4 ms·

> I think they are going to fail, and quietly increase the human involvement once they realize that their AI isn't good enough. Then they will slowly scale back

> I think they are going to fail, and quietly increase the human involvement once they realize that their AI isn't good enough. Then they will slowly scale back the humans as their AI improves, until the next controversy when they realize that they still need humans. Exactly. AI is not good at subjective decisions of qualitative data. For example, nobody knows any political candidate's net worth apart from IRS, until they make those records public. And say political candidates make statements that they are way less or way more than their net worth they cannot detect it is true or false. That's why I was surprised. Pichai is smart enough that AI can't combat fake news. Hence he was just saying to save face.

abtinf 8y ago
Short of launching a forensic investigation, subpoenaing records, and carefully watching you, how would the IRS know your net worth?
ehsankia 8y ago
Realistically there doesn't exist a good solution for this. That's why every single site out there with user content that is large enough is struggling with moderation. There is no solution that scales up to billions of users, and while it's true that AI most likely won't work, it is the best they've got right now. Do you have a better solution? Because Google has hired some of the smartest people and even then they still are having issues with Youtube every other week, so I'm sure you'd be paid a pretty hefty sum if you could solve this.
- raxxorrax 8y ago
  > Do you have a better solution? Let the receiver evaluate the information? Sure, some will have false believes. Cannot change that and you shouldn't even attempt to do so. Because authority has been wrong to an equal degree. The whole premise falls flat in my opinion. There won't be any viable solution, if the problem cannot even be defined. Currently it is based on a feeling that there are mean and false statements on social media. And this fact is emphasized by parties who like more control about content. Anyone aware of the current capabilities of AI should know, that it is no solution, at least in its current state. Sure, big tech likes to signal otherwise. But these statements just underline their business interest in that sector. Nothing wrong with that. Wanting to tighten control about content is.
  - toofy 8y ago
    > Let the receiver evaluate the information This ignores one of Google's primary missions. As jlebar posted in a different comment > evaluating trust on the internet using computers (call it AI or not) is literally Google's core competency. This idea that we should pretend as if there is no problem to solve seems incredibly strange to me. Particularly when we know unreliable information is being weaponized by bad actors to influence entire populations of people. This is a very real problem which needs solving. Humans have limited time for research and we value reliability. Consider all of the hype around the recent severe decline in trust of authenticity and reliability of products purchased through Amazon. We tend to prefer shopping at places where we know the products we purchase will be of a certain standard. We go to trusted sources (as amazon was previously) because we don't have the time to carefully evaluate each of the many many items we purchase every week. We choose a store we trust will have already done that quality check. It seems as if Google believes it is important for it's business that the information it prioritizes has a certain quality or reliability level. Google prefers their information be more like a Target store than a back alley tent. And if the amount of people who shop at Target stores over back alley tents is any indicator, I'm guessing Google is probably on to something. I'm certainly skeptical about who should be able to decide what is "true" on subjective issues, but I don't see how we can pretend as if there isn't a very real problem of actual provably false information being weaponized. I for one am fairly busy and just as I don't have time to do a deep dive of research into whether or not the shampoo, deodorant, milk, cereal, orange juice, whiskey, dish soap, cold medicine etc etc etc are fake; I also don't have time to deep dive research if the 30+ articles I skim everyday are provable outright lies. Not only don't I have the time, I don't necessarily have the inclination.
    - leetcrew 8y ago
      > We tend to prefer shopping at places where we know the products we purchase will be of a certain standard. we don't just prefer; we pay to shop at places where we trust that the products will be of high quality (or at least authentic). no one i know around my age pays for any news whatsoever (including myself, admittedly). this should tell you something about how much we value quality news.
- snaky 8y ago
  > Do you have a better solution? There is a somewhat working solution, but it might be incompatible with American voter (for some time) https://www.scmp.com/news/china/poliitics/article/2162036/china-launches-piyao-platform-crack-down-online-rumours https://www.scmp.com/news/china/poliitics/article/2162036/ch...
toast0 8y ago
The IRS doesn't know your net worth. They know your income and deductions, mostly.
allthenews 8y ago
>AI is not good at subjective decisions of qualitative data That's exactly the core strength of AI. It is what differentiates AI from hard coded solutions. Estimations(subjective decisions) based on correlations in fuzzy (qualitative) data. In your example, you could take a set of data containing the net worths and other characteristics, e.g. birth zip code, spending habits, affiliations, and if there are any trends related to net worth, a properly architected neural network trained on the appropriate data could easily estimate subjectively on what amounts to qualitative data.
jlebar 8y ago
"Fake news" has lost all meaning in the intervening years, but at the time, it is hard to remember, we were concerned about essentially phishing websites that made themselves out to be (say) CNN, but were not. As others have said, evaluating trust on the internet using computers (call it AI or not) is literally Google's core competency.

