Softmax | softmax.com | San Francisco (onsite) | Full-time
THE MOST IMPORTANT UNSOLVED PROBLEM IN AI IS THE MOST POWERFUL OF ALL CAPABILITIES: ALIGNMENT.
Hi, I'm Emmett. You may know me as the founder of Twitch (YC S05), and as the (extremely) interim CEO of OpenAI.
I believe the question of how AI systems learn to understand themselves and others is one of the most consequential technical problems of our time. If we get it right, the agents we build today will become our friends and collaborators tomorrow. And the infrastructure required to train agents that are good at working in teams and societies will be extraordinarily valuable.
Softmax is building large-scale multi-agent, multi-policy environments that will benchmark AI agents – both our own and others – on their capacity for theory of mind, adaptive goal setting, and teamwork. We’re looking for the engineers who will make that possible.
The work is unusually hard. The problems are ambiguous and challenging. High autonomy, high expectations. You'll work directly with me and the rest of our small but mighty team.
If this sounds exciting, I’m looking forward to meeting you.
E
———————————————————
Hiring: Software Engineer https://softmax.com/jobs/softwareengineer
Mention “HN Who’s Hiring” in the "How did you hear about this” field
———————————————————
Active Inference is kinda "a joint learning of the state representation with the search algorithm"...or at least, related to the idea. I like framing it as a joint learning problem.
I can’t quite reconcile this article with what I’ve seen before like this below…the methodology in the OP seems pretty questionable, so I trust this other article more.
Practicing gratitude for what you have is of course wise, and there are techniques that are helpful...gratitude journals, meditation, etc can be helpful.
However when I read your post it sounds to me like perhaps the issue is not a failure to be sufficiently grateful for what you have. It sounds like there is some desire, deep within you, that is going unmet. You have checked off everything from your list of Stuff You Thought You Wanted and yet something is still missing. It seems likely to me that that something is in fact missing.
You say "it's never enough", you "want more". But what exactly is it that you're wanting? What's the "enough" that it isn't? If you can look closely at that feeling with real curiosity and some gentleness...maybe there's a clue there as to what's going on. What's the reality that is going on (eg. "I'm often feeling deep failure and shame", a fact) vs. what's the narrative you're telling yourself ("Nothing is ever enough for me", a judgement you're making)?
That's not what the SAFE says. It's only on subsequent securities, not prior ones.
"If the Company issues any Subsequent Convertible Securities with terms more favorable than those of this Safe (including, without limitation, a valuation cap and/or discount) prior to termination of this Safe, the Company will promptly provide the Investor with written notice thereof, together with a copy of such Subsequent Convertible Securities (the “MFN Notice”) and, upon written request of the Investor, any additional information related to such Subsequent Convertible Securities as may be reasonably requested by the Investor."
Do you even know what the Thiel Fellowships are? Whatever you think of his politics, he literally gives away money to promising young people to help them realize their dreams...in exchange for no equity. Are you really saying that, because you dislike Thiel, you'd never do business with anyone who accepted free money from him?
Come on...before you post complaints about how someone else is literally giving away millions of dollars to the wrong promising young people, maybe actually read the list of projects?
"Zeno Power is developing small-scale nuclear power systems to provide clean and resilient power in austere environments." He's literally funding clean energy.
And even if this year's list didn't have that in the list, let's celebrate someone doing good instead of lobbing stones from the sidelines about how they didn't do it the exact way we would have.
This is silly. An investment is an expenditure of resources today, which pays off down the line. Dinner at a restaurant? Not an investment (unless you believe in investing in memories, which I believe stretches the metaphor slightly). An Xbox? Sure that’s an investment that pays dividends in entertainment over time. It’s a pretty short term since it’ll be obsolete in 5-7 years. Investing in a house w a neighborhood you love? Assuming you stay and love it, absolutely an investment and a long term one — maybe pays off for the rest of your life!
Not all returns pay out in cash. To think otherwise is to financialize your whole life, to know the value of money but nothing else.
It's obvious that there is some upper limit of companies per batch, past which it will be impossible for YC to source quality companies and advisors for batches. But why should we believe we've hit the point where the quality of advisors and peers decline as YC gets bigger? There are a lot more than 300 companies started per year...YC still only makes up a small fraction of the total. (And anecdotally just looking at this batch...I wouldn't care to bet that the quality of peers has decreased!)
That's not perfectly parallel as a belief. The parallel statement would be "If you believe that flat-Earth advocates don't actually care about whether the Earth is flat or not, they just want to prevent space exploration to keep us trapped on Earth"...that would be equivalent. And I can't speak for Scott, but I think he'd say if you believed that, your ability to reason about the Earth being flat is basically dead. It happens to be that you have the right answer (the Earth is round), but only by chance...if you were wrong, you wouldn't be able to be convinced to change your mind.
That quote is NOT Scott saying that pro-life advocates are right about abortion or wrong about abortion. That quote is Scott saying that once you start to believe that "the enemy" is a cackling evil villain, who is lying about their true motivations and secretly is only advocating for their preferred policies because they want to do evil things...you are doomed, at least in terms of seeking truth. Because once you believe that, there is nothing that "the enemy" can ever say or do that will change your mind. It's always just a tactic to advance their evil plot.
Of course, Scott could be wrong about this thesis. It's possible that maybe believing that pro-life conservatives don't care at all about fetuses doesn't actually indicate that you have a trapped prior that will prevent you from ever hearing new arguments on the abortion issue. But that is the argument that Scott is advancing, not that you can't form opinions about the truth or falsehood of someone's political views at all.
Paul doesn’t seem to have gotten in any trouble for saying it. What makes you think that someone else would? The problem only comes up when you start trying to label other people’s beliefs as dogma.
This is about examining your own beliefs to root out dogma, and avoiding pushing your dogmas into others. Not about being a self-perceived victim of some other group oppressing you.
I’m not enough of an astrophysicist to do this math myself, but what would this kind of burst look like from Earth? Observable w a telescope? W the naked eye? Enter a bomb shelter and save yourself?
If you could really start the company for $10k, I'd recommend self-funding out of savings. Save up $25k working (to give yourself a cushion) and give it a shot. If you have the skills to start a technology business, you definitely have the skills to earn $25k of working capital with your labor. People start small businesses that way all the time, and you can too.
THE MOST IMPORTANT UNSOLVED PROBLEM IN AI IS THE MOST POWERFUL OF ALL CAPABILITIES: ALIGNMENT.
Hi, I'm Emmett. You may know me as the founder of Twitch (YC S05), and as the (extremely) interim CEO of OpenAI.
I believe the question of how AI systems learn to understand themselves and others is one of the most consequential technical problems of our time. If we get it right, the agents we build today will become our friends and collaborators tomorrow. And the infrastructure required to train agents that are good at working in teams and societies will be extraordinarily valuable.
Softmax is building large-scale multi-agent, multi-policy environments that will benchmark AI agents – both our own and others – on their capacity for theory of mind, adaptive goal setting, and teamwork. We’re looking for the engineers who will make that possible.
The work is unusually hard. The problems are ambiguous and challenging. High autonomy, high expectations. You'll work directly with me and the rest of our small but mighty team.
If this sounds exciting, I’m looking forward to meeting you. E
——————————————————— Hiring: Software Engineer https://softmax.com/jobs/softwareengineer Mention “HN Who’s Hiring” in the "How did you hear about this” field ———————————————————