emcq

26 ms·

181.
by emcq 11y ago
In general it's an abstract concept like Kolmogrov complexity but there are some practical approaches. People often try to intuit the intrinsic dimensionality of a dataset by using techniques like looking at singular values above some
182.
by emcq 11y ago
The qualifier there is "intrinsic" dimensionality. There is smoothness to the image such that there is a lower dimensional embedding than 1m dimensions for a 1 megapixel image, particularly with any applied setting. That said ther
183.
by emcq 11y ago
FYI asking people what they do outside work is often a big HR no-no because it can easily fall into illegal but accidental discriminatory questions that are unrelated to the job which can be unnecessary PR and legal risks for your company.
184.
by emcq 11y ago
Many selective companies have many phone screens and coding projects, plus hours of on site interviewing. Google's process notoriously takes months. Some companies interviews likely take 10+ hours all in all. That's more like 3rd
185.
by emcq 11y ago
If you look at the graduates of the top bootcamps, these are folks who studied music theory at Harvard, physics at Yale, math at Stanford, EE at UMich, mechanical engineering at UC Berkeley, etc. These are very smart folks who have proven t
186.
by emcq 11y ago
That sounds great! That should happen regardless. Put another way what does the investor gain putting money in a company that stays private?
187.
by emcq 11y ago
What is a fair compensation package for early hires if companies continue to keep stocks private and less liquid?
188.
by emcq 11y ago
Unfortunately I'm not familiar with a single resource for this, which probably contributes to developers being unfamiliar. There is a lot of domain specific knowledge depending upon if you're working with graphics, distributed com
189.
by emcq 11y ago
My favorite is when you jump on a project and do a simple estimation of compute throughput for the highly complex distributed system, its something like hundreds of kilobytes per second. You could literally copy all the files to one compute
190.
by emcq 11y ago
This result came out of a large team that Google paid a lot of money for. Google Brain has been kicking ass internally solving really hard problems like realtime translation or ranking. Most major companies from Toyota to Baidu to IBM have
191.
by emcq 11y ago
What I find confusing is that he says compensation is crazy at companies like Google, where employees can sell their equity on a public market with an expected value significantly above zero. Does this mean they are fundraising to pay their
192.
by emcq 11y ago
Valve is indisputably an amazing company with kickass products! But if you're a mainstream consumer, perhaps someone voting for Trump, you've never heard of them. And like the name "Trump", Sony has turned the name &quot
193.
by emcq 11y ago
That's fair. If you ask people outside the VR bubble if they would trust Facebook, some company called Valve, or Sony PlayStation to make a quality VR entertainment device most would put their money on PlayStation. Not only is the cos
194.
by emcq 11y ago
The absolute price difference is more important here than the multiple. Spending roughly $1k more for entertainment is certainly within the budget of many folks, versus $60k+ for a Tesla over a new Honda fit. But you're right that the
195.
by emcq 11y ago
Seems like it's only one 1920x1080 screen, so half that resolution per eye at 120hz. That's going to be some chunky pixels. I'd rather have double the resolution at 90hz with a Vive or Rift :)
196.
by emcq 11y ago
I don't think comparing win loss distributions is particularly insightful. By a single machine winning many games relative to the distributed version, really it's just saying that the value/policy network is more important th
197.
by emcq 11y ago
Interesting point, but which do you think is a more reasonable prior: that intelligence is a deterministic process similar other's we've encountered or that there is something uniquely different about intelligence than other physi
198.
by emcq 11y ago
I disagree. In engineering we've shown great advancements not by exactly reproducing biology from cases like transportation (wheel vs legs) or flight (engine and wing vs flapping feathers). We can't even mass produce synthetic mus
199.
by emcq 11y ago
This is simply not true, and there are many tasks the architecture would fail with and the networks presented are architecturally different from the highest performing vision networks like GoogLeNet. The techniques behind AlphaGo have no me
200.
by emcq 11y ago
Doubtful; I don't think comparing such small W/L distributions will be illustrative. On the other hand, the Nature paper shows the single 8 GPU machine performs similar to the 64 GPU cluster, but the larger clusters perform a comf
201.
by emcq 11y ago
I'm hoping someone creates adversarial formations for AlphaGo if Google ever releases their model :) One thing I've been pondering is if many adversarial samples exist. The board is rather low dimensional (19 x 19) and discrete. W
202.
by emcq 11y ago
If the value/policy model is predictive with a dataset containing only amateur games, but fails to generalize to unseen data with professional games, that seems like a case of overfitting to a dataset only containing amateur games. In
203.
by emcq 11y ago
This is an interesting question - if Lee Sedol simply replays the game exactly, does he repeat a win? I think the answer would be most likely not - the monte carlo tree search is randomized so AlphaGo's responses to Sedol may not be ex
204.
by emcq 11y ago
I dont think they were bugs in a traditional sense. I think AlphaGo picked moves to try and maximize the probability of winning, and at some point that was only by the opponent making a suboptimal response. I remember reading somewhere most
205.
by emcq 11y ago
That was really cool! It seemed after the brilliant play in the middle the most probable moves for winning required Lee Sedol to make impossibly bad mistakes for a professional, which would be a prior that AlphaGo doesn't incorporate.
206.
by emcq 11y ago
Extrapolating Deep Blue's 11GFLOPs supercomputer to today with Moore's law would be equivalent to a 70TFLOPs cluster. AlphaGo is using 1+PFLOPs of compute. While they likely aren't actually achieving that compute throughput,
207.
by emcq 11y ago
I'm not comparing brain wetware to hardware. The parent's post was interested in how we achieved such great AI go performance today that was supposed to take 10 years. If you look at the components that fueled this, the performanc
208.
by emcq 11y ago
Part of this timing jump is Google throwing hardware at the problem with a large 280 GPU + 1920 CPU cluster. I would venture this is almost 100x bigger than most of the Go AI hardware we've seen to date. The nature paper suggests witho
209.
by emcq 11y ago
You're absolutely right. The power budget for these cars is more like 30 watts. A Tesla driving is around 240 watts per km or 14kw per hour at 60kmph. If you're okay reducing your range but half you could get a dozen beefy GPUs if
210.
by emcq 11y ago
This 1000 times. Extrapolating Deep Blue's 11GFlop supercomputer to today with Moore's law would be a 70TFlop cluster. AlphaGo is using 1+PFlops of compute (280GPUs referenced for competition in [0]). That's an insane amount

More ›