Douglas Gray
1.9K posts
Douglas Gray
@de3ug
Visual Search, Multimodal @amazon. Here for computer vision, ML, IR, & NLP research.
Portola Valley, CA Sumali Haziran 2009
1.9K Sinusundan474 Mga Tagasunod
@AmandaAskell The medical system is a powerful AI that will turn findings into profit
I think that’s more what people are acknowledging
English
The view that we shouldn't do more medical scans because incidental findings cause a lot of harm doesn't sit well with me. It seems like the issue it points to isn't the scan but the response to it. If you see something on a scan but have no other symptoms, you could ignore it.
English
@kunley_drukpa >"orang putih"
>not "mat salleh"
fake malayboo
Indonesia
Becoming a ‘Malay-boo’ Malaysia nerd because Japan is too mainstream
Zoso 🇺🇳(🇮🇶)@MesoAnglic
One of the most interesting countries to me right now is Malaysia. Despite its seemingly low nominal GDP per capita figure it is pretty close to having developed country living standards. It's also very spiritually English in a lot of ways.
English
@ProfKuang It’s the flip side of “nobody really writes papers anymore anyways”.
I actually think papers have gotten easier to read in the last 2 years as a result. Obviously written by AI >> obviously written by someone with poor English skills.
English
Over the last few months more and more people I talk to (and perhaps alarmingly, many more phd students) start saying that "nobody reads papers anymore anyway".
And the conviction is higher the "closer" the person is to "AI research".
The argument being that the reader simply tosses a paper into an LLM to ask for a summary of the key highlights anyway, so what's the point of polishing the 3rd sentence of the 4th paragraph.
Is this really the norm now? Yes, most papers are badly written, but I still greatly enjoy reading through the subtle subjective tonalities of a good paper, teasing out what the authors say versus what they *really* believe.
Honest question.
English
@antoniogm There's only 10 successful calendar year triple crown hikers.
en.wikipedia.org/wiki/Triple_Cr…
English
Number of people who've sailed solo around the world is about the number of people in this photo (or less).
Number of people who did it non-stop wouldn't fill a subway car. Far fewer than the number of humans who've been in space.
There are still frontiers out there.
sam buntz@SamBuntz
No image has ever better captured our sense of cultural exhaustion, our sense that it's all been done before, than this one:
English
@_fat_ugly_rat_ Nobody wants to live east of Chualar, but there is lots of development happening on the former Fort Ord. Lots of space to grow there too. Just need to address the traffic situation on 156/101.
English
Douglas Gray nag-retweet
all the labs will now race to get similarly banned by govt to avoid serving their frontier big chungus at a loss and subsequently write-off big chungus R&D costs as patriotism
English
Very pleased to hear Anthropic have walked back this policy simonwillison.net/2026/Jun/11/an…
SemiAnalysis@SemiAnalysis_
BREAKING NEWS: Anthropic's latest model will NOT help you if it thinks your ML research/ML engineering is interesting, and/or will secretly degrade its IQ so that the average engineer won't notice. We are already seeing Anthropic's latest model's moderation filters our GPU inference research and programming 😭
English
Buried in the 319 page system card for Fable 5:
We have hidden sophons in the model to quietly sabotage your LLM related research.
English
@adamac @dpetrou @karpathy Yes there's 2 separate pieces: 1) if apparently doing cyber sec stuff, downgrade to opus and warn; 2) if apparently trying to improve frontier LLMs, silently sabotage the work.
(That 2nd one reads like an insane conspiracy theory! But it's actually documented by Anthropic.)
English
@karpathy This is not a day for celebrating, Andrej.
It's a very dark and very sad day, and the damage may be impossible to undo.
English
@btrmasaladosa @henstepping I’ll take that over a wall of text that gets read off the slides verbatim any day.
English
@henstepping i swear the amount of tokens people burn on ai slop images for presentations is so frustrating. no one wants to see cartoon images in a 30 min ppt
English
@BellaBaddie__ San Pellegrino.
Pro tip: you can get cases delivered monthly via readyrefresh.com
English
Mfs out here quitting big labs to start their own right in the middle of compute shortage and recursive super intelligence
English
This is actually how you do performance management in a post-scarcity society.
Needs improvement? straight to the label mines.
Plot twist: top rating still has to label data 10% of the time, but your data gets more training weight.
Gergely Orosz@GergelyOrosz
@__apf__ Fulltime! Forcefully assigned. It's why so many devs at Meta are actively searching for new jobs. There's around 5,000 of them reassigned for FT data labeling
English
@GergelyOrosz @__apf__ Feels like the Japanese concept of “window sitting”, but with extra homework.
English
@__apf__ Fulltime! Forcefully assigned. It's why so many devs at Meta are actively searching for new jobs. There's around 5,000 of them reassigned for FT data labeling
English
Just learned:
Software engineers used to do manual data labeling at Scale AI while Alex Wang was CEO. After he left, new leadership joined, and were HORRIFIED to learn this. Stopped it ASAP
Now at Meta, software engineers are assigned manual data labeling... see the pattern?
English
@oscardumlao Is it a flow/diffusion model, or did they just combine photoshop with a computer use agent?
English
The level of detail possible in the Reve 2.0 model is insane — now it's possible to create designs with fine print and micrographics using straight image generation
English
@hanlonbt @billclerico They closed the train station to keep those folks out.
English
@billclerico As someone looking for a decent place in SF now, I very much wish Atherton remained the aspirational landing spot for the tech elite!
English
Nothing says “I was a big deal in a prior version of Silicon Valley” like a house in Atherton
English