abtinf 8y ago

Short of launching a forensic investigation, subpoenaing records, and carefully watching you, how would the IRS know your net worth?

ehsankia 8y ago

Realistically there doesn't exist a good solution for this. That's why every single site out there with user content that is large enough is struggling with moderation. There is no solution that scales up to billions of users, and while it's true that AI most likely won't work, it is the best they've got right now. Do you have a better solution? Because Google has hired some of the smartest people and even then they still are having issues with Youtube every other week, so I'm sure you'd be paid a pretty hefty sum if you could solve this.

raxxorrax 8y ago
> Do you have a better solution? Let the receiver evaluate the information? Sure, some will have false believes. Cannot change that and you shouldn't even attempt to do so. Because authority has been wrong to an equal degree. The whole premise falls flat in my opinion. There won't be any viable solution, if the problem cannot even be defined. Currently it is based on a feeling that there are mean and false statements on social media. And this fact is emphasized by parties who like more control about content. Anyone aware of the current capabilities of AI should know, that it is no solution, at least in its current state. Sure, big tech likes to signal otherwise. But these statements just underline their business interest in that sector. Nothing wrong with that. Wanting to tighten control about content is.
- toofy 8y ago
  > Let the receiver evaluate the information This ignores one of Google's primary missions. As jlebar posted in a different comment > evaluating trust on the internet using computers (call it AI or not) is literally Google's core competency. This idea that we should pretend as if there is no problem to solve seems incredibly strange to me. Particularly when we know unreliable information is being weaponized by bad actors to influence entire populations of people. This is a very real problem which needs solving. Humans have limited time for research and we value reliability. Consider all of the hype around the recent severe decline in trust of authenticity and reliability of products purchased through Amazon. We tend to prefer shopping at places where we know the products we purchase will be of a certain standard. We go to trusted sources (as amazon was previously) because we don't have the time to carefully evaluate each of the many many items we purchase every week. We choose a store we trust will have already done that quality check. It seems as if Google believes it is important for it's business that the information it prioritizes has a certain quality or reliability level. Google prefers their information be more like a Target store than a back alley tent. And if the amount of people who shop at Target stores over back alley tents is any indicator, I'm guessing Google is probably on to something. I'm certainly skeptical about who should be able to decide what is "true" on subjective issues, but I don't see how we can pretend as if there isn't a very real problem of actual provably false information being weaponized. I for one am fairly busy and just as I don't have time to do a deep dive of research into whether or not the shampoo, deodorant, milk, cereal, orange juice, whiskey, dish soap, cold medicine etc etc etc are fake; I also don't have time to deep dive research if the 30+ articles I skim everyday are provable outright lies. Not only don't I have the time, I don't necessarily have the inclination.
  - leetcrew 8y ago
    > We tend to prefer shopping at places where we know the products we purchase will be of a certain standard. we don't just prefer; we pay to shop at places where we trust that the products will be of high quality (or at least authentic). no one i know around my age pays for any news whatsoever (including myself, admittedly). this should tell you something about how much we value quality news.
snaky 8y ago
> Do you have a better solution? There is a somewhat working solution, but it might be incompatible with American voter (for some time) https://www.scmp.com/news/china/poliitics/article/2162036/china-launches-piyao-platform-crack-down-online-rumours https://www.scmp.com/news/china/poliitics/article/2162036/ch...

toast0 8y ago

The IRS doesn't know your net worth. They know your income and deductions, mostly.

allthenews 8y ago

>AI is not good at subjective decisions of qualitative data That's exactly the core strength of AI. It is what differentiates AI from hard coded solutions. Estimations(subjective decisions) based on correlations in fuzzy (qualitative) data. In your example, you could take a set of data containing the net worths and other characteristics, e.g. birth zip code, spending habits, affiliations, and if there are any trends related to net worth, a properly architected neural network trained on the appropriate data could easily estimate subjectively on what amounts to qualitative data.

jlebar 8y ago

"Fake news" has lost all meaning in the intervening years, but at the time, it is hard to remember, we were concerned about essentially phishing websites that made themselves out to be (say) CNN, but were not. As others have said, evaluating trust on the internet using computers (call it AI or not) is literally Google's core competency.