ypcx

GPT-3 is (the process of) general AI being born in front of our eyes, catching us unprepared, making us realize that the we may be a bit rigid in understanding of what intelligence really means. If primed correctly, can GPT-3 produce worthwhile improvements to its own paper or implementation? Can GPT-3 be made to prime and query itself in a loop, by asking it to produce a better query for itself? Can its knowledge be already used to disrupt existing frontiers of human thought? The simple and safe answer here is that we don't know yet.

In a world full of ideological adversity and cold-blooded corporate competition, slowing down the release of technology which may or may not be capable of a runaway effect (a.k.a. technological singularity) in the hands of an adversary is an intelligent and prudent thing to do.

Okay. But how do we determine if something is logical? At the very least we have to abstractly infer and compare with what we have been taught is logic (because hardcoding it manually doesn't map to the [hierarchical] intricacies of reality very well). So logic is a high level reasoning which has to be powered by the low-level abstract/infer/mimic as exhibited e.g. by the transformer.

I wonder if expert systems could be used to generate a "logical reasoning" training dataset (or a cost/fitness function) to help train/evolve neural networks on. Or if there are other ways of integrating these two.

If prediction cannot replace reasoning, then how would you define reasoning? If reasoning is the process of inference of the most likely from the most similar known, then how does the multi-head attention transformer not fit that description?

The transformer model as presented in GPT-3 may be a few tweaks away from a human-acceptable reasoning, at which point we may realize that human brain is just a neat party trick as well. This may come difficult for some people to internalize, especially those who understand the technology in depth. Because it means that the medium of our reality is the consciousness.

I think most of us would agree that all drugs (drug use) should be decriminalized. But even if you go as far as to imagine the Mexican cartels being (eventually) legitimate businesses, the strong competition between them would not subside, so they would still fight each other and keep their grip on the government and the population. (Then there's also the ties they have to the governments of the target markets, and their policy making.)

I think to solve this situation, you'd have to diminish the demand for their product in the target markets, which is next to impossible due to the addictive nature of the product. One way to do this would be to legalize and promote therapeutic use of drugs which are known (among many other positives) to break drug addiction (in just a few sessions), like psilocybin mushrooms or Ayahuasca.

2020 version: https://openai.com/charter/

There was this Google Inbox thing which was loved by many including myself. A true successor to Gmail, in my opinion. But Google scrapped the project.

Edit: clarify

For me one of the worst were Docker for Desktop on both Mac and Windows, especially when used for local Kubernetes. I fixed this with a project running Kubernetes directly on a local virtual machine(s) and local Docker (without polluting the machine with Docker for Desktop) is a bonus that comes with that[1].

[1] https://github.com/youurayy/hyperctl

ConsoleZ on Windows that I use with Cygwin, being on one hand the best terminal app for Windows, on the other there were things that were driving me insane. The main project seems to be sort of abandoned, so I eventually fixed them myself[1].

[1] https://github.com/youurayy/console/releases/tag/1.19.0-pers...

From the point of UI / UX, I cannot think of a worse software than Skype (all platforms including Android).

It's always a hard struggle, effort and suffering to make a call to a phone number, even if this phone number is in the contacts of the phone, or has been called before.

Adding to that, when Microsoft acquired Skype, they were offering to merge the Skype and Live accounts, which I agreed to, which rendered my Skype account unusable (with credit on it and lot of contacts registered).

The. Worst.

I only use the web version. The native version is a resource hog (or was, when I last tried to use it).

The network connection quality on Slack calls seems to be very lacking compared to other call apps.

I wonder what the GPT-3[1] has to say about this. While it can't exhibit reasoning, I wouldn't bet that it is completely off the path getting there. Call me crazy but I think we are about a year from first general reasoning algorithms.

[1] https://www.gwern.net/GPT-3

The good. To have 100% reliable Kafka, your data (both transfer and storage) needs bot replicated by the factor of 3 (you will need to transfer and store 3 copies of all your data). Pulsar will happily do with just 2x. Next, thanks to the "failover" subscription type and producer real-time deduplication, high-availability consumers are easier/cleaner done on Pulsar. On Kafka, you have to fiddle with Consumer Groups and being very aware of partition numbers of your topic.

