5 ms·

1) We require food manufacturers to list on the side of their products the ingredients, as well as the nutrition facts so people can make an informed decision a

by blhack 6y ago

1) We require food manufacturers to list on the side of their products the ingredients, as well as the nutrition facts so people can make an informed decision about what they are putting in their body. There is a growing pressure to require your companies to do something similar, and publish the algorithms and metrics you use to keep people on your services longer. Would you support this movement? 2) Do you believe your products are addictive (this was already asked. I hope it gets asked more.) 3) What does it mean to you for something to be true? Could you give an example of something that you find to be uncontroversially true?

ithkuil 6y ago
These are actually good questions. There might be even more good questions; possibility too many. What's the best way to crowdsource such good questions and select a good sample and bring it to the attention of whoever organizes such hearings?
- ethbr0 6y ago
  Write your senator. All questions come from somewhere.
  - ithkuil 6y ago
    But then you don't inspire other people to come up with better questions if the question is only known to the senator ('s team) until eventually revealed to the public once asked.
gizmo385 6y ago
> There is a growing pressure to require your companies to do something similar, and publish the algorithms and metrics you use to keep people on your services longer. Growing pressure from who? I've occasionally seen these suggestions within niche tech communities, but I've never heard a layperson suggest something remotely similar to this. I'm also unsure how useful it would be. How detailed would such disclosures be? Even if they list "We use this machine learning algorithm", that algorithm description seems like it would be largely meaningless without the training data (which I don't think you could convince them to publish at all). > Do you believe your products are addictive? What company CEO _wouldn't_ say yes to this? It feels, to me at least, like a fluff question that you already know the answer to and won't yield any interesting insight into the company or how the CEO thinks about the product. > What does it mean to you for something to be true? Could you give an example of something that you find to be uncontroversially true? I think this is a much better question, although I worry that it might yield itself to some "gotcha" moments from Senators. For example, if one of the CEOs were to say "The sky is blue", I think you might get a lot of Senators saying "Well what if it's raining". That might not be the best example, but just something off the top of my head. I agree that it might be hard to come up with a unequivocal definition of what is and is not true, as that's ultimately a complicated philosophical question, but I think we can approximate truth by relying on a variety of sources who we know will do their due-diligence on research and are examining the full context of issues.
vb6sp6 6y ago
These CEOs seems really skilled at using a lot of words to say and commit to very little and it seems like all these questions could be pretty easily dodged.
- CodeGlitch 6y ago
  I bet they've had training on this. If you watch Zuck's earlier interviews, he really squirms.
  - realce 6y ago
    Mark's way worse than Jack is IMO. Jack's go-to is to say they made a mistake last time and they'll try to do better, which is kind of a dodge but at least he gives up a bit of ground. Mark never admits a single failing and remarks on how "meaningful" your interaction with his service is. Like yeah - smoking crack is pretty meaningful when you're a crack addict.
thehappypm 6y ago
The answer to number 1 is an easy "no" because trade secrets are crucial to just about every industrial sector and an algorithm is clearly a trade secret. Even the food industry allows trade secrets -- the secret Coca-Cola recipe is not printed on every can. Even the Pfizer and Moderna vaccines have proprietary stabilizing agents in them. As much as it sounds nice, the outcome would be vagueries like "we use algorithms to select the content we think you'll find most interesting based on your past browsing behavior".
tshaddox 6y ago
I’m not a big fan of the last question. It seems like it’s hinting at the line of argument (which is very popular on HN) that goes something like “you can’t provide a formal specification for what it means for something to be “true,” and you cannot determine “truth” with 100% reliability, therefore you should have no processes or policies which involve any determination whatsoever about the state of the world.”
BurningFrog 6y ago
The only practical way to do 1) is to publish all source code related to this, right? Live, after each merge/release.
UnpossibleJim 6y ago
Not to be a naysayer, but when you get into legal terms, you have to be careful with terms like "addictive". Addictive implies a component of physicality (by the legal/medical definition, not the common definition). It's horribly habitual, which is as bad as, and hard to differentiate except in the physical symptomology of withdrawal. I know this sounds like splitting hairs on message boards, but I can guarantee you that all of these people who have been called to testify have been briefed heavily by squads of lawyers.

