hadsed

8 ms·

31.
by hadsed 7y ago
Some distractions are good!
32.
by hadsed 7y ago
For ML practitioners out there, this is a great method for a field also in a replication crisis.
33.
On Creativity in Academia(timdettmers.com)
3 points by hadsed 7y ago|0 comments
34.
by hadsed 7y ago
Link doesn't work on mobile?
35.
by hadsed 7y ago
Whenever the topic of building online open source communities comes up I feel compelled to share the work of the great Pieter Hintjens, the guy who wrote zero MQ. He wrote a book about this topic which I thought was quite good: https:&#x2F
36.
by hadsed 7y ago
Yeah, in some sense any layer is a "memory module". Perhaps more specifically, attention solves the problem of directly correlating two items in a sequence that are very, very far away from each other. I'd generally caution a
37.
by hadsed 7y ago
If you think a longer context length might be helpful consider stacking convolutions to give higher units a bigger receptive field, or try the convolutional LSTM. If that helps and you have a further argument for why an even larger context
38.
by hadsed 7y ago
How are people using knowledge graphs at companies? Immediate examples that come to mind are product categorisation and perhaps trivia questions. What else?
39.
by hadsed 7y ago
An excellent video lecture on this by Max himself which is brilliant and very intuitive: https://youtu.be/5MdSE-N0bxs
40.
by hadsed 7y ago
The architecture doesn't change but the way it works often changes dramatically with scale.
41.
by hadsed 7y ago
I recently read, The Uninhabitable Earth. It came out this year I believe, and pulls absolutely no punches... Not even in the title! https://www.goodreads.com/book/show/41552709-the-uninhabitab...
42.
by hadsed 7y ago
Outperform on what tasks?
43.
by hadsed 7y ago
Since you like simple responses I will try my hand at it. :) The problem is, "filtering" for jobs across multiple industries is harder than doing it for one industry.
44.
by hadsed 7y ago
I used SO for candidate search. It worked well for SF, but I recently moved to LA to start a team down here and it was just ridiculously... bad. LinkedIn was better, and that's saying something. But I think it's the scale effects
45.
by hadsed 7y ago
A little off topic... but I love this because it gives me an idea. I've been looking at LinkedIn with some immensely negative feelings for some time now. I've always wished someone would just disrupt the hell out of them, because
46.
by hadsed 7y ago
Alright alright, you worded it well enough, but your questions seemed to be a little sharp given that the guy is just asking for some help. More people should do that generally speaking, instead of not... everyone is better off, as long as
47.
by hadsed 7y ago
Why the hate? OP's looking for people to help them grow. No reason for you to assume they can't get there with the right help.
48.
by hadsed 7y ago
Woz admits that there are benefits to FB. What he's talking about is the privacy angle. So what's the optimal usage pattern here to enjoy the benefits and still keep your privacy safe? I think it's actually pretty straightfor
49.
by hadsed 7y ago
So, does anyone here use Heron? Or decided not to use it? Why are there so many stream proceeding engines?? (No disrespect to the project, I'm sure there is good reasoning here I'm just not aware of it.)
50.
by hadsed 7y ago
So what is the difference between saturation and recession? To me it seems like you're going to invest and pile on until you hit saturation, in which case you have to cut back because the demand isn't there. That cut back is reces
51.
by hadsed 7y ago
Amazing to find a VC who not only isn't concerned but actively derides concerns around privacy. Maybe he has a terrible personal opinion (in my opinion), but this seems like a poor perspective from someone who should be worrying about
52.
by hadsed 7y ago
Not sure I agree with this perspective. Many of these companies are started by engineers or product managers who have seen the boring solutions not work for them. In some cases they either see it not work many times or see a future where ot
53.
by hadsed 7y ago
It doesn't have to reach its end game goal immediately. It's an iterative process of slowly gathering people together, showing others that there are at least some people who care, technologists seeing the demand and creating produ
54.
by hadsed 7y ago
I'm not sure it's the slowest. In fact according to this chart it could potentially be one of the fastest: https://www.wired.com/2016/01/phil-kennedy-mind-control-comp... (Seems i can't link the act
55.
by hadsed 7y ago
It is getting easier to scale deep learning in difficult domains. It'll likely be some combo of pretrained semi- or self-supervised models that are transferred and fine-tuned. You mention large data, but we also have the knobs of induc
56.
by hadsed 7y ago
You should read a bit about the revolution of physics in the late 1800s-early 1900s. There was just as much carrying on with theories that are popular despite mounting evidence to the contrary. The tide was so great that Ludwig Boltzmann co
57.
by hadsed 7y ago
The negative comments here are disappointing. ML is a fantastic tool for science, where it can propose a model that works as a starting point for getting to a model that works AND that you can understand. This is quite common in physics, fo
58.
by hadsed 7y ago
Black boxes are better than nothing. The ultimate black box is the universe, where experimental scientists can fiddle to try and understand. They give a great starting point to make progress, if that's what you want, and if not (like i
59.
by hadsed 7y ago
In fact I found the second and third talks more interesting than the first! Which sadly are not included in the slides from the OP.
60.
by hadsed 7y ago
Well, a better video title would be "How Starlink Works". I think it's high time that communications gets disrupted. And this is Elon's brilliance yet again. Targeting the financial industry with a low-latency communicat

More ›