skoocda

11 ms·

61.
by skoocda 9y ago
Probably not very hard. The biggest headache is probably that the recurring credit card purchases might be flagged and blocked for being an unrecognized international transaction. But then again, $1.50 won't raise alarms for many banks
62.
by skoocda 9y ago
And dozens of "I left google" : https://hn.algolia.com/?query=%22I%20left%20google%22&sort=b...
63.
by skoocda 9y ago
This paper suggests that increasing vitamin D intake from 800 IU/d to 2000 IU/d doesn't increase the risk of falling over and fracturing a bone. I wouldn't call that especially relevant, and it's definitely not an o
64.
by skoocda 9y ago
I can't find exact numbers on Snips.ai, but generally there's a linear relation between the size of the inference model in RAM and the accuracy it can obtain. I'd have to assume DeepSpeech outperforms anything running on a Ra
65.
by skoocda 9y ago
Oh my mistake, it was actually I who misread the first comment. Regardless, I don't think a screen tap logger would be necessary in many cases, since you could just open the keyboard on the phone and manually check the bounding boxes f
66.
by skoocda 9y ago
You're only accessing accelerometer / gyro data here, so yes this could occur in the background. On iPhone it restricts background activity after 10 minutes, but on Android you'd have free reign until they manually close the
67.
by skoocda 9y ago
This is somewhat common during brain surgery. If the patient plays a wonky note (or stops playing altogether) it might be because the surgeon cut or poked the wrong spot. See another example from a month ago [1] where a musician had dystoni
68.
by skoocda 9y ago
Do you have a website? Am interested to learn more but I don't have any Alexa-enabled devices, so I can't really use your link.
69.
by skoocda 9y ago
Recently got turned down for a job here, now I'm remarkably glad I did - this is an incredibly difficult problem to tackle, and they need absolutely brilliant people to achieve success. I wish them the best of luck in their continued p
70.
by skoocda 9y ago
This should probably be re-titled now that it's been resolved. @dang is this possible?
71.
by skoocda 9y ago
Here's the full interview[1] - the above comment linked a shortened version. For reference, this is Dr. Jordan Peterson, the University of Toronto professor who was fired a few months ago for his stance on using gender pronouns. He&#x2
72.
by skoocda 9y ago
Left <-> right interchange doesn't really happen anymore since radial tires have become standard on most vehicles. Radial tires are unidirectional, so if you go left -> right your tires will be rotating backwards, and then the
73.
by skoocda 9y ago
This comes up semi-regularly when Twitter is mentioned on HN, but I've moved all my Twitter attention over to Nuzzel and have enjoyed the change. Twitter started to feel like such a poor return on my time due to the format- having an a
74.
by skoocda 9y ago
What I find even more exciting is that they're finally deploying their predictive suspension technology in a production model. What this system does is utilize the front cameras to scan and map the road surface ahead. It can then apply
75.
by skoocda 9y ago
Some would further argue that learning underlying meaning (which can only derived from context, AKA the pragmatic), is completely impossible without agency . Until these AI systems can grok information relative to some sort of spiritual&#
76.
by skoocda 9y ago
A few people here have said the Moto Z Play. If you're in EU / CAN, the Moto X Play is (was?) available, which has even better battery life... and no bloatware to boot. However, you forfeit a gyroscope, so no VR with this one.
77.
by skoocda 10y ago
In the truck/SUV/CUV segment, the engines are dated- sure. But that segment barely exists in other markets, at least compared to small cars. Particularly in Europe and Asia. Regardless, I don't think it's a tech issue, b
78.
Jeff Hannan: Training Neural Networks to Drive a Rally Car (2001)(ai-junkie.com)
1 points by skoocda 10y ago|0 comments
79.
by skoocda 10y ago
End of term is fast approaching!
80.
by skoocda 10y ago
'cute dog videos 10 hours HD' is a 216 bit UTF-8 string input. Youtube's video result is a 288 000 000 000 bit HD video output. So for the (server - disregarding the client) computer, output exceeds input. For the human, inpu
81.
by skoocda 10y ago
The 'whole techno-futurism stuff' that Musk focuses on is more of a reflection of the narrow set of problems which are defined well enough to be solved by investing adequate capital; because they're mostly dependent on techno
82.
by skoocda 10y ago
https://cloud.google.com/speech/
83.
by skoocda 10y ago
Very cool! Reminds me of Voxer
84.
by skoocda 10y ago
It still relies on 3rd party proprietary STT- which is a huge problem for most of these speech interface projects. The best results in the open-source STT field are accomplished using a library called Kaldi. However, it's a pain to set
85.
by skoocda 10y ago
Children's voices are a surprisingly tough problem in speech recognition. Mostly because there isn't much labeled data with children's voices- ASR follows the trend seen in other deep learning fields of working best for North
86.
by skoocda 10y ago
I wonder what he used for the speech recognition component- quite certain it's an API, but is it internal or 3rd party? Maybe something by Wit.ai which is still unreleased?
87.
by skoocda 10y ago
Can't access- flagged as a 'fair use violation'. No details provided, first time I've been on this site. Any idea what's going on?
88.
by skoocda 10y ago
See my adjacent comment. Also, to add some more info here; the site will soon allow you to share the transcripts, so we don't all have to upload a copy for ourselves. We're also building editing / revision tools.
89.
by skoocda 10y ago
https://app.spreza.co We've been working on this recently- it's not yet a public release, but it is useful (and free). Just make an account and upload the MP3. Allows you to read, listen or both at the same time. Takes
90.
by skoocda 10y ago
Not easily! I'm doing speech recognition; I might be able to run a segment through two separate models, with half the lexicon available in each run, then just combine the results. However I don't think the overall savings will be

More ›