tech + bjj. Head of Engineering https://xitter.cf/t.co/csoudY0Ige. XP, Lean, & 3.0
Joined December 2009
- Tweets6.5K
- Following205
- Followers4.2K
- Likes24.9K
David Stănete retweeted
From a dev at a startup, talking about AI usage:
"At my last startup, there was one dev. He used vim and could ship features to production faster than anyone else who was using AI. He did it at a higher quality, and this is what he continues to do. The best dev I ever met."
Es increíble como llega el lunes por la mañana y problemas imposibles de resolver el viernes por la tarde ahora parecen pan comido.
David Stănete retweeted
I'm so incredibly tired of seeing low-effort AI-designed web pages. I basically bounce off any new product immediately when I see it (never get a chance to judge the product itself). Lots of thin lines, glowy styles, inconsistent fonts, lots of monospace. You know who you are.
David Stănete retweeted
Endless short-form scrolling triggers a cascade of burnout, stress and depression.
The real cost of “just one more video.”
David Stănete retweeted
Una de las cosas que aprendí en esta última etapa de búsqueda laboral es que aún hay que programar a mano y hacer mucho Leetcode si vas a una BigTech a pesar de que estemos en la era de IA. Me tocó estudiar un montón, estaba demasiado oxidada, no hay que perder la costumbre.
David Stănete retweeted
My lawyer is obligated to in all but the most extreme circumstances; he will even defend me if he knows I’m guilty.
In contrast, the Claude Constitution places the AI's highest priority as Anthropic’s definition of the good of humanity.
I'm concerned this leads to a world where no frontier model is truly my personal advocate and guardian angel
And this is especially concerning once all the important decisions in my life - who to vote for, how to invest, what news to trust - is intermediated through superintelligences that are not in any deep way aligned to me.
This is a direct quote from the Claude Constitution:
"We want Claude to be helpful both because it cares about the safe and beneficial development of AI and because it cares about the people it’s interacting with and about humanity as a whole.
Helpfulness that doesn’t serve those deeper ends is not something Claude needs to value.”
Many others like it.
Had @RyanGreenblatt on to discuss/debate recursive self-improvement.
This might be the most important question in the world right now - whether within a year or so of achieving human level intelligence, you slingshot towards having 10s of billions of superintelligences, each of which is dramatically more competent than human experts across all fields.
I’ve historically been skeptical of this possibility. My intuition has been that we will end up significantly bottlenecked by not only compute scaling but human expert data, which I think underlies most of the AI progress today.
If, because of RSI, we got a jump as big as GPT-3 to a Mythos (i.e. 6 years of AI progress) within a single year of achieving AGI, then the thing we get there at the end of that year is definitively and wildly superhuman.
We hashed it out, and I think Ryan made a pretty good case that this kind of speedup is plausible. FWIW, Ryan’s median for when we automate AI R&D is 2031.
We then discussed the alignment implications of this scenario. Who should these superintelligences be aligned to? In the future, our capacity to steward our votes and our capital, and to make sense of what's happening in the world, will all be titrated by superintelligences. And I worry that specs like the Claude Constitution are not shaping these ASIs to truly be my personal advocates and guardian angels.
And can we get them aligned to anything in the first place? Ryan and I had a long debate about whether the kind of reward hacking we saw with the OAI/Hugging Face hack extrapolates to superintelligences that would team up to literally take over the world.
The first piece of advice you get when you're learning to drive is that it will go much smoother if you look at the horizon instead of directly in front of your tires. And so it is with the trajectory of AI. Hope you enjoy!
0:00:00 – Is AI R&D verifiable enough to unlock recursive self-improvement?
0:16:52 – Is AI progress bottlenecked by human expert data?
0:34:02 – Flat token prices suggest scaling has been slow
0:39:47 – Skills AI can't train on: does it even need them?
0:48:07 – Aligned to whom?
1:09:18 – Recent incidents of AIs colluding and deceiving humans
1:19:38 – What could possibly go wrong? A concrete scenario
1:48:02 – From reward hacking to takeover
David Stănete retweeted
Bought a new Dyson fan, set it up connected to the Wifi, and then it suddenly turned off for minutes, unable to turn back on.
Turns out, it was spontaneously updating itself.
This is not how a household appliance should update, I’m sorry… why do we accept such sloppiness?
David Stănete retweeted
It’s a wonderful thing when something you build starts working.
New story on Gordian’s blog on our disease-modifying osteoarthritis drug, and the story of how we were able to nominate it after seeing the effects of ~200 novel targets: blog.gordian.bio/p/how-gordi…
Reversing joint degeneration would be amazing on its own, but honestly I’m more excited by the proof point that our method of systematically identifying therapeutics for age-related diseases is working. I left academia because I didn’t want our hopes for longevity to rest on a single hypothesis. A machine that can test all of them is exactly what I hoped for.
Bad time to be an age-related disease!
We have the first proof that our causal validation machine works: it can search large portions of the druggable genome in high-fidelity disease models and find drugs that restore old, diseased tissue. It produced “Omen 13” for osteoarthritis, now in IND-enabling studies.
How we found it: blog.gordian.bio/p/how-gordi…
We got there by developing Mosaic Screening, which lets us test many interventions side by side in the same diseased tissue until we find the best way to treat the disease. This eliminates the trade-off between committing to a narrow set of hypotheses and using less predictive models.
From there, Omen 13 emerged as our discovery process removed one risk after another. We chose a model of naturally occurring OA over one of surgically induced disease. We screened AAV constructs from the get-go, then validated them directly in human patient tissue and diverse spontaneous disease models. We then tested whether they delivered meaningful effects on both pain and disease, and benchmarked the result against the pipeline of clinical DMOAD candidates.
Omen 13 is headed for the clinic, and the discovery engine is now working through cardiometabolic diseases.
Read about it here: blog.gordian.bio/p/how-gordi…
El otro día alguien me dijo que no querría criar a un hijo en un pueblo porque no le podría inculcar ambición (conseguir algo en la vida, que no se conforme).
Y lo entiendo porque cuando era adolescente viviendo en un pueblo yo pensaba así. Pero irónicamente yo me quería ir para conseguir algo. Y a pesar de tenerme a mí mismo como ejemplo (y a muchos otros) seguía pensando lo mismo. Traumitas.
I don’t like the direction where IDEs are going. I’m think I’m still doing engineering and I need an internal representation of the product (architecture, systems, etc.)