by blhack 6y ago

ithkuil 6y ago

These are actually good questions. There might be even more good questions; possibility too many. What's the best way to crowdsource such good questions and select a good sample and bring it to the attention of whoever organizes such hearings?

ethbr0 6y ago
Write your senator. All questions come from somewhere.
- ithkuil 6y ago
  But then you don't inspire other people to come up with better questions if the question is only known to the senator ('s team) until eventually revealed to the public once asked.

gizmo385 6y ago

> There is a growing pressure to require your companies to do something similar, and publish the algorithms and metrics you use to keep people on your services longer. Growing pressure from who? I've occasionally seen these suggestions within niche tech communities, but I've never heard a layperson suggest something remotely similar to this. I'm also unsure how useful it would be. How detailed would such disclosures be? Even if they list "We use this machine learning algorithm", that algorithm description seems like it would be largely meaningless without the training data (which I don't think you could convince them to publish at all). > Do you believe your products are addictive? What company CEO _wouldn't_ say yes to this? It feels, to me at least, like a fluff question that you already know the answer to and won't yield any interesting insight into the company or how the CEO thinks about the product. > What does it mean to you for something to be true? Could you give an example of something that you find to be uncontroversially true? I think this is a much better question, although I worry that it might yield itself to some "gotcha" moments from Senators. For example, if one of the CEOs were to say "The sky is blue", I think you might get a lot of Senators saying "Well what if it's raining". That might not be the best example, but just something off the top of my head. I agree that it might be hard to come up with a unequivocal definition of what is and is not true, as that's ultimately a complicated philosophical question, but I think we can approximate truth by relying on a variety of sources who we know will do their due-diligence on research and are examining the full context of issues.

vb6sp6 6y ago

These CEOs seems really skilled at using a lot of words to say and commit to very little and it seems like all these questions could be pretty easily dodged.

CodeGlitch 6y ago
I bet they've had training on this. If you watch Zuck's earlier interviews, he really squirms.
- realce 6y ago
  Mark's way worse than Jack is IMO. Jack's go-to is to say they made a mistake last time and they'll try to do better, which is kind of a dodge but at least he gives up a bit of ground. Mark never admits a single failing and remarks on how "meaningful" your interaction with his service is. Like yeah - smoking crack is pretty meaningful when you're a crack addict.

thehappypm 6y ago

The answer to number 1 is an easy "no" because trade secrets are crucial to just about every industrial sector and an algorithm is clearly a trade secret. Even the food industry allows trade secrets -- the secret Coca-Cola recipe is not printed on every can. Even the Pfizer and Moderna vaccines have proprietary stabilizing agents in them. As much as it sounds nice, the outcome would be vagueries like "we use algorithms to select the content we think you'll find most interesting based on your past browsing behavior".

tshaddox 6y ago

I’m not a big fan of the last question. It seems like it’s hinting at the line of argument (which is very popular on HN) that goes something like “you can’t provide a formal specification for what it means for something to be “true,” and you cannot determine “truth” with 100% reliability, therefore you should have no processes or policies which involve any determination whatsoever about the state of the world.”

BurningFrog 6y ago

The only practical way to do 1) is to publish all source code related to this, right? Live, after each merge/release.

UnpossibleJim 6y ago

Not to be a naysayer, but when you get into legal terms, you have to be careful with terms like "addictive". Addictive implies a component of physicality (by the legal/medical definition, not the common definition). It's horribly habitual, which is as bad as, and hard to differentiate except in the physical symptomology of withdrawal. I know this sounds like splitting hairs on message boards, but I can guarantee you that all of these people who have been called to testify have been briefed heavily by squads of lawyers.