Systems engineer Yacine argues independent builders can outperform well-funded teams on AI simulation and reinforcement learning bottlenecks

Many people's frame of reference, entire environment is the social structure around them rather than the actual physical properties of the world that God gave them. They will quickly give up in the face of any social signal. But if you act like this, you'll never be great

Anyways, here is me trivially passing the "Nvidia DeepMind" bench 90k SPS by 20x by simply texting my server while at an OB appointment with my pregnant wife. I literally pasted his tweet into codex verbatim

All the things I do are easily done by others, especially with an AI subscription. Just like whatever they've done with mujoco is also easily done by me, if I had 4 years. Nothing is difficult. What is difficult is having the courage to be different, to believe you know better

Consider this: the very smart people at DeepMind and Nvidia have created infrastructure that I could never have created. They are optimizing for building the infrastructure, and if they optimized for what I think should be optimized, they would surely succeed. But they haven't

There is something to this line of thinking, like how a beginner may not know the limitations so they never constrain them.

Everyone said a double backflip on a dirt bike was impossible then one guy did it and everyone is now, same for sub 2 hour marathon. every major progress that was “impossible” but eventually solved becomes common, it just takes one person to do it first

@yacineMTB uhm bro, how you ever gonna accomplish this without 7 layers of product managers and miles of corporate red tape, simply impossible

@yacineMTB It is not money that is a scarce resource; creativity, determination, and trust (including self-trust) are the scarce resources.

@yacineMTB they never worked in a big highly funded group and don't realize people there spend 90% of their time either filling out timesheets or feeling important

@0xBoesgaard None of this would be possible if it wasn't for the Nvidia / DeepMind teams working together. The people who orient the red tape so that this is possible, funneling the people in the right direction are just as responsible for making this work

@yacineMTB FLASH BACK. TO ONE OF MY FAVORITE TWEETS ON THIS PLATFORM.

@yacineMTB @BryanMcAnulty Crazy how often I see this losing attitude

@yacineMTB Thanks for sharing this.

I recently did something similar.

Beat a paper’s result by 2x just by prompting codex with a half baked idea.

@yacineMTB the reasonable man adapts himself to the world; the unreasonable one persists in trying to adapt the world to himself. therefore, all progress depends on the unreasonable man

- george bernard shaw

@yacineMTB I think these people haven’t worked in a big company to know how inefficient, bureaucratic and mind numbing things get. You can’t just do things without 79273648 permissions.

A small group can just cut through the noise and create without a care.

@yacineMTB They don't understand every great big oak was once a tiny acorn

@yacineMTB People often don’t take a deeper look at incentive structures too. Just because a big org is well funded doesn’t mean their incentive structure at a high level and then at the base employee level will make them actually do it. I’m with you 100% here.

@yacineMTB These companies are just made up of people, some of which are lazy and probably shouldn't be there. Of course we can do what they are doing

@yacineMTB totally agreed, there are few advantages individual has over labs and if only they could stop doubting and just do it

I spent half an hour explaining to a friend a technical solution that encapsulated a lot of problems (I was so hyped) and at the end he asked why hasn't big tech solved it or thought of it before me.

I took it as an offense, and said are you this simple minded that the thought of an individual solving a big problem gets questioned by default because he is an individual?

@yacineMTB amazing