prestonh

It's always bothered me too, obviously when people say "we're killing the planet!" they refer to the biosphere, not the hunk of rock floating around in space.

No, I mean for productivity as well. 10 hour days (for many) aren't all that taxing.

There won't be a universally appealing solution. I have had this discussion with colleagues and most agree a 4x10 week is preferable to the 5x8 work week.

> This is just not true. Do you know the game? Are you speaking as a player? Or are you telling us what's written in the OpenAI blog post? There is nothing a player couldn't figure out. The caster team lost because of the broken game and because few of them were rusted (Merlini hadn't played for months). The bots were garbage at the TI, and got beaten without any problem by the pro teams.

A former player. I'm not regurgitating the blog post, I'm regurgitating what the players themselves said. The AI got beaten by pro teams (top <<1%), but the matches were competitive in the early game, and only later did the bots run into trouble. To give a couple of specific examples of novel behavior, the AI figured out a solid deathball strategy and was able to exploit that to beat a lot of teams, it liberally used fortify to protect creeps and sustain pushes, and it was way more aggressive in rotating its supports to critical lanes in the early game. Now, no single one of those things is entirely novel, but the combination of all of them (especially by a machine that learned it on its own) is what is novel, and what allowed the strategy to be successful.

> The hero pool is still 18 heroes. Dota 2 has over 110 heroes. Can you please try to think what makes you say something so wrong with so much confidence?

Admittedly this was a mistake, the language they used in their blog post was "Removed our last major restriction from what most pros consider 'Real Dota” gameplay'", which is poorly explained and made me think the hero pool was entirely open.

DotA is basically 2 games, the drafting part and the gameplay part. The bots made huge progress in figuring out the gameplay part, which is super impressive.

>> Casters don't count as serious players.

Maybe we're using different language. The casters are definitely in the top 1% of players, or more, which I consider "serious", but not "the best". But no one was arguing that the bots are "the best", which is self-evident from their loss at the International.

Anyways, this is all beside the point. What OpenAI was able to do was really impressive and is only helping to advance the state of reinforcement learning. You argued that there's nothing impressive about what they've done, but I'd love to see you point me to an example of an ML algorithm that learned to play a team game as complex as DotA at a competent level.

Lua looks like such a fun language to program in, but unfortunately I cannot justify its use in my work aside from small side-projects.

>> No complicated behavior was produced.

The bots were routinely pulling off coordinated team behavior that players couldn't figure out, but that worked. This has to qualify as complicated.

>> Dota 2 game was not played. A tiny subset of the game was attempted. Since people seem to be just buying whatever OpenAI propaganda sells them, let me be specific: only 18 heroes are in the game. The combinatorics explode when you go from 18 to 110. Go on a 5x5 board is a joke compared to 19x19 Go. This is not hard to understand.

Already mentioned that the hero pool has been opened up. Along with the removal of the other restrictions (invincible courier, items) this is basically pure DotA.

>> Just because OpenAI Five appears to play Dota 2 (still doesn't beat any serious players though) doesn't imply anything fundamental about tactics beating strategy.

The bots beat a team of 5 casters (granted, with some of the older restrictions in place) who are individually in the top 1% of DotA players by MMR.

You're flat out wrong or at the very least inaccurate in all three statements that you made.

What is the alternative besides chaining, devise a new variable name for every operation? I understand that daisy chaining can be error prone, but the alternative is so difficult to maintain in long chains.

I come from the background you describe, have lots of friends from the same background, and my experience has been the opposite. Most people know that data is messy business and that as data scientists we will often serve more as engineers in our day to day work.

Ive nearly been hit by cars twice as a law abiding pedestrian in the 6 months I have been in Seattle,so theres no way I am cycling here. And as a city on the West Coast, I am sure it is more bike friendly than most in the states. Fortunately, I work and live near downtown, so I can walk my daily commute.

Could you give some examples? Not challenging you, just curious.

Calorie counting can be wildly inaccurate while still producing the desired effect of weight loss. It probably has more to do with mindful consumption of food than getting the numbers exactly right.

That's correct, but what I mean is that this isnt a traditional software problem, it is a problem related to the workings of NNs themselves, whether they are embedded in software or in pure hardware.

I really doubt that at TI level the drafting advantage is significant enough to give one team 90% advantage. My guess would be closer to 2/3, granted thats still just a complete guess. I think often analysts will use the draft as an excuse to in retrospect justify the outcome of the game, but that doesn't bear on the reality of the situation at all. Though, I would be really interested to see some data that shows the significance of drafting.

The barriers preventing self-driving cars from taking off right now aren't software related though, they're fundamentally related to the limitations of neural networks that people didn't think would be an issue 1 or 2 years back. Our understanding of the capabilities and limitations is constantly changing, which is why many (including myself) thought full self-driving cars would be available in the next several years.

I don't have much experience with Matlab, so forgive me if this is incorrect, but numpy should be able to do everything that matlab can at comparable speeds. No one performing matrix/vector-like operations is doing so with standard Python lists if numpy is available.

