrs.
The problems seem absolutely impossible.
Vadim Berman retweeted
Wake up.
If a single bug can take down
airlines, banks, retailers, media outlets, and more, what on earth makes you think we are ready for AGI?
The world needs to up its software game massively. We need to invest in improving software reliability and methodology, not rushing out half-baked chatbots.
An unregulated AI industry is a recipe for disaster.
Vadim Berman retweeted
Artificial intelligence was asked to make a picture of Mother Teresa fighting against poverty and this is the result
Vadim Berman retweeted
The real threat to LLM chatbots isn't that adoption levels have stalled -- greater capabilities could reignite adoption. It's the fact that 1.5 years of massive investments haven't resulted in meaningfully better models, that would make a real difference in adoption potential.
Facebook still having odd issues like these despite optimistic reports about accuracy.
"Facebook's Algorithmic Erasure of History"
We wish to express our deep concern regarding the recent action of @facebook simultaneously flagging 21 posts published by the Auschwitz Memorial.
The posts, which serve as tributes to individual victims of Auschwitz, have been unjustly targeted by this platform's content moderation system, citing absurd reasons such as "Adult Nudity and Sexual Activity," "Bullying and Harassment," "Hate Speech," and "Violence Incitement."
The Auschwitz Memorial, an institution dedicated to preserving the memory of the victims of the German Nazi camp, has been met with unprecedented action at the hands of Facebook. The posts in question feature nothing but respectful remembrance and historical documentation, showcasing the faces and biographical information of those who suffered and were murdered in a place that symbolizes one of the darkest chapters of human history.
Despite our immediate appeal and request for review, only a fraction of the flagged posts have been reinstated.
Moreover, we also noticed that a post commemorating the tragic fate of Jewish children from an orphanage in Izieu from 6 April was summarily removed without the possibility of recourse.
The rationale provided by the Facebook system, claiming that these posts resemble content that violates Community Standards, is utterly confounding. For years, the Auschwitz Memorial has been sharing similar content without any issues, yet suddenly, 21 posts are simultaneously flagged. This raises serious doubts about the efficacy and accuracy of your content moderation technology.
This egregious incident not only undermines the important work of our institution. Such an act is unacceptable and offensive to the memory of the victims of Auschwitz that we try to preserve. It is hurtful to Survivors, descendants, and all those committed to preserving historical truth.
We demand an immediate and thorough review of this matter, along with a transparent explanation of why these posts were flagged in the first place.
@Meta @MetaNewsroom
