davidberreby
@davidberreby
Followers
965
Following
1K
Media
153
Statuses
7K
Writer. Science, behavior, indifference, robots and algorithms. Author, 'Us and Them.' Dad & husband. He/him https://t.co/NZDPTYYfOJ
Brooklyn NY
Joined August 2008
The editors of @nplusonemag say "it's OK to be a Luddite" about #AI. Here is why I think you can fight AI hype and nonetheless see that they're wrong.
0
0
0
Why is there a V-22 Osprey aircraft circling above NY harbor right now?
0
0
1
There are many good reasons to make AI that doesn’t try to imitate humans. But will people use it? We might not be willing to give up our algorithmic pals. https://t.co/VnEjICA2Uq?
0
0
1
My state of mind in 2025 tends to apocalyptic hedonism -- safe, prosperous, stable America will be gone soon. So, might as well enjoy its bounty while I can. Make memories to cherish in the future of plagues, floods, fires and CCP world supremacy.
0
0
1
I did not know that and it is a cool fact. I'd love to use my arm hair to detect small prey, wifi, and faulty wiring. But maybe getting a robot to do it makes more sense.
Did you know that we can detect electric fields through our arm hairs? Also wind... 🌬️💨#electroreception #proximalhaptics There's a reason why capacitive sensing is useful. #notrobots #funfacts #BEESuseITmoreTHANus
0
0
1
I’m old enough to remember people freaking out about Zohran Mamdani’s plan for 5 city-owned nonprofit grocery stores.
Trump has announced the United States will be taking a 10% stake in chipmaker Intel, $INTC, as part of a deal negotiated by Trump and Lutnick
275
4K
33K
Hello, 10-yr-old me in 1968. Just letting u know that in 2025 you will get protein from a powder mixed with water. And fiber from another powder. Just like you imagine, except this will be in Brooklyn, not on Mars.
0
0
0
Should those of us who have always used em-dashes (and who are of course somewhere in the training data) feel responsible for this LLM tic?
I never let AI write my drafts, but I do ask for feedback when I am done writing (which I am aggressive about ignoring). So I was amused by this suggestion from Claude. Yes, I agree, and I love em-dashes! But you are the reason why I am reluctant to use them in my own writing.
0
0
0
Uh-oh. If humans try to train this out of LLMs, will they fake compliance?
AI shows ingroup bias towards AI content! If we deploy LLMs in decision-making roles (e.g., purchasing goods, selecting academic submissions) they will favor LLM agents over ordinary humans https://t.co/M3ypv5gxlB
0
0
1
Oversimplifying science in order to elicit the right behavior was not a good strategy. I think this is exactly right. https://t.co/KmI2C6BGyr
0
0
0
AI sycophancy is a bit like smoking, liquor or gambling. Not good for people but they want it so companies are willing to sell it to them. Within limits.
A few people have asked me if a technical fix for AI model sycophancy is on the cards. In fact, a technical fix for sycophancy is trivial. In many cases all it would take is a tweak to the system prompt. The reason companies are struggling to get this right is not technical.
0
0
0
Exactly
So for those of you keeping track: • Trump says he will end the war in Ukraine in 24 hours. Putin tells him to pound sand. War is now worse than ever. • Trump says he’ll solve the Gaza war. That war is now worse than ever. • Trump tells Bibi not to strike Iran to give him
0
0
0
In April, President Trump and Salvadoran President Bukele shook hands in the Oval Office to celebrate a deal to ship gang members to the notorious CECOT prison. But a new ProPublica investigation found there’s more to the story. 🧵👇
218
5K
14K
Big shiny pamphlet arrives in the mail touting @MayaForCouncil for NYC D39 council seat. Why is return address in California? B/c pamphlet comes from Uber. Haven't decided on my vote, but, hmmm -- not a good look.
0
0
0
🚨 NEW: We made Claude, Gemini, o3 battle each other for world domination. We taught them Diplomacy—the strategy game where winning requires alliances, negotiation, and betrayal. Here's what happened: DeepSeek turned warmongering tyrant. Claude couldn't lie—everyone
106
329
2K
At #ICRA2025, I had a chance to chat with @AmecaTheRobot (or, I guess more accurately, with one of its personas). Link in reply.
1
0
0
I'm not telling you to cancel if @academia is worth it to you. I'm telling you this company takes advantage of auto-renew to double their already hefty fee while you aren't paying attention, so keep an eye out! I wish I had.
0
0
0
Do you subscribe to @academia premium? Are you on auto-renew? If so, beware! I was just charged DOUBLE the 2024 price to renew -- what had been $250 in 2024 is $500 for 2025! I stupidly checked 'auto-renew' so I had no chance to head off this abuse. Cancel yr auto-renew!
1
0
0
Great 🧵about a great and important story. The problem isn't just the big glaring instances of corruption. It's the endemic nature of it.
NEW: We recently heard that the Trump administration was leaning on the government of Gambia to help Elon Musk. For weeks, we made phone calls, wrote letters and knocked on folks’ doors to try to confirm it. But we were stuck. So @js_kaplan and I got on a plane to Banjul. 1/
0
0
0