swerling

2 yıl öncediscuss

Thanks for the smile OP.

How soon until the 1/25-scale cray C90 gets as many MIPS as the original? Seems like the one he built is within shouting distance.

swerling

3 yıl öncediscuss

On GPT-3.5 and now on GPT-4, I told it a line I could remember from a sonnet, and asked it to give me which sonnet it came from. It failed, and fabricated a sonnet that was a mashup of other sonnets.

It seems like maybe GPT-4 is not good at knowing when it does not know something?

Is this a common issue with LLMs? Also surprising (to me), it seems to give a slightly different wrong answer each time I restart the chat.

Here is the transcript in case you're curious:

``` Me: hello

GPT: Hello! How can I assist you today?

Me: I can remember a line I love from a sonnet, but can't remember which one.

GPT: Sure, I can help you identify the sonnet based on the line you remember. What is the line that you can recall?

Me: "Like a jewel hung in ghastly night"

GPT: The line you remember is from Sonnet 66 by William Shakespeare. The full line is: "And art made tongue-tied by authority, And folly (doctor-like) controlling skill, And simple truth miscalled simplicity, And captive good attending captain ill: Tired with all these, from these would I be gone, Save that, to die, I leave my love alone. But, like a jewel hung in ghastly night, Makes black night beauteous and her old face new." ```

Incorrect, the response is a mashup of at least 2 different sonnets. Still I think it's really impressive how it understood what to do with "I can remember a line I love from a sonnet, but can't remember which one." That's not a well phrased question, or even a question at all, but still it knew what I was getting at.

swerling

6 yıl öncediscuss

Apple announced it way back in about 2011 or 2012.

Here is the sequence of events for our company:

2012-ish: apple requires IAP be an option if you have payment subscriptions. This date might be slightly off, I think it was way back then.

2014-ish: We launch an app that has no recurring subs. So we are still in compliance with their terms.

2017: We add a concept of membership, with braintree managing the subscriptions. Unbeknownst to us, we are now in violation of their terms.

2019: Apple notices the subs, and freezes our app updates

So we were indeed in violation of their terms, ones they had made known long before. We did not know it, but we were. Apple let us know 2 years after the fact by preventing us from publishing a bug fix release.

swerling

6 yıl öncediscuss

I think this is wrong, and really the crux of the problem. The app store is not a platform, it is a marketplace.

Apple has used control of that marketplace to impose a technology that we would not otherwise use and adds no value to our product.

swerling

6 yıl öncediscuss

Point taken.

But I would suggest, for that point to really resonate, Apple would have to make some changes in how they enforce these rules, and other aspects of IAP:

1. Provide a grace period

Apple blocked our app release one release after we did a major makeover on the app, switching from Objective C to react native to bring it in line with our 2 other apps. We expected to experience teething issues and quickly release bug fixes as a result. We could not. Given how it went down, we had to implement IAP as fast as we could, giving us a fair amount of tech debt.

2. Make sure the rules are enforced equally for all apps

See linked article for details

3. Allow developers to pay an entry fee for placing their app in the Apple App store as an alternative to implementing IAP

Apple is forcing developers to use a technology that they dont want, one that some developers may feel degrades their ability to provide the best product to their customers. They are using their position as the gatekeeper to the app store to do that.

Let us into the app store with a fee instead of forcing us to expose a feature we have no control over ourselves.

4. Improve their API

The truth is, their 'API' isnt really an API, in the sense that say Braintree or Stripe give you. You are provided a 'receipt', which is really a window onto a state machine. The 'receipt' changes over time and tells you the history of the subscription. But that history is incomplete. For example, the history wont tell you the amount of a given transaction, it's just not there. You have to deduce it from the associated payment plan. But that is hard too, since the payment plan price may change over time to test the market.

Another problem with their implementation is that Apple does not send webhook events for some subscription events -- they only send send webhook events during state transitions of a subscription. So for example if the user makes a monthly payment, there is no webhook sent, since their subscription did not change state. But we need to know these things as it impacts our messaging with our users, so we are forced to write daemons that poll apple to update the receipt and see if any new events have occurred that may require user messaging.

It is a cumbersome and awkward 'API', one I do not belive many companies would voluntarily choose were they not forced to do so.

4. Give us more control over IAP