You're right, it definitely depends on the field. The paper I am referencing looked at psychology, I believe. It is likely that a social science would have greater issues with reproducibility than a physical science.

Serious defense: p-values have been in use for a long time, and while they are error prone a larger number of true results has been found than false results (according to recent research the ratio is 2:1 reproducible to non-reproducible).

In the cartoon, the scientists are making multiple comparisons which is something strictly forbidden in frequentist hypothesis testing. One way to get around it is to apply a correction by dividing the significance theshold ("alpha") by the number of comparisons being made, in this case 20. The cartoon does not state it's actual p-value as most journals will require, but the hope would be that by dividing by the corrective factor the significance of that particular comparison goes away.

So p-value methods still lead to a lot of Type I and Type II errors, but in the past they have been the best science has been able to come up with. Actually, probably the greatest issue with false results in the scientific literature is that null results are not publishable. This leads to a case where 20 scientists might independently perform the same experiment where the null is true, for only one to find a significant result. The demand for positive results only acts as a filter where only Type I errors get made! This is just one problem with the publishing culture, and doesn't take into account researchers' bias to manipulate the data or experiment until p < .05.

An alternate approach to the frequentist methodology of using p-values is the Bayesian method, which has its own problems. First there are practical concerns such as choosing initial parameters that can affect your results despite sometimes being arbitrarily chosen, and also the high computational demand to calculate results (less of an issue in the 21st century, which is why the method is seeing a revival in the scientific community). Probably their main problem right now is that practitioners simply aren't familiar with how to employ Bayesian methods, so there's some cultural inertia preventing their immediate adoption.

This 'assumption' always bothered me when studying for DS roles because it's something that you're expected to know if asked, but isn't really true/accurate. Another is the non-collinearity assumption between variates, which is violated all the time in ML tasks but an 'assumption' of the model nonetheless.

Uber basically does own the Carnegie Mellon CS department, they hired up most of the faculty a few years back.

https://www.wsj.com/articles/is-uber-a-friend-or-foe-of-carn...

That looks completely identical to all bike lanes I've seen in suburban areas. I think bike lanes are just inherently dangerous. Would be interesting to see what a well-designed suburban bike lane looks like.

I think you hit the nail on the head with a lot of your points. In particular:

>> 1. Smart, capable students have the "inferiority complex" of being surrounded by CalTech, USC, UCLA. They know they need to push harder to have equal footing when stepping into the industry.

I just ended my graduate studies at a UC school and had some miscellaneous encounters with CSU students, both undergrad and graduate. I found that for the most part they were just as capable as many UC students, but what set many of them apart was their maturity. This may be because many of them came from unconventional backgrounds, they were either lucky to be attending college or were attending college later on in life. At the same time, a lot of them felt like UC was a more legitimate college, or that they were somehow less intelligent because they were going to a state school.

It shows correlations between colors. Consider two images, one with lots of purple and one with equal amounts of red and blue. Since the color purple is equal parts red and blue, we don't expect to see much difference in the one dimensional histograms between the red/blue image and the purple image. On the other hand, when we look at the 2D histogram we do find such a difference. The purple image's 2D histogram will have large values near the diaganol (red scales with blue). The red/purple image might have no correlation depending on the image, or could possibly be anti-correlated, with a diaganol as before, but with negative slope.

I think you are really missing the point here. The audience doesn't care whether the dinosaurs depicted are scientifically accurate, they only care whether they look like they could be accurate, and that is a big distinction. It is about matching the viewer's expectations, not reality. In JP the dinosaurs look like they could exist, whereas in JW they looked like CGI (the so-called 'uncanny valley' effect in CGI). That being said, of course JW succeeded from the perspective of the investors since it made so much money. Still, that aspect of the film is completely orthogonal to the aesthetical aspects that people here are discussing.

And with anaconda distribution, you can't have any special characters, nor spaces in the path of the installation directory. And, in some machines that have multiple users, the default installation directory is a hidden folder which causes all sorts of issues and requires a reinstall.

Just some of the issues I've encountered running python workshops. Linux and Mac users have far fewer issues.

It is in that she assumed he was American because he was white though.

Spot on, but whoever is to blame the grad students shouldn't have to bear the burden unless we want a society of undereducated people.

In equilibrium that's true, but the whole problem is that we're not in equilibrium at the moment. That's what climate change is all about. If we can increase the albedo (reflectance coefficient, basically) of the planet, we can push the Earth down into an equilibrium position at lower temperature.

EDIT: Of course, we would have to significantly alter the albedo, which is probably not possible with rooftop panels alone.

I would assume that this means that you may work at home everyday, but must be accessible for occasional in-person meetings.

Haven't read the paper, but the large energy requirements for desalination right now come from the rough interiors of the channels creating large frictional forces which resist the flow of water. Carbon nanotubes don't suffer that drawback.

Can't say I could answer all these questions, but I think they're really great and motivate me to spend more time studying statistics. With the rise in data science and machine learning I've noticed a lot of resources devoted to teaching people how to create and train models for a given problem, but less resources on how to interpret those models and on statistical inference in general.