9 ms·

This article is mostly a straw man, while still containing some valid ML criticism. I am a ML s(c|k)eptic too, in that popular conceptions of what ML is current

This article is mostly a straw man, while still containing some valid ML criticism. I am a ML s(c|k)eptic too, in that popular conceptions of what ML is currently overpromise, often don't even understand what ML actually is, and are often just some layperson's imagination about what "artificial intelligence" might do. This article is the opposite. He's treating ML as basically a simple supervised architecture that doesn't allow any domain knowledge to be incorporated and simply dead-reckons, making unchecked inferences from what it learned in training. Under these constraints, everything he says is correct. But there is no reason ML has to be used this way, in fact it is extremely irresponsible to do so in many cases. ML as part of a system (whether directly part of the model architecture and learned or imposed by domain knowledge) is possible, and is generally the right way to build an "AI" system. I think ML has its limitations and will be surprised to see current neural networks evolve into AGI. But I also don't think the engineers working in this space are as out to lunch as the author seems to imply, and would not write off the possibilities of what contemporary ML systems can accomplish based on the flaws pointed out in relation to a very narrow view of what ML is.

3gg 5y ago
> But I also don't think the engineers working in this space are as out to lunch as the author seems to imply. Are you at all close to this space? It sounds you may be underestimating corporate politics and the lack of rigour and ethical thought with which these systems are applied. The example Cory puts on policing -- and the many other examples you can find in Evgeny Morozov's book or "The End of Trust" -- are solid proof of this.
- vletal 5y ago
  My first thought was that I'm not the target audience of this article. I'm a ML practitioner. This seems more like an overstated opinionated wake up call to mgmt and sales people. Is not it?
  - version_five 5y ago
    Agreed. What I called a straw man in the OP could also be characterized as a simplification to get his point across to lay-audiences. (Personally I dont agree with the simplification, per my other post). It's meant for popular audiences (as someone else points out, this is from a sci-fi magazine)
  - 3gg 5y ago
    If you are an ML practitioner and you think you're not part of the target audience, then you're probably part of the target audience.
- foobiekr 5y ago
  There are three entirely different groups at work here. The deepmind team etc type of group who actually know what they’re doing and the boundaries of what they are working with the “AI-washing” startups, corporate groups who know they are faking it and that what they’re doing is extremely limited the corporate project team types who are just doing random tool play and honestly don’t understand what they are doing or that they are absolutely clueless with no self-awareness at all I’ve worked with all three and they really are just totally different things that are all being lumped together. They also are listed in terms of increasing proportion. For every self-aware AI-washer team I’ve seen 50 “we are doing AI” Corp team types spinning out one trivial demo after another to execs who know zero.
  - 3gg 5y ago
    Where does Google Vision Cloud sit in your categorization? https://algorithmwatch.org/en/google-vision-racism/ https://algorithmwatch.org/en/google-vision-racism/
    - foobiekr 5y ago
      First group. You’re observing that they aren’t doing a perfect job, which is true, but my grouping isn’t related to perfection of results.
      - 3gg 5y ago
        > The deepmind team etc type of group who actually know what they’re doing and the boundaries of what they are working with. You claim that they "know what they are doing and the boundaries of what they are working with" -- and yet they recklessly make public a racist vision product?
        foobiekr 5y ago
        Your argument is that knowing what you are doing means error free output.
- bhntr3 5y ago
  > Are you at all close to this space? I am. > The example Cory puts on policing My most upvoted comment on this website was discussing this exact scenario. https://news.ycombinator.com/item?id=23655487 https://news.ycombinator.com/item?id=23655487 Could you perhaps clarify the generalization you're making about me and people like me so I can understand it?
  - 3gg 5y ago
    Excellent. One problem in my mind that I don't see discussed enough -- and also not in your other post -- is that there is a large divide between those who use the technology (the cops in this case) and those who supply it, and there is no accountability in any of the two groups when something goes wrong. Like you write in your other post, "the system works (according to an objective function which maximizes arrests.)", and that is as far as the engineer goes. On the other hand, the cop picks up the technology and blindly applies it. To make any improvement to the system would require both groups to work together, but as far as I know, that is not happening. A recent example can be found in the adventures of Clearview AI. So from that perspective, I do think that the engineers (and the cops, and everybody else) are out to lunch, each doing their own work in a bubble and not paying enough attention to (or caring about) the side effects of the applications of this technology. Also, the lack of thought and accountability that I mention above I think is fairly general from my experience, even outside of policing. That is why I don't generally agree with the lunch statement. Guys are having a hell of a party as far as I can tell -- at the expense of horror stories suffered by the victims of these systems.
    - salawat 5y ago
      I second this. I spend a great deal of time digging through where we've positioned big data models to steer population scale behavior, and very infrequently do the implementers of the system ever stop to analyze the changes they are seeding or think beyond the first or second degree consequences once things take off. That is all part of engineering to me, so by definition, I think many in the field are in fact, out to lunch.