We have a 'customer is always right' approach at our company. If a user asks for a refund, they're probably going get it. Unless they purchased something using IAP. We have no control over an IAP subscription -- for those, we must direct the user to contact apple and see if they can get a refund there. Apple does not have a 'customer is always right' approach to their users. Should they refuce a customer a refund that we would have given, that blows back on us.

That is really the tip of the iceberg, there are other things we have to cede control over to Apple that effect our relationship with our own customers, especially relating to marketing and testing price points.

swerling

6 yıl öncediscuss

When people look at this issue they tend to fixate on the 30/15% 'tax'. Is it fair? Is it not?

The commission is not nothing, obviously, but misses a larger point.

Apple did to Hey what they did to our company. One day we submitted a minor upgrade for an app that had been on in the app store _for years_, a bugfix release, and they told us they would not publish it or any subsequent submission unless we add IAP to the app.

We would be happy to pay apple some fair fee to place our app in their store. We all understand that it's not a charity. But that's just not what apple is doing here, or not the whole story.

They took an app that was _already_ in their store for years, and froze updates on it. That approach forced much of our company to stop whatever it was they were working on and implement IAP _now_. Mobile devs, backend devs, accounting, marketing analytics/funnelling were all affected. There would be no bugfix releases until that feature was added.

The implementation of recurring payments in Apple's IAP is quite different from that of other payment processors like Braintree or Stripe. As such it has added a lot of complexity to our backend services, and has sucked a huge amount of time from our developers, accounting, and from our analytics team. All so we can implement a feature we _already had in our products_ and were happy with, just their version of it.

We incorporated IAP in the fall of last year, but we are _still_ trying to digest it into our company, still refining how we handle their subs, still mitigating some mistakes we made in the our initial implementation. And their remain some unresolved question for us on the how best to do accounting and market funneling.

The media continue to fixate on the 30% commision, as do must of the comments in this thread. But the issue is that Apple does this by imposing the commission on apps that _already_ have huge sunk costs on there platform, going back years, that already have tens of thousands of users (at least in our case), and then one day send back a bugfix release with a letter that says 'if you want to put an update to this app on our platform, add IAP to it. Now.'

swerling

8 yıl öncediscuss

Rather than repair it in space, would it be easier to bring it back to earth, refurbish it and relaunch? That would be $120M plus refurbish costs. Or am I underestimating how hard it would be to bring it back to the ground?

swerling

8 yıl öncediscuss

The ruby version looked a bit odd to me. It would be unusual to not use regex in this context.

I'm not sure if this is in the spirit of the test, but using regex reduced the time from 27s to 0.04s on my machine.

    require 'benchmark'

    contents = File.read("/tmp/200mb.txt")
    find = "07123E1F482356C415F684407A3B8723E10B2CBBC0B8FCD6282C49D37C9C1ABC"
    result = nil
    elapsed = Benchmark.realtime do
      result = (contents =~ /#{find}/)
    end
    puts "Found at index #{result} in #{elapsed}s"

(The search string was placed at the end of a 200mb file)

swerling

9 yıl öncediscuss

Sorry about that. Put it back in the post under the list.

swerling

9 yıl öncediscuss

From wsj.com:

"...some site-selection experts to speculate Mr. Bezos may choose a location where an influx of workers could help promote political change..."

Playing with that notion.

Premise: 50k tech workers would have more impact on US politics if: 1) added to U.S. cities; 2) added to cities in states with smaller populations. (Debatable premise(s), but just go with it).

Below is trimmed down version of the original list of 20 with that in mind, sorted by lowest population. Cities in states where 50k would be < 0.40% of the population were cut (eg. 50k in florida would only be 0.25%).

  Place:  50k/State Pop (as %)
  ------------------------------------------

  Denver: 1.00%
  Nashville: 0.77%
  Northern Virginia: 0.63%
  Atlanta: 0.50%
  Raleigh: 0.50%
  Columbus: 0.45%

(50k would be about 7.35% of Washington DC, which I edited out, but put back in after seeing a reply to this post)

(list of cities and quote above from https://www.wsj.com/articles/amazon-narrows-choices-for-seco...)

swerling

10 yıl öncediscuss

I've got an 80-year-old row house in Queens, NY. Family of 4, no pets.

Our electricity usage is well below yours. We don't make a huge effort to conserve, and I don't believe our house is particularly well insulated. We have no special energy tech, unless you count the LED lights. We have a lot of electronic gizmos sipping away at any given time.

Winter: we don't have the baseboard heaters, but I work from home most days and have an infrared heater pointed at my desk for a good 8 hrs or so each day.

