rytill

33 ms·

361.
by rytill 7y ago
That makes sense. But do you see why it might be conflicting to some people if they have the chance to do, say, 40 units of good for 8 pleasure instead of 10 for 10? Where do you draw the line?
362.
by rytill 7y ago
That’s crazy that you don’t know anyone like that. It’s a common thing for people to take very seriously.
363.
by rytill 7y ago
351 117 3141?
364.
by rytill 7y ago
That’s great. Thank you for sharing. Seems a rather cryptic way to go about things with your health on the line.
365.
by rytill 8y ago
Really? Sell me on this one. Keras is great, but socially he leads people to be careless about AI safety, which is hugely negative.
366.
by rytill 8y ago
It would be cool to run a machine learning model server written in Python and start a connection to it with JS, all in a tab.
367.
by rytill 8y ago
No, that would be awesome.
368.
by rytill 8y ago
What makes you so confident that you’d say “anyone who’s done a modicum of AI research” would come to the same conclusion as you? Also, do you believe AGI is currently more a compute/hardware problem, or an algorithmic problem?
369.
by rytill 8y ago
It's not totally a heuristic, though it's intractable to use it the real way. Which theory has a shorter program that generates it?
370.
by rytill 8y ago
That’s a really awesome idea. Also, maybe a few of the performance boosts in the final combination are mutualistic, but if measured independently would not make a difference.
371.
by rytill 8y ago
Actually wouldn't be terrible if the comments appeared in a more orderly fashion. Right now, they spawn randomly and can even overlap which makes it actually unreadable.
372.
by rytill 8y ago
This is currently very prolific and successful. You don’t have to imagine anything. Latest popular paper on the specific application of evolution you’re talking about: https://arxiv.org/abs/1808.00193
373.
by rytill 8y ago
You don’t see the catch-22? Or do you believe that no startups require a staff of more than a few cofounders to build their product?
374.
by rytill 8y ago
What kind of remote operation/visualization interfaces do you have for deployed systems? Does your company receive the alerts/visuals/notifications or do the farmers that buy from you receive these? Do you want farmers to eve
375.
by rytill 8y ago
"Medium AI" is definitely useful, as long as you have the correct interfacing and apply it to the correct problems. Obviously you can't just slap it onto an AI-complete problem and have medium AI perform well enough to ship.
376.
by rytill 8y ago
Spacetime is pretty slippery already. At least, I haven't been able to stop moving through it yet.
377.
by rytill 8y ago
Normalization of a layer’s outputs kind of does this already.
378.
by rytill 8y ago
> one that we could turn on and off as we pleased This is a research problem. Your phrasing makes it sound like you think “Singularitarians” choose to believe that rogue AIs are a potential risk, rather than deriving that knowledge.
379.
by rytill 8y ago
Perhaps interesting: Shannon's original information theory article is titled 'A Mathematical Theory of Communication', and when later released in book form was renamed 'The Mathematical Theory of Communication', due
380.
by rytill 8y ago
If you look at it from a utility perspective and discount the network effect, at a certain point, it becomes economically incentivized sharding. "We accept any of the 2021 or prior World Bank approved ETH forks." Your multi-wallet
381.
by rytill 8y ago
I see what you mean about not being able to sell back into a different currency profitably. I need to think about this a little more myself. On a slightly different note, another relatively minor but important thing to take into account whe
382.
by rytill 8y ago
> ETH executes all VISA payments > ETH Value: $0.088 (There are 100M ETH) Correct me if I made a mistake, but it seems like this is not a stable market. What's to stop me from buying a large chunk of the supply and sitting on it?
383.
by rytill 9y ago
Become more creative.
384.
by rytill 9y ago
What’s the latency?
385.
by rytill 9y ago
No, the key requirement which makes it difficult to use on real-world tasks is that you must be able to do a forward rollout of your environment in your decision-making process.
386.
by rytill 9y ago
The AlphaZero algorithm (monte carlo tree search with value estimator trained by reinforcement learning) works on any environment you can simulate during play time, single player or not.
387.
by rytill 9y ago
Thank you. That makes a lot of sense.
388.
by rytill 9y ago
I would argue the primary purpose is to abstract away low-level and repetitive actions while retaining the ability to tell a computer what to do. A fully visual programming environment would have more in common with standard programming lan
389.
by rytill 9y ago
Can you elaborate on “leveraging syntax”? Like commenting on the structure of their argument?
390.
by rytill 9y ago
Compare it to something else then.

More ›