argonaut

67 ms·

691.
by argonaut 10y ago
Nope. That car just does lane keeping - it doesn't even do turns or lane changes. This is all stuff solved 20 years ago. And even then it only achieves 98% autonomy on lane keeping - this is a task that needs 100% accuracy. You should
692.
by argonaut 10y ago
What does that data give you? Raw sensor data + steering angle + acceleration/braking is not particularly useful in of itself. Nobody is trying to shove large amounts of data through a blackbox model and directly trying to predict car
693.
by argonaut 10y ago
It is a supervised learning problem, but that doesn't mean it will be solved by just shoving large amounts of data into a black box algorithm. This is not MobileEye's approach, and it almost certainly is not Tesla's approach.
694.
by argonaut 10y ago
I meant actual success, obviously. North Korea is culturally homogenous, and culturally similar to South Korea. There is no comparison to Afghanistan and Iraq. It also goes without saying that if America was willing to fully occupy Iraq f
695.
by argonaut 10y ago
The poster I was replying to is really the one who needs to prove that deep RL is super easy like they were claiming.
696.
by argonaut 10y ago
Errr... The US was fighting China in the Korean War. And the US military was not yet the undisputed global military superpower it is today. A land invasion of North Korea would almost certainly be successful, the casualties and cost is the
697.
by argonaut 10y ago
IANAL, but the first link does not support that conclusion. It is not made clear whether that employee was in fact committing theft and selling stolen goods to TMZ (the hotel threatens to press charges (what charges?) on the employee but di
698.
by argonaut 10y ago
Errr... No. Just no. Deep reinforcement learning is not some pixie dust that magically works for any problem with a reward function that you throw it at. It's astounding how commenters on HN think this is all "easy".
699.
by argonaut 10y ago
On the contrary, my experience with HN is that people are across-the-board far too accepting, credulous, and unskeptical of machine learning posts, especially given how most commenters have little ML experience and given how incredibly eas
700.
by argonaut 10y ago
Buying stolen property is a crime. TMZ would be committing a crime if they bought photos from hackers. Doing that would be the end of TMZ.
701.
by argonaut 10y ago
The company nixed from the crowd selected funding fellowship was offered the investment but directed it to charity instead.
702.
by argonaut 10y ago
95% of people are incentivized enough to not sell to hackers by the incentive of not becoming a criminal .
703.
by argonaut 10y ago
But someone isn't. That's the point. These bugs don't go for $10k on the black market.
704.
by argonaut 10y ago
This dichotomy isn't the actual choice, though. You'd hire the person with straight A/A-'s (read: nobody cares about perfect grades) from Stanford/Berkeley/MIT/Caltech who also worked directly with som
705.
by argonaut 10y ago
To add on to amluto's point, that's because your average elite college admit is cookie-cutter: great grades, great test scores, probably did a sport, did some extracurriculars with some leadership/responsibility. A cookie cut
706.
by argonaut 10y ago
Breaking the law requires intent or gross negligence, and the FBI said there was no evidence of either (gross negligence being a higher standard than "extreme carelessness").
707.
by argonaut 10y ago
Yes, but this is true of most presidents.
708.
by argonaut 10y ago
I'm pretty sure this entire affair will only be a tiny blip in Obama's historical legacy. Addendum: People on HN tend to vastly over-estimate the importance of privacy to the American populace. It's a very niche issue in poli
709.
by argonaut 10y ago
I think that perspective of exhaustion with contrarianism is part of the problem. There an excellent quote, which is "the amount of energy needed to refute bullshit is an order of magnitude bigger than to produce it." It is really
710.
by argonaut 10y ago
The courts are also the weakest branch of government. They have no power to enforce any of their decisions and the other branches of government possess numerous "nuclear" options to sideline the courts (court-packing, impeachment)
711.
by argonaut 10y ago
Well, the "enormous difference" is so enormous this really isn't a good comparison. The judiciary is the weakest branch of government. The judges can be impeached by Congress. Congress can arbitrarily expand the size of the S
712.
by argonaut 10y ago
Conditional random field. It's a model that comes from the probabilistic/statistical side of ML, which was the "hot" ML area before deep learning.
713.
by argonaut 10y ago
Yes. Investing in miracles is what VCs do. Investing in 50 companies and only having 5 of them succeed is what they do. Investing in 50 companies and expecting 45 of them to succeed is called being a small business bank.
714.
by argonaut 10y ago
Yes, and? It's exactly that kind of focus - focusing on the average outcome rather than the power law outlier outcomes that return your entire fund - that characterize the people who laughed at FB's valuation, as opposed to the VC
715.
by argonaut 10y ago
Less regulation = not breaking up a monopoly = higher prices.
716.
by argonaut 10y ago
You're dancing around, and haven't really defined, what it means to "learn Go." Sure, humans wouldn't become better at Go. But that's a limitation of the human brain (we're not good at mathematical memoriz
717.
by argonaut 10y ago
Where did you get that quote? It's not in the article.
718.
by argonaut 10y ago
In the abstract, breaking up monopolies is the canonical example.
719.
by argonaut 10y ago
Uber claims to be profitable in the US already, and it's their massive competition with Didi in China where they're burning cash ($1B/year). http://fortune.com/2016/02/18/uber-profitable-us&#x2F
720.
by argonaut 10y ago
Yes, and? Obviously people continued to investigate in subsequent years.

More ›