      - 5y ago
coding123 5y ago
That's how I felt too. Most of the article is trying to pull us with an emotional attachment (mostly to racist things a computer will do if tasked to do important things). While that criticism is welcome, it's not specifically meaningful towards an argument against AGI. The only part that was seemed to be that statistical inference is not a path to AGI which is somehow backed up by the emotional stuff. What deep learning seems to step into more and more is time-based statistical inference. AGI is not: seeing that a girl has a frown on their face. seeing that a girl has a frown, because someone said "you look fat" seeing that a girl has a frown because her boyfriend said you look fat seeing that Maya has generally been upset with her boyfriend who also most recently told her she is fat. But keep going and going and going and we might get somewhere. Do we have the computer power to keep going? I don't know.
- salawat 5y ago
  AGI is that capability to orchestrate layering of topical filters and feature detections in order to create an actionable perception. Note that it isn't anything to do with the implementations of said filters and detectors, but with the ability to artistically arrange them to satisfy a goal, and very possibly, must be coupled with the capacity to synthesize new ones. That executive and arranging function is the unknown. From whence cometh that characteristic of Dasein? That preponderance of concern with the act of being as Being? It's a tough nut to crack, even in philosophical circles. To think that we're going to articially create it by any means other than accident or luck is hubris of the highest order.
mistrial9 5y ago
I like your comment here starting with "straw man" .. and agree with some of the statements.. I have seen lengthy, detailed and authoritative reports that say some of the same things, but in a formal, long-winded way with more added.. This meta-comment of restatement in various contexts, with various amounts of story-telling and technical detail, brings up the educational burdens of communication -- to be effective you have to reach a reader where there are today .. in terms of assumptions, technical learning, and focus of topic.. since this is such a fast-moving and wide subject area, its super easy to miss the distinction between "low value, high volume audio clips recognition" and "life and death medical diagnosis for less than 100 patients". hint - that matters a lot in the tech chain AND the legal structure, and therefore combined, the "do-ability"
ziggus 5y ago
Agreed. The article reminds me of the arguments that religious fundamentalists make against evolution: "there are still monkeys, so how could it be that we evolved from monkeys, wouldn't all the monkeys have evolved as well?" Clearly, no biologist claims that humans evolved from modern primates, just like no modern AI researcher seriously thinks that current machine learning methods will lead to "True AI".
karaterobot 5y ago
> This article is mostly a straw man, while still containing some valid ML criticism. I don't think this is an example of a straw man, given that his audience is readers of Locus, a science fiction magazine. While researchers and practitioners in ML understandably hold a more nuanced, informed view, the position he's arguing against is pretty common among the general public, and certainly common in science fiction.
zwaps 5y ago
The author conflates AI and statistical, in particular, causal inference. A ML model and most of what we currently call AI, is a statistical model. It predicts things, and in the process of "training" this model we can also learn about the world it interacts with. Anyone who has ever spend any amount of time on the issue knows that the idea of "generalizing" or "adapting"in ML corresponds directly to causal inference. This is not some secret to be uncovered by every other pundit, it's a direct consequence of ML models being.. statistical models. If our AI can understand when statistical relationships hold and when they do not, then it is inferring causality from data. Currently, we are dealing with ML algorithms adapting or generalizing: learning which regularities to "trust" and which to "discard" when the "DGP" changes. This sounds nebulous and difficult, but that's only the case because the models are complex. Nevertheless, the underlying statistical problem is old. Ancient. Talked to death in every scientific field that tries to do an experiment, to infer some causal parameter from observational data, or construct counterfactuals to guide policy. Similarly, the divide between quantitative and qualitative approaches, are decades (if not centuries) old. These problems are so well understood that we can state very precisely what needs to happen to make our inference causal. The catch is, that first it depends on the DGP and our model thereof, and further that none of these things can be proven to be true in the same framework. Whether the DGP is as we need it to be, or whether the model identifies what we want, is something we can be confident about given a set of assumptions, but it is not something we can know to be true. And guess what, if the model is complex, then so is thinking about its inferential capabilities. The discussion seems tired, because it is. It's not a deep philosophical issue. It's a practical one. That doesn't imply there is a good solution to it either, but the basic issue hasn't changed for a long time. Maybe we are not trying to predict a treatment's effect on some randomly selected population, maybe we are instead trying to land a plane under conditions we can barely foresee. But what needs to happen for these two predictions to be unbiased, consistent, low variance, whatever... has not changed. Researchers are frustrated with these article, because pundit after pundit claims to uncover some general problem with AI or ML in face of reality. But it is not about AI. It's about statistics, and we all want to say: "Yes, we know. Now what?"