Now with that said, these are the current downsides of Pulsar as recently perceived by me: More complex architecture (in the Kubernetes context) (zookeeper + broker + bookie + proxy + autorecovery [+bastion/prometheus/grafana]) vs Kafka's (zookeeper + broker [they are working on removing zookeper now]). Pulsar has extremely active development, but not that many active developers or active community size, large number of active bugs, the project is far less mature than Kafka, the documentation body is much smaller and worse, and the respective Stack-overflow (etc.) knowledge-base is much smaller than Kafka's. My experience with Pulsar's deployment to Kubernetes was that is that it wasn't ready. E.g. a lot of startup synchronization is being done by k8s yaml-embedded startup scripts and is extremely brittle and in some cases broken (you have to manually restart the proxy pod on startup, etc.). Due to the metric exposure on Kafka I find some critical devops scenarios (e.g. handling a full node loss) to be more transparent and somewhat more possible to approach than with Pulsar which appears more complex and more of a black box to me in this respect at this time. One big advantage often brought up in Pulsar vs Kafka comparison is that Pulsar has active sharding rebalance, but to my knowledge Kafka now has something like that too, albeit maybe not as dynamic as Pulsar has, not sure.

So to summarize, Pulsar is great, but actively developed and quite complex given the size of the (serious) active user base and the active developer base. I think the active user base being the main developmental driving force, it's a chicken-egg issue which simply takes its own time to evolve in parallel. We have to realize that Kafka is a 10 year old open source Apache product, thus very mature, and that's why I'd recommend it for new projects (which need to reach production quickly) over Pulsar at this time.

It is worse than just crazy. Government is made of people, and people are corruptible. If the encryption backdoor key leaks from the government to a bad actor, it will create a "national security" issue of magnitude never seen before. If this line of reasoning is correct, then proposing an encryption backdoor is akin to committing an act of treason in itself, because it is purposely weakening the technological infrastructure of the businesses and people in the country and thus the country itself. Attempts like these are either doomed to fail, or they will doom the country to fail.

There seems to be an ongoing work for Vulcan Compute support for Tensorflow. But the mlir repo moved at the end of 2019 and I don't see where (or if) the discussion and PR continue, because the new repo doesn't even use Github Issues.

  https://github.com/tensorflow/mlir/issues/60
  https://github.com/tensorflow/mlir/pull/118
  https://github.com/tensorflow/mlir/

I'm thinking what would be the purpose or added value of a separate "public spec" document (which would require a significant maintenance effort) when you already have [1] a healthy Github (or Gitlab) project governance going on, [2] Bitcoin feature proposal process.

  [1] https://github.com/bitcoin/bitcoin/projects
  [2] https://github.com/bitcoin/bips

When you upgrade your Bitcoin node software, the act of the upgrading is you casting your Yes vote for the changes introduced in that version. The BIPs etc. can help you understand these, but if you want to be 100% diligent, the only fully authoritative source is the state of (the code sources in) the git repo as referenced by the commit hash in the Bitcoin software you are upgrading to.

> How do we make this better?

I do believe that a great library (package) is either self contained (e.g. lodash, moment, etc.) or has a minimum number of dependencies on other such libraries.

I think we can think of dependencies as adding to the code size and code complexity of the library which uses them, and then derive the quality of said library from those metrics.

JavaScript is great for writing code which can do a lot, and writing it fast, and I think this power shouldn't be misused to manufacture a lot of small libraries with a lot of dependencies. Perhaps we should use something like Github Gists or Stack Overflow answers for that.

This why I am personally gravitating towards Deno (https://deno.land/) now -- it has a nice set of base APIs and modules, and it allow me to pull a self-contained library straight from the CDN.

Seems like the relevant blog entry is here: https://blog.c0nrad.io/posts/hydrogen-pt2/

Edit: the Reddit discussion has more info: https://www.reddit.com/r/Physics/comments/gt1set/interactive...

And people would start losing trust in the network if this was happening, thus the value of the currency would start dropping, thus all the network participants including the miners would be in a losing position in the long run, which is not what you are looking for after you have invested money in the mining equipment and the whole operation.

That's why the total "hash rate" (the computational effort spent by all the participating miners) is in direct relation to the perceived value of the currency.

Deno is a JavaScript runtime built in Rust, so I guess it depends what you mean by "types". Definitely not a strong typing, although the example could be rewritten to TypeScript, which Deno supports natively. I use Deno sometimes, so it was of interest to me to compare it to the advancements presented here for Go.

I've tried to parse those test files in Deno (in a very rudimentary test[1]), the results (on my i9-9980HK @ 2.40GHz) are:

  canada.json          -->   31 ms,   73 MB/s
  citm_catalog.json    -->   13 ms,  135 MB/s
  code.json            -->   17 ms,  113 MB/s
  example.json         -->    0 ms,   73 MB/s
  sample.json          -->    6 ms,  124 MB/s
  twitter.json         -->    6 ms,  114 MB/s

[1] https://gist.github.com/youurayy/18553475c5a9f81a17345cddeeb...

