beernet

17 ms·

181.
by beernet 4y ago
I think this post explains tensors fairly well: https://www.kdnuggets.com/2018/05/wtf-tensor.html Quote: "A tensor is a container which can house data in N dimensions. Often and erroneously used interchangeab
182.
by beernet 4y ago
I find the relationship (or equivalence?) between data compression and ML highly interesting and expect more advances in data compression fueled by progress in ML research. See also: "Text Compression Is Equivalent to AI", https:
183.
by beernet 4y ago
Pretty sure this answer is machine-generated...
184.
by beernet 4y ago
Degrading "developing your people and networking" as "politics" just shows the lack of overall business and people acumen that engineers tend to suffer from. This is not necessarily a bad thing if you are aware of this,
185.
by beernet 4y ago
Generative AI and the applications built on top of it. I'm confident we will look back at things like GPT-3, Stable Diffusion or DALL-E as the first child steps of a whole new universe of technological possibilities across a wide range
186.
by beernet 4y ago
It depends on the "DL model", which is a highly vague term. Both a model with 10K parameters and a model with 10T parameters fit this description equally Well.
187.
by beernet 4y ago
Then look at some of the past time series related Kaggle challenges, plenty of evidence there in the winning solutions.
188.
by beernet 4y ago
It's true simply because large ANNs have a higher capacity, which is great for large, nonlinear data but less so for small datasets or simple functions. In any case, Transformers are eating ML right now and I'm actually surprised
189.
Palm-SayCan: Using LLMs to Improve the Planning Component of Robots for Free(twitter.com)
1 points by beernet 4y ago|1 comments
190.
by beernet 4y ago
As usual in ML, the appropriate solution depends on the problem and context. ML (particularly DL) tends to outperform "classical" statistical time series forecasting when the data is (strongly) nonlinear, highly dimensional and la
191.
by beernet 4y ago
Thanks for the post, I think these are good and valid points. Question out of curiostiy as this is not my field: On the proxies website you state that there are around 10 million proxies. How are you able to offer this massive amount withou
192.
by beernet 4y ago
Nice, 8M USD for seed in the current market climate is not bad.
193.
by beernet 4y ago
A big part of the business claim lies in the generalizability. LLMs are basis technology that, once trained, have applications in practically all sectors and businesses, because text and data of other modalities are everywhere. Massive scal
194.
Remote companies can be too asynchronous(posthog.com)
3 points by beernet 4y ago|0 comments
195.
Gazprom: Nord Stream 1 supply to Europe to be cut further(bbc.com)
3 points by beernet 4y ago|0 comments
196.
by beernet 4y ago
This article says basically nothing. It contains the typical message "Data has biases and we don't know how to solve it. In fact, we don't even know to what extent this is a problem."
197.
by beernet 4y ago
I think it’s important to have a system that works for you. For me, the key is having an organized way of capturing and organizing my notes so they are easy enough find when needed but not too much effort required at any given time (i don&#
198.
by beernet 4y ago
The protocols used for space communications are a mix of standards and proprietary. For example, the SpaceX Falcon 9 uses an Ethernet-based protocol called XDR (evolved from NASA's TDM) to communicate with ground stations during launch
199.
Ask HN: Effective data viz. tool with data privacy to report business KPIs?
2 points by beernet 4y ago|1 comments
200.
Flashcards for ML Engineering(github.com)
1 points by beernet 4y ago|0 comments
201.
Microsoft, SAP launch “Coding school for 21st century learning” in Berlin(42berlin.de)
2 points by beernet 4y ago|0 comments
202.
by beernet 4y ago
Scary this is just the beginning... a recession has not even started
203.
by beernet 4y ago
Do not develop if you cannot communicate. These skills directly depend upon another. If you can still manage to get the position, go get the money - either be very disciplined or quit soon, no inbetween.
204.
by beernet 4y ago
Is this 2018? The comments are the same. Not a crypto head but nay sayers will be nay sayers. Kinda disappointing
205.
by beernet 4y ago
While I appreciate findings like this it feels more like alchemy than science. "Oh, we found this prompt X and it works better than prompt Y but pretty likely there is at least one prompt Z which we don't know yet that works even
206.
by beernet 4y ago
It's still "just" next token prediction based on basic math. Yes, we see many astonishing results, at the same time, there is no magic to it. So I don't think that it's not clear what's going on here.
207.
by beernet 4y ago
Misses a lot of context. Where is your target group? This is where you go.
208.
by beernet 4y ago
London, Berlin, Lissabon Amsterdam less dense but an astonishing city.
209.
by beernet 4y ago
2B is a high valuation even for a great product like HuggingFace - are there any estimations on their ARR available? Even if their revenue was at 10M per year (I'd assume it is less actually) that would be a multiple of 200X.
210.
by beernet 4y ago
> Unlike BERT, GPT models are unidirectional, and their main advantage is the magnitude of data they were pretrained on: GPT-3, the third-generation GPT model, was trained on 175 billion parameters, about 10 times the size of previous mo

More ›