3gg 5y ago

> But I also don't think the engineers working in this space are as out to lunch as the author seems to imply. Are you at all close to this space? It sounds you may be underestimating corporate politics and the lack of rigour and ethical thought with which these systems are applied. The example Cory puts on policing -- and the many other examples you can find in Evgeny Morozov's book or "The End of Trust" -- are solid proof of this.

vletal 5y ago
My first thought was that I'm not the target audience of this article. I'm a ML practitioner. This seems more like an overstated opinionated wake up call to mgmt and sales people. Is not it?
- version_five 5y ago
  Agreed. What I called a straw man in the OP could also be characterized as a simplification to get his point across to lay-audiences. (Personally I dont agree with the simplification, per my other post). It's meant for popular audiences (as someone else points out, this is from a sci-fi magazine)
- 3gg 5y ago
  If you are an ML practitioner and you think you're not part of the target audience, then you're probably part of the target audience.
foobiekr 5y ago
There are three entirely different groups at work here. The deepmind team etc type of group who actually know what they’re doing and the boundaries of what they are working with the “AI-washing” startups, corporate groups who know they are faking it and that what they’re doing is extremely limited the corporate project team types who are just doing random tool play and honestly don’t understand what they are doing or that they are absolutely clueless with no self-awareness at all I’ve worked with all three and they really are just totally different things that are all being lumped together. They also are listed in terms of increasing proportion. For every self-aware AI-washer team I’ve seen 50 “we are doing AI” Corp team types spinning out one trivial demo after another to execs who know zero.
- 3gg 5y ago
  Where does Google Vision Cloud sit in your categorization? https://algorithmwatch.org/en/google-vision-racism/ https://algorithmwatch.org/en/google-vision-racism/
  - foobiekr 5y ago
    First group. You’re observing that they aren’t doing a perfect job, which is true, but my grouping isn’t related to perfection of results.
    - 3gg 5y ago
      > The deepmind team etc type of group who actually know what they’re doing and the boundaries of what they are working with. You claim that they "know what they are doing and the boundaries of what they are working with" -- and yet they recklessly make public a racist vision product?
      - foobiekr 5y ago
        Your argument is that knowing what you are doing means error free output.
bhntr3 5y ago
> Are you at all close to this space? I am. > The example Cory puts on policing My most upvoted comment on this website was discussing this exact scenario. https://news.ycombinator.com/item?id=23655487 https://news.ycombinator.com/item?id=23655487 Could you perhaps clarify the generalization you're making about me and people like me so I can understand it?
- 3gg 5y ago
  Excellent. One problem in my mind that I don't see discussed enough -- and also not in your other post -- is that there is a large divide between those who use the technology (the cops in this case) and those who supply it, and there is no accountability in any of the two groups when something goes wrong. Like you write in your other post, "the system works (according to an objective function which maximizes arrests.)", and that is as far as the engineer goes. On the other hand, the cop picks up the technology and blindly applies it. To make any improvement to the system would require both groups to work together, but as far as I know, that is not happening. A recent example can be found in the adventures of Clearview AI. So from that perspective, I do think that the engineers (and the cops, and everybody else) are out to lunch, each doing their own work in a bubble and not paying enough attention to (or caring about) the side effects of the applications of this technology. Also, the lack of thought and accountability that I mention above I think is fairly general from my experience, even outside of policing. That is why I don't generally agree with the lunch statement. Guys are having a hell of a party as far as I can tell -- at the expense of horror stories suffered by the victims of these systems.
  - salawat 5y ago
    I second this. I spend a great deal of time digging through where we've positioned big data models to steer population scale behavior, and very infrequently do the implementers of the system ever stop to analyze the changes they are seeding or think beyond the first or second degree consequences once things take off. That is all part of engineering to me, so by definition, I think many in the field are in fact, out to lunch.
    - 5y ago

coding123 5y ago

That's how I felt too. Most of the article is trying to pull us with an emotional attachment (mostly to racist things a computer will do if tasked to do important things). While that criticism is welcome, it's not specifically meaningful towards an argument against AGI. The only part that was seemed to be that statistical inference is not a path to AGI which is somehow backed up by the emotional stuff. What deep learning seems to step into more and more is time-based statistical inference. AGI is not: seeing that a girl has a frown on their face. seeing that a girl has a frown, because someone said "you look fat" seeing that a girl has a frown because her boyfriend said you look fat seeing that Maya has generally been upset with her boyfriend who also most recently told her she is fat. But keep going and going and going and we might get somewhere. Do we have the computer power to keep going? I don't know.