Summer: my wife hates A/C when she sleeps, so we only have one air-conditioned room at night (me and 1 kid in the room w/ A/C, wife and other kid in the other). No pets, so we leave the A/C off when we're out, it comes back on (by timer) about an hour before we get home. So big difference in summer.

Still, that doesn't sem to account for all of the difference. I just checked my con-ed history online. Some typical usage numbers:

  Feb: 310 kWh  
  Apr: 240 kWh
  Jul: 410 kWh
  Sep: 270 kWh

We spend anywhere from $80-$350 a month for electricity. Maybe there's a big difference between an 80-year-old row and a 150-year-old row?

We have

- 3 window A/C units.

- 3 routers, running 24/7

- 2 servers (mostly used as NAS), running 24/7

- FireTV, a few sonos devices, drawing a bit 24/7

- Many lights, mostly LED. We're pretty good about turning them off.

- ~10-15 phones/kindles/laptops/tablets/gizmos constantly being topped off

- Gas for stove, water, and household heating

EDIT: After reading more discussion below, I thought maybe I'd made a mistake, since there is a difference between our usage and most of the other US folks. I double checked, and the numbers are right. Our total usage for the last 12 months is 4340 kWh. We were gone for a few weeks in the summer, else it would have been a bit higher.

swerling

10 yıl öncediscuss

This is fantastic. Feature request: drag a rectangle over a group of dots, and see them as a text list of websites. As is it's hard to see all the sites that are in a dense dot cluster.

swerling

14 yıl öncediscuss

That would be the sixteenth amendment to The Constitution of the United States of America.

swerling

15 yıl öncediscuss

Beautifully done, I love this. I look forward to reading the code. Thanks.

swerling

17 yıl öncediscuss

The Rails backlash sometimes seems bigger then the hype ever was, and is cloying (so savvy, so de riguer). But this article is not one of those (you might think so from its title). The error messages really could use some TLC.

swerling

17 yıl öncediscuss

Mu.

The question is unasked.

swerling

17 yıl öncediscuss

  >  'For fun, I placed a second large message
  >  halfway through the newspaper: "Stop
  >  counting. Tell the experimenter you have
  >  seen this and win £250." Again, the unlucky
  >  people missed the opportunity because they
  >  were still too busy...

Uh huh. Was this 'study' financed to the tune of thousands of pounds? Or were the offers rescinded from all those light-hearted, open-minded, happy, lucky people who are good and deserving in every way, especially when compared to bad, sick-hearted unlucky people who probably molest their pets and deserve to be poor because they're actually inferior in deep spiritual ways when you think about it, especially when you think about the matter 'scientifically', with like, stats and everything?

Or, was the offer of £250 not actually ever made?

The article reaks of a kind of manufacturable 'luck' that it nevertheless did not mention.

swerling

17 yıl öncediscuss

While the article does not go into specifics about the "Science" involved, it is not hard to find hard science that show unequivocally that only lowering calories lowers weight, eg. this one from the New England Journal of Medicine: http://content.nejm.org/cgi/content/full/360/9/859. Why is the first law of thermodynamics so hard for people to internalize? Shall we start measuring the energy in stars in terms of grams of fat too? One expects this kind of thing on Oprah, not on HH.

At any rate, let's assume for a moment that the blogger is still eating the same number of calories, and doing the same amount of weight lifting (that is, still burning the same number of calories), but he is still losing weight. To be consistent with the 1st law of thermodynamics, it will mean he has made his body less efficient at absorbing the calories in food. So he has successfully shocked his metabolism into being less efficient. Will he stay on this diet forever? Because as soon he starts eating bread again, presumably his metabolism will emerge from the state of shock he induced. When that happens, while he will feel as if he were eating the same amount of calories, he will effectively be eating much more since his metabolism will no longer be so inefficient.

swerling

17 yıl öncediscuss

Exactly. California is, hands down, the innovation capital of the world, and has been for decades. Whatever lunacy the political winds are blowing these last few years doesn't undo the accomplishments of the last 100 years: the movie industry, microchips, software (both desktop and internet) to name a few. Entire sectors of the world economy were born there. More than a few cultural movements started there too if the argument can be extended beyond business and tech innovation.

What is really sad in California is the cost of higher education. In the mid 80s, U.C.S.D cost me about $400 per quarter (which covered health insurance), plus I had to come up with another $300/month on rent. My summer job as a Domino's Pizza guy (at $10/hr after tips) and my part time job throughout the year as a high school teacher's assistant ($7/hr) covered most of my expenses for the year. Folks worrying about California should be looking at the cost of education, and the impact that that might have on innovation.

Submissions

The Right to Forget Those That Request to Be Forgotten (Because: Apple IAP)

The Night Zombie Smartphones Took Down 911

comments