I think I killed mine with the shutup10 tool. I can't imagine having my OS telling me when to update (and potentially kill the features I need, e.g. the Hyper-V) and when to restart. I understand the need to force an Office-running, email-attachment-clicking user to update periodically, but as a dev this is way past the line I would ever tolerate. And thanks to (also) folks like you, I don't have to, so thanks for that.

Another first thing I do after Windows 10 install is killing the task scheduler. Among other things, it contains a broken task related to Xbox support, which activates on idle and takes up like 50% of the CPU. I think that people think that the machine is running a defrag or some other optimization.

If Linux got its UI business together, I would flee from the Windows immediately. I've been 10 years on Mac but we all know where that's going. Now with WebKit and Node.js so advanced, I'm not sure what's stopping us from creating a really nice Linux UI ecosystem. Perhaps the state of Linux laptop drivers, mainly the touchpad ones.

I've quit Dropbox when they capped the free account to (I think) 3 devices only. I'm now using Mega.nz which has end-to-end encryption and works well. I actually have a paid account with them to backup photos from my phones. My local cloud directory is still named Dropbox.

I have recently analyzed Kafka as a message bus solution for our team. My take is that Kafka is a very mature and robust message system with clearly defined complexity and thus when used correctly, very dependable. The recent versions even have online rebalancing (but don't tell that to the Pulsar people). However "Kafka Streams" just make me shiver - I don't understand why someone would masquerade a project directly competing with other processing frameworks (e.g. Apache Flink or Spark) as being part of Kafka itself, when apart from utilizing Kafka's Producer and Consumer interfaces it really has nothing else in common with it. Streams is a giant, immature, complexity-hiding mess which as a responsible systems designer you have to run away from really, really fast.

What's the difference between thinking that you are conscious, and actually being conscious? How can we define our perception of being conscious? Is it simply the ability to observe our own thinking (or feeling, or perceiving) process? If so, what is the output of such observation, and why do we think that there has to be (or should be) something more than just the programmed response for that given data input? I.e.:

  if(is_currently_observed(my_own_thought)) {
    invoke_feeling('being_conscious')
  }

Edit: I swear that I am conscious. I can feel it.

Thanks for writing this.

Do you think they gave you the false EOL information knowingly, by mistake, or was it a coincidence that all the parts got binned prematurely?

Why do we open new tabs?

1. We want to do multiple things at once (very rare, very limited, let's say we have 5 such things, Gmail, Slack, Gitlab, monitoring, etc.)

2. We want to have websites preloaded and ready to go as soon as we get to that tab (often)

3. We want to "bookmark" links temporarily without actually saving them into our Bookmarks (this goes hand in hand with point 2. but not always)

4. Observation: large numbers of pre-opened tabs will stay opened for a very long time until we actually find the time to get to them, consuming memory, tab bar space, as well as our mental space as we have their tab icons visible.

So I think based on the above, an ideal tab management system would:

- understand "sessions/groups/trees", e.g. I opened HN and alt+clicked on links and discussions that interest me as I went through the list

- gave me an opportunity to transfer such "trees" to another browser window, session, computer

- had an option to automatically, or on a command, archive (and later restore from) these into a tab history, which would retain the hierarchy

- had other functionality that I cannot foresee at this moment

Additionally, the browser should probably load & pause a background tab once its opened using alt+click, and only run it once its actually opened by the user.

Edit: formatting

Now, can we evolve a ML algorithm that would in turn produce a better AutoML? Ladies and gentlemen, the Singularity Toolkit v[quickly changing digits here].

I test both via keto urine strips and via keto breath analyzer. But when you are on a strong keto, you simply know it, you can feel it by the increased amounts of body and brain energy, mental clarity. Also your urine is more yellow and has a more specific odor.

>Since mid-January this year

That's a sample too short, a lot of that can be the momentum carried over from whatever you have been doing previously. You have to understand that as a carb/sugar eater, you may have effectively destroyed your body to the point when it's much harder for it to digest nutrients from food (e.g. fatty liver). And/or you may have a starting diabetes.

Before jumping into the meds death spiral, check Dr. Berg's videos on YouTube, and I'm sure you'll find a way forward. When your body is broken, nutrition isn't a simple thing, trust me, I know.