salawat 5y ago
AGI is that capability to orchestrate layering of topical filters and feature detections in order to create an actionable perception. Note that it isn't anything to do with the implementations of said filters and detectors, but with the ability to artistically arrange them to satisfy a goal, and very possibly, must be coupled with the capacity to synthesize new ones. That executive and arranging function is the unknown. From whence cometh that characteristic of Dasein? That preponderance of concern with the act of being as Being? It's a tough nut to crack, even in philosophical circles. To think that we're going to articially create it by any means other than accident or luck is hubris of the highest order.

mistrial9 5y ago

I like your comment here starting with "straw man" .. and agree with some of the statements.. I have seen lengthy, detailed and authoritative reports that say some of the same things, but in a formal, long-winded way with more added.. This meta-comment of restatement in various contexts, with various amounts of story-telling and technical detail, brings up the educational burdens of communication -- to be effective you have to reach a reader where there are today .. in terms of assumptions, technical learning, and focus of topic.. since this is such a fast-moving and wide subject area, its super easy to miss the distinction between "low value, high volume audio clips recognition" and "life and death medical diagnosis for less than 100 patients". hint - that matters a lot in the tech chain AND the legal structure, and therefore combined, the "do-ability"

ziggus 5y ago

Agreed. The article reminds me of the arguments that religious fundamentalists make against evolution: "there are still monkeys, so how could it be that we evolved from monkeys, wouldn't all the monkeys have evolved as well?" Clearly, no biologist claims that humans evolved from modern primates, just like no modern AI researcher seriously thinks that current machine learning methods will lead to "True AI".

karaterobot 5y ago

> This article is mostly a straw man, while still containing some valid ML criticism. I don't think this is an example of a straw man, given that his audience is readers of Locus, a science fiction magazine. While researchers and practitioners in ML understandably hold a more nuanced, informed view, the position he's arguing against is pretty common among the general public, and certainly common in science fiction.

zwaps 5y ago

The author conflates AI and statistical, in particular, causal inference. A ML model and most of what we currently call AI, is a statistical model. It predicts things, and in the process of "training" this model we can also learn about the world it interacts with. Anyone who has ever spend any amount of time on the issue knows that the idea of "generalizing" or "adapting"in ML corresponds directly to causal inference. This is not some secret to be uncovered by every other pundit, it's a direct consequence of ML models being.. statistical models. If our AI can understand when statistical relationships hold and when they do not, then it is inferring causality from data. Currently, we are dealing with ML algorithms adapting or generalizing: learning which regularities to "trust" and which to "discard" when the "DGP" changes. This sounds nebulous and difficult, but that's only the case because the models are complex. Nevertheless, the underlying statistical problem is old. Ancient. Talked to death in every scientific field that tries to do an experiment, to infer some causal parameter from observational data, or construct counterfactuals to guide policy. Similarly, the divide between quantitative and qualitative approaches, are decades (if not centuries) old. These problems are so well understood that we can state very precisely what needs to happen to make our inference causal. The catch is, that first it depends on the DGP and our model thereof, and further that none of these things can be proven to be true in the same framework. Whether the DGP is as we need it to be, or whether the model identifies what we want, is something we can be confident about given a set of assumptions, but it is not something we can know to be true. And guess what, if the model is complex, then so is thinking about its inferential capabilities. The discussion seems tired, because it is. It's not a deep philosophical issue. It's a practical one. That doesn't imply there is a good solution to it either, but the basic issue hasn't changed for a long time. Maybe we are not trying to predict a treatment's effect on some randomly selected population, maybe we are instead trying to land a plane under conditions we can barely foresee. But what needs to happen for these two predictions to be unbiased, consistent, low variance, whatever... has not changed. Researchers are frustrated with these article, because pundit after pundit claims to uncover some general problem with AI or ML in face of reality. But it is not about AI. It's about statistics, and we all want to say: "Yes, we know. Now what?"