Smerity

26 ms·

151.
by Smerity 11y ago
You might need to clarify what you mean - I'm confused. What you're discussing doesn't seem grounded in modern ML/AI and seems to be comparing Tay to how humans learn? Almost none of the modern machine learning algorithm
152.
by Smerity 11y ago
Repeating what I wrote in a blog post[1]: > Humans have the tendency to imbue machine learning models with more intelligence than they deserve, especially if it involves the magic phrases of artificial intelligence, deep learning, or neu
153.
TayAndYou – toxic before human contact(smerity.com)
24 points by Smerity 11y ago|0 comments
154.
by Smerity 11y ago
I have doubts as to whether this system was even performing online learning yet. Even if it was, that wasn't the cause of many of these issues. Like conversational bots from the past, they tried to appear intelligent by copying previou
155.
by Smerity 11y ago
I was testing disaster recovery for the database cluster I was managing. Spun up new instances on AWS, pulled down production data, created various disasters, tested recovery. Surprisingly it all seemed to work well. These disaster recovery
156.
by Smerity 11y ago
Thanks for the good discussion :) + The issue with MCTS was not that it couldn't be extended to non-determinism (you're correct re: POMCP) but that it requires a simulator which produces at least a reasonably accurate model of the
157.
by Smerity 11y ago
The companies may have plenty of resources but those resources were not solely dedicated to this problem. You mention Facebook and they were indeed on the verge of putting time into this with - though their team is far smaller (1-2 people)
158.
by Smerity 11y ago
The best discussion in terms of hardware adjusted algorithmic process I've seen is by Miles Brundage[1]. The additional hardware was the difference between amateur and pro level skills. It's also important to note that the non-dis
159.
by Smerity 11y ago
Preface: AlphaGo is an amazing achievement and does show an interesting advancement in the field. Yet ... it really doesn't mean almost anything that people are predicting it to mean. Slashdot went so far as to say that "We know n
160.
by Smerity 11y ago
Thanks for the great comment! For a direct link to the visual saliency maps from the paper (Figure 6) which @nicklo mentions: http://i.imgur.com/DRfaNxB.png Being able to see where a neural attention mechanism is fascinatin
161.
by Smerity 11y ago
Congratulations to the TensorFlow team on this massive step! It's also hilarious that using distributed TensorFlow in Keras is a trivial change[1]. I honestly think high level libraries with transparent backends will be the future in t
162.
by Smerity 11y ago
I agree with you. The addition of the meta tag actually allows you to see the traffic is coming from HN, even if you are a HTTP site, which was not the case when the story was written. HN also specify only to send the "origin". Th
163.
by Smerity 11y ago
The site can signal, either via server headers or meta tags, whether to send or withhold referrers. Adding <meta name="referrer" content="never"> will prevent referrers from being sent, whilst <meta name="r
164.
by Smerity 11y ago
From the article: > HTTPS websites will send referrers to any other HTTPS website even if it contains sensitive information As such, having a website which is HTTPS will get you referers from both HTTP and HTTPS if they're willing t
165.
by Smerity 11y ago
As the author, this was a pleasant surprise! :) It was written in 2013 (potentially note that in the title?) though people still frequently say they find it useful. As @jstayton noted, browser support has come a long way[1], and @davecardwe
166.
by Smerity 11y ago
Keras has a low-level compatibility library[1] (`from keras import backend as K`) that people have reported as useful independently. The interface offered by K[2] seems quite similar to that of TensorFuse[3] As the K backend is used for Ker
167.
by Smerity 11y ago
The diagram was generated using a DMN I've implemented in TensorFlow, part of my work at MetaMind. Those diagrams are useful not just in visualizing the architecture but also for spot checking certain issues. I was planning on writing
168.
by Smerity 11y ago
Model serving in production is a persistent pain point for many ML backends, and is usually done quite poorly, so this is great to see. I'm expecting large leaps and bounds for TensorFlow itself. This improvement to surrounding infrast
169.
by Smerity 11y ago
I love the tool and tweeted about it yesterday. It's brilliant, I love it, excited to try some more complex probabilistic distributions. Only request would be to allow for private spreadsheets. I can download and run the code locally b
170.
by Smerity 11y ago
Given how many deep learning toolkits have popped up and the complexities involved, these evaluations are quite useful. Code comparisons would be even better but I feel that's asking too much given how quickly things move :) One import
171.
by Smerity 11y ago
Honestly, there are enough issues with TensorFlow right now due to CUDA 3.0 that using it with AWS is highly problematic. I appreciate the author's attempt, but there's no way the five lines of code he changed to allow CUDA 3.0 ha
172.
by Smerity 11y ago
"Prison and jail communications is a $1.2 billion a year business, whose handsome profits come from serving a captive and inelastic market." Securus is used by more than 1.2 million inmates and make $404 million in revenue, equati
173.
Why small files are a curse for big datasets(smerity.com)
4 points by Smerity 11y ago|0 comments
174.
by Smerity 11y ago
Discussing this with friends, we pondered the advantages of university degrees being cryptographically signed. If universities were created today, the idea that they would only use paper (or a scan) to certify a degree is laughable. The pro
175.
Certificates, Reputation, and the Blockchain(medium.com)
45 points by Smerity 11y ago|11 comments
176.
Certificates, Reputation, and the Blockchain(medium.com)
3 points by Smerity 11y ago|0 comments
177.
by Smerity 11y ago
I'm very excited by this work. As they say, testing for distributed systems leaves a great deal to be desired, but there work is a great step! I was excited for FoundationDB, who put a great deal of focus on testing[1], but they were a
178.
by Smerity 11y ago
Common Crawl ( http://commoncrawl.org/ ) Position: Crawl Engineer / Data Scientist Location: SF or Remote Email: jobs@commoncrawl.org Common Crawl is the non-profit organization that builds and maintains the single large
179.
by Smerity 11y ago
If people are interested in playing with the dataset from "Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks", I added an example to Keras that uses recurrent neural networks[1]. It achieves similar accuracy
180.
by Smerity 11y ago
As you state from looking at the website, the stated purpose is for education, but NLTK is used (and advertised) as far more than that. Due to this mixed message, NLTK is used in places it shouldn't be - the author's main concern.

More ›