
prev created @aisecurityinst, AI Safety Summit, UK AI Research Resource, EU AI Code of Practice.
San Francisco, California
Joined May 2012
- Tweets2.1K
- Following2.5K
- Followers2.5K
- Likes8.3K
Nitarshan retweeted
We decided to provide more transparency into pretraining progress at OpenAI for the sake of moving toward shared safety bars. In “Perplexity and Intelligence” I argue for common limits based on perplexity and other measurements, with progress beyond them gated on our ability to understand AI behavior.
openai.com/index/perplexity-…
Nitarshan retweeted
Me, looking up from the petty grudge matches, corruption and dysfunction of academia just in time to see the Riemann hypothesis solved
Replying to @mosslines
Glad to see there is a machine better than those who think they’re the best :)
Nitarshan retweeted
Between this and The Altruists, I feel like Hollywood does not understand the Bay, does not want to understand the Bay, and will pump out generic 'nerds suck' movies regardless of what complex villainy and hubris and ideological, catastrophic recklessness they ostensibly depict
Nitarshan retweeted
I hate how, whenever an AI resolves a millennium problem, A human will claim they did it first. The media then amplifies the claim and states [AI company] stole their work, leading normies to doubt that AI can genuinely solve hard math/science problems. If you possess a proof for a millennium problem, publish it immediately; otherwise, I prefer not to hear about it when the next one is solved.
Nitarshan retweeted
Terence Tao on Numberphile said he believes with the help of AI, we could solve the reimann hypothesis before 2050.
It took 35 days
Nitarshan retweeted
I covered Myanmar's slaughter of the Rohingya, and talked to the mother of a baby thrown onto a bonfire, to a girl locked in a hut that was set on fire. So it is heartening to see the Int'l Court of Justice rule that this was genocide. Shout-out to Gambia for bringing the case.
Nitarshan retweeted
fuck you, and fuck everyone with a fund logo in their bio pretending this is innovation
i'm a recovering gambling addict. here's what you don't put in the pitch deck:
people with gambling disorder die by suicide at 15x the rate of everyone else. for men 20–49 it's 19x. about 1 in 5 attempt it, the highest rate of any addiction
kalshi lets 18 year olds bet on sports in states where the legal age is 21. addicts who self-excluded from draftkings and fanduel are relapsing on your app because "federally regulated" means none of the state protections apply
that's not the "longer, more difficult path." that's the path around every guardrail that was built for people suffering from this addiction
you're not backing a prediction market. you're backing a sportsbook with better lawyers, and you're getting rich off people who will lose their savings, their families, and some of them their lives
Nitarshan retweeted
Talked to @nytimes about handoverism: the idea of handing over control to an AI God. Mentioned two things that I find striking:
1. With AI, there's an inversion of Silicon Valley's founding myth: for the first time, the stated goal of tech is to reduce human agency. This is very different from basically all tech that came before.
2. That there's a feeling of inevitability and helplessness, among those who create AI. The reporter who wrote the piece asked me if people here realize how much backlash there is against tech across the country. I said some here want to be stopped by an external force, but they feel they can't stop on their own.
Nitarshan retweeted
the defining psychological condition of san francisco is having already won the lottery and still being haunted by all the other lotteries you didn’t win
Nitarshan retweeted
Replying to @hcndashwood
Not AGI pilled enough; everything is downstream of model intelligence in the long-run. Trust the exponential.
No models; no party.
Nitarshan retweeted
My coworker @thlarsen makes a good point: While AI jaggedness is a real thing, modern AI is a lot less jagged than most people expected. A LOT less, unless I'm misremembering conversations a decade ago. Like, people used to expect a giant ecosystem of different AI services, rather than just a few flagship models that do everything.
Nitarshan retweeted
When I first heard that millions of adults were taking a week off for the release of GTA VI, my first reaction was the same as yours: Really? For a video game? Grow up.
Then I remembered May, 1999.
Grown men in bathrobes camped out on sidewalks for weeks with plastic lightsabers for Star Wars: Episode I. Offices emptied across America so millions could stand in line just to buy back a piece of their childhood.
Every generation looks at the next one’s obsession and usually mocks it. Your mom said it about rock and roll. Your dad said it about Star Wars. Now we’re saying it about video games. And every time, the older generation completely misses the point.
I barely remember the plot of The Phantom Menace. Honestly… I’ve spent decades trying to forget it. What I remember is who I stood in line with and who I sat next to in the theater. The thing you waited for fades, but the people you shared it with never do.
So when your son, your grandson, or your nephew disappears into the basement tomorrow, you have a choice.
You can sneer at it, or you can walk in.
You’re not going to understand the controls. You’re probably going to drive straight into the ocean. And yes, you might commit some of the most heinous fictional crimes imaginable… maybe some grand theft auto.
But you’ll be there.
Because twenty years from now, when he’s 50 and the next one finally comes out, he won’t really remember what happened in that game. He’ll remember the dad or granddad who came downstairs and picked up the controller.
Nitarshan retweeted
I recently became aware of how confused even many academics are about Pangram. Please, please, look at the audits that have been done before you deliberately sow doubt.
(link in thread)
Nitarshan retweeted
pause your cyberattack doomscrolling to confront the looming threat of philanthropy
Who Pays for What We Know
When economic change outpaces our understanding, those profiting finance the research that poses the questions we ask, steers the policies we consider, and shapes the rules that govern their power.
Read our new article by @poastmilque (link below):
Nitarshan retweeted
ultimately it's just very difficult to beat a philosophical/ideological/movement that can inspire Jane Street traders or Google engineers to quit their high-paying jobs in California to work with a bunch of relatively clownish coworkers on the Hill for $60k/yr in 84° DC summers
Nitarshan retweeted
I don’t think you should respond to questions about your credibility by telling your interlocutor not to ask the question, with the implied point being that the underlying nobility of your quest is so unimpeachable that your credibility is without question.
Nitarshan retweeted
If your AI characters sound fake, the fix is almost always in how you write and DIRECT the dialogue in your prompt.
I recently started on the dialogue-heavy scenes for my AI feature film about Beethoven, and too many of the lines kept coming out wrong.
They’d sound too theatrical, too monotone, or too much like a practice read-through between actors.
(all this despite the fact that Seedance 2.5 is fantastic when it comes to realism.)
So through a whole lot of trial and error, I've found that these 5 things made the biggest impact on the quality of the dialogue delivery in my generations:
1. Paying close attention to the written punctuation used.
2. Breaking long lines into shorter ones, with intentional pauses.
3. Using bodily actions as punctuation (this one gave the biggest improvement BY FAR! In the attached clip, note the sniffles in between Beethoven's lines. That alone made the delivery feel much more natural.)
4. Using "anchor gestures" (he lifts his finger as if to say "one sec")
5. Just split your generations in two (stupid simple, but it really makes a difference.)
So if your characters keep sounding like they're reading from a teleprompter, give this a try!










































