Associate Prof @uni_copenhagen & Senior Researcher @NSCR_Instituut: violence, bystanders, open science, proudly practicing positivist
Copenhagen, Denmark
Joined November 2012
- Tweets1.1K
- Following908
- Followers488
- Likes2.6K
Lasse S. Liebst retweeted
Some reflections coming out of a faculty meeting on AI in stats teaching today. With apologies for cannibalizing the thoughts of my smarter colleagues, a thread. 1/6
Lasse S. Liebst retweeted
When the Gino fraud hit a few days ago, I told my mom this is why I leave the room when she listens to Hidden Brain. Because shows like that are so much more likely to cover research like Gino’s. I said this w/o checking if Gino had been on Hidden Brain, but of course she has. 😢
Lasse S. Liebst retweeted
The replication crisis in the social sciences can end. That's what a new paper by @JProtzko found.
Four laboratories made discoveries with large samples, confirmatory tests, and preregistered methods and hypotheses.
The findings replicated.
Do good science and it's replicable.
Lasse S. Liebst retweeted
Whistleblower, David Charles Grusch, 36, a decorated former combat officer in Afghanistan, is a veteran of the National Geospatial-Intelligence Agency & the National Reconnaissance Office. He says U.S has retrieved intact & partially intact craft of non-human origin.
#UFO 🛸👽 – at Tennessee, USA
Lasse S. Liebst retweeted
One of my favorite papers in recent years included this diagram. It shows the impact of controlling for three different types of variables: confounders, colliders, and mediators.
With confounders, control is good. With the others, you ruin your result by controlling.
Lasse S. Liebst retweeted
Have you watched "The Prevalence of Replications in Psychology"?
Beth Clarke shares the result of their analysis.
What's our replication ratio? It's 0.2%. No, not 20%, not 2%, it's 1/5 of a percent. 0.2%. 1 in 500 gets replicated.
Video:
youtube.com/watch?v=teOvjoJb…
Lasse S. Liebst retweeted
Nominate me for chief editor if you'd like to see a journal fit for the future including:
-Registered Reports, Exploratory Reports, Verification Reports, Results-blind review, TOP Level 2+
-Close ties to preprints & preprint-based review, inc. @PeerCommunityIn & @PCI_RegReports
Please write APS at [email protected] and nominate Simine Vazire and Chris Chambers for Chief Editor of Psychological Science.
We should try and support open-science supportive editors for such roles.
Maybe, just maybe, if enough people write in...
Lasse S. Liebst retweeted
The @NSCR_Instituut video group started eight years ago with a crazy anthropological idea of becoming a fly on the wall as violence unfolded. Now we are 14 people doing video analysis of rule breaking, law enforcement, peers, criminal acts & peacekeeping nscr.nl/en/theme/nscr-video-…
Lasse S. Liebst retweeted
Every organization needs to plan for four possible AI futures:
1) This is it. LLMs never get better. (Current capabilities are still disruptive)
2) Slow gains in ability
3) Continued exponential improvements, but no AGI
4) AGI
I think people are way too fixated on 1 & 4 vs 2 & 3
Lasse S. Liebst retweeted
I recently taught a data science class here in Aarhus and went full chatGPT
I explicitly encouraged its use for both coding and using the API for natural language modelling tasks
The latter went very well, quality of results increased dramatically relative to pre-GPT times 1/
Lasse S. Liebst retweeted
Our new paper suggests that #GPT is a cheap, easy & effective tool for text analysis.
We find that GPT (3.5 & 4) is able to detect psychological constructs (eg emotions) WAY better than dictionary methods across 12 languages!
See paper for details: psyarxiv.com/sekf5/
Lasse S. Liebst retweeted
If you are supervising an international student who is trying to write in their non-native language - tell them how awesome they are. They often feel incompetent, frustrated and even stupid. In reality they're really smart - just doing something very complicated. Tell them.
Lasse S. Liebst retweeted
Hvis du på nogen måder arbejder med data - så skal du prøve @noteable_io's plugin til ChatGPT!
Det er Code Interpreter light!
Her har jeg bedt den om at loade noget data fra mit offentlige datasæt om dkpol fra 2019-2022.
Lasse S. Liebst retweeted
Several billion people just got free AI interns.
They are weird, somewhat alien interns that work infinitely fast and sometimes lie to make you happy, but interns nonetheless.
Here’s one approach for figuring out what they can (and can’t) do for you: oneusefulthing.org/p/on-boar…
Lasse S. Liebst retweeted
Brilliant when journalists set out to validate your research findings - here with undercover recordings - and bring you on new ideas. Bystanders intervene in all kinds of aggression, apparently also in racist intimidations. We need to know more about this! npo3.nl/hoe-is-nederland/16-…
Lasse S. Liebst retweeted
1. EditGPT
Transform ChatGPT into @Grammarly with EditGPT
→ Try now: editgpt.app
Lasse S. Liebst retweeted
🗞 ICYMI: #ChatGPT gets a settings UI revamp and teases upcoming 2FA support! Check out our latest blog post for all the details and learn how these updates bring ChatGPT closer to GDPR and security compliance 🔒🤖
Lasse S. Liebst retweeted
Gigantic Announcement for Language Models That Run on your CPU!💥📣
We are releasing:
- GPT4All-Snoozy: the strongest local LLM that runs on your private CPU hardware!
- The first local OS native LLM app verified by Apple !
Try it at: gpt4all.io
Lasse S. Liebst retweeted
GPT generated this academic paper from a dataset in 30 minutes.
It may not be a blockbuster but it tests something kind of interesting in a sophisticated way - in short, it may be publishable, somewhere.
The text is prone to hallucinations, but the data analysis less so. 1/2
Lasse S. Liebst retweeted
We just gave the world’s most powerful AI access to tools, and things are starting to get very strange.
Some experiments with ChatGPT Code Interpreter & Bing Sidebar, and what they might mean for the future of work. oneusefulthing.org/p/it